CN105338352B

CN105338352B - Adaptive inverse quantization method and device in video coding

Info

Publication number: CN105338352B
Application number: CN201410355790.5A
Authority: CN
Inventors: 赵寅; 杨海涛; 吕卓逸
Original assignee: Huawei Technologies Co Ltd
Current assignee: Huawei Technologies Co Ltd
Priority date: 2014-07-24
Filing date: 2014-07-24
Publication date: 2019-04-19
Anticipated expiration: 2034-07-24
Also published as: JP6388355B2; KR101893049B1; EP3163875A1; EP3163875A4; CN105338352A; EP3163875B1; KR20170031202A; JP2017527181A; US20170127062A1; US10257514B2; WO2016011796A1

Abstract

The invention discloses quantification method adaptive in a kind of Video coding and devices, method provided by the invention utilizes space neighborhood information, time neighborhood information or the space neighborhood information and time neighborhood information of transform block, estimate the statistical property of background area locating for Current Transform block, the quantization regulatory factor of inverse quantization is adaptively exported, inverse quantization processing is neatly adjusted.Method and apparatus provided by the invention do not need to provide additional overhead bit and transmit to quantify to adjust information compared to the existing scheme of the transmission quantization adjusting information in code stream, therefore further improve code efficiency.

Description

Adaptive quantification method and device in a kind of Video coding

Technical field

The present invention relates to quantification method adaptive in data processing field more particularly to a kind of Video coding and devices.

Background technique

Current video coding techniques includes a variety of, for example, H.264/AVC, H.265/HEVC, audio/video encoding standard Video encoding standards such as (Audio Video coding Standard, AVS), above-mentioned video encoding standard is usually all using mixed Coding framework is closed, which mainly includes the following links:

Predict (prediction), transformation (transform), quantization (quantization), entropy coding (entropy The links such as coding).

Predict that link generates present encoding block pair using the reconstruction pixel (reconstructed pixel) in encoded region The prediction pixel (predicted pixel) for the original pixels (original pixel) answered.Prediction mode mainly includes in frame Predict (intra prediction) and inter-prediction (inter prediction) two major classes.Intra prediction uses present encoding The reconstruction pixel of block spatial neighborhood in current encoded image generates the prediction pixel of present encoding block, such as H.264/AVC In level (horizontal), the multi-direction prediction such as vertical (vertical) and be based on stencil matching (template ) and the prediction mode of intra frame motion compensation (intra motion compensation) matching.Inter-prediction uses current Encoding block corresponding prediction pixel for rebuilding pixel as present encoding block in one or more encoded images, such as based on The prediction of motion compensation (motion compensation)；Inter-prediction includes single directional prediction and both bi-directional predicted forms, Wherein single directional prediction generates the prediction pixel of present encoding block using the reconstruction pixel in an encoded image, and bi-directional predicted The prediction pixel of present encoding block is generated using the reconstruction pixel in two encoded images.

Value differences between original pixels and prediction pixel are known as residual error (residual).In order to improve the volume of residual error Code efficiency usually first converts residual error, is translated into transformation coefficient (transform coefficient), commonly Transformation has discrete cosine transform (Discrete Cosine Transform, DCT), discrete sine transform (Discrete Sine Transform, DST), wavelet transformation etc..Later, quantification treatment is made to transformation coefficient, such as utilizes vector quantization or scalar quantity Change.Then, by the transformation coefficient and coding mode information (such as coded block size, prediction mode, motion vector after quantization Deng) by entropy coding handle be converted into code stream.Common entropy coding method has arithmetic coding (arithmetic coding), becomes Long codes (variable length coding, VLC), block code, Run- Length Coding (run-length coding) etc..

The quantization of transformation coefficient can use scalar quantization mode.Remember i-th in N number of transformation coefficient of present encoding block residual error A transformation coefficient is C (i) (1≤i≤N, N are related with the size of transform block, usually 16,64,1024 etc.), then the change after quantifying Changing coefficient Q (i) is

Wherein sign { X } indicates to take the symbol of X, that is, hasRound { X } is to be rounded One of (rounding) it operates, usually can be downward rounding, round, round up；| X | expression takes X's Absolute value or amplitude；Qs (i) indicates transformation coefficient C (i) corresponding quantization step (quantization step)；O1 (i) is house Enter biasing (rounding offset).

Video decoding be the process that code stream is converted to video image, it include entropy decoding (entropy decoding), Several key links such as prediction, inverse quantization (dequantization), inverse transformation (inverse transform).Firstly, by code Stream parses the transformation coefficient after coding mode information and quantization by entropy decoding process.Then, on the one hand believed by coding mode Breath and decoded reconstruction pixel obtain prediction pixel；On the other hand the transformation coefficient after quantization is obtained into weight by inverse quantization The transformation coefficient built, then inverse transformation, the residual information rebuild are carried out to the transformation coefficient of reconstruction.Later, by the residual of reconstruction Poor information is added with prediction pixel, obtains rebuilding pixel, to recover video image.

The transformation coefficient R (i) that inverse quantization is rebuild is carried out to the transformation coefficient Q (i) after quantization, can be described as:

R (i)=sign { Q (i) } round { Q (i) Qs (i)+o2 (i) } ... ... ... (formula 1)

Wherein Qs (i) can be floating number, and o2 (i) is rounding-off biasing.In general, being adopted in order to avoid using floating point arithmetic With addition of integer and the mode of displacement come approximate substitution floating number multiplication, such as the inverse for H.265/HEVC stating (formula 1) Change process is approximately:

R (i)=sign { Q (i) } (Q (i) Qs'(i)+(1<<(bdshift-1)))>>bdshift ... ... (public affairs Formula 2)

Wherein, bdshift is shift parameters, Qs'(i) be integer, Qs'(i)/2^bdshiftThe quantization being similar in (1) formula Step-length Qs (i), o2 (i)=0.5, the mode that is rounded are to be rounded downwards at this time.Qs'(i) by level scale (levelScale) l (i) It is codetermined with zoom factor (scaling factor) m (i),

Qs'(i)=m (i) l (i) ... ... ... ... ... ... (formula 3)

And l (i) is the function of quantization parameter (Quantization Parameter, QP), that is, is had

... ... ... ... (formula 4)

Wherein, level scale table levelScale [k]={ 40,45,51,57,64,72 }, k=0,1 ..., 5；It indicates to QP (i) except 6 are rounded, % is remainder operation.

Generally speaking, inverse quantization is directly related with quantization step, and quantization step is then by QP, zoom factor, level ruler Spend the influence of table.Adjusting quantization step can be realized by various ways.For example, when zoom factor, level scale table are fixed, often H.264/AVC and H.265/HEVC the corresponding quantization step of the quantization parameter of a rank defines the quantization ginseng of 52 ranks Number, therefore quantization step can be adjusted by changing quantization parameter.In another example adjustable zoom factor walks to change quantization It is long, typically can choose multiple zoom factor matrixes, also referred to as quantization matrix, in one determine zoom factor.On although It is different to state the data that two examples change, but essence is adjusting quantization step.

For lossy coding, rebuild pixel with original pixels may be it is different, numerical value difference between the two is known as It is distorted (distortion).Due to the presence of a variety of visual masking effects, such as brightness masking effect and contrast masking sensitivity effect, The characteristic of background has close connection where eye-observation to the intensity and distortion of distortion.That is, the distortion of human eye is sensitive Degree is related with the background luminance of distortion position and background contrasts, and usually U is presented in distortion sensitivity and background luminance Shape curved line relation, and be in monotone decreasing relationship with the variance of background or standard deviation.It is special in conjunction with above-mentioned vision in Video coding Property, increase quantization step to insensitive region (i.e. the lesser region of distortion sensitivity) is visually distorted, sensitizing range is subtracted Small quantization step can make coding distortion distribution more meet human eye visual perception compared to unified quantization step is used, thus Subjective quality is promoted under same code rate, that is, improves code efficiency.

The method of the adjustment quantization step-length provided in the prior art are as follows:

Video sequence is analyzed in coding side, determines the corresponding quantization parameter of the transformation coefficient of each encoding block (QP), and will Offset (delta QP) the write-in code stream of quantization parameter or the quantization parameter relative to band level quantized parameter (slice QP)； Decoding end carries out corresponding quantization step adjusting according to the quantization parameter being resolved to each encoding block.

Above-mentioned prior art, which transmits quantified parameter information through coding side decision quantization parameter and in code stream, to be made to solve Code end knows that the quantization step of each encoding block adjusts numerical value, to realize that adaptive quantizing step-length is adjusted.But quantization parameter pair The side information (side information) answered limits the promotion of code efficiency to a certain extent.

Summary of the invention

The present invention provides adaptive quantification method and device, method provided by the present invention and dress in a kind of Video coding It sets and solves the problems, such as that the corresponding side information of quantization parameter limits the promotion of code efficiency to a certain extent in the prior art.

In a first aspect, adaptive quantification method in a kind of Video coding is provided, this method comprises:

Determine the first transformation series manifold corresponding image-region in currently decoding image；The first transformation series manifold packet N number of transformation coefficient is included, the transformation coefficient is the transformation coefficient of any color space component in the current decoding image, In, N is positive integer；

The decoded pixel information in described image region is obtained, the decoded pixel information includes the space in described image region Neighborhood territory pixel information；

Quantization regulatory factor is calculated according to the decoded pixel information；

Inverse quantization is carried out to the first transformation series manifold according to the quantization regulatory factor, obtains the second transformation coefficient Collection.

With reference to first aspect, in the first possible implementation, the decoded pixel information further includes described image The time neighborhood territory pixel information in region and/or the Pixel Information in described image region.

Any one with reference to first aspect or in the first possible implementation of first aspect, at second In possible implementation, according to the decoded pixel information, calculating quantization regulatory factor includes:

According to pixel mean value P_avgFirst function f₁(P_avg) calculate quantization regulatory factor；Wherein, the pixel mean value P_avg It is the mean value or the multiple picture that the luminance component of multiple pixels corresponding to the decoded pixel information is averaging The mean value that the same color space component of element is averaging, the color space component are the first transformation series manifold pair The color space component answered.

Any one with reference to first aspect or in the first possible implementation of first aspect, at the third In possible implementation, according to the decoded pixel information, calculating quantization regulatory factor includes:

According to pixel variance P_varSecond function f₂(P_var) calculate quantization regulatory factor；Wherein, the pixel variance P_var For the variance of the luminance component of the corresponding multiple pixels of the decoded pixel information, or the same color for the multiple pixel The variance of color space component, the color space component are the corresponding color space component of the first transformation series manifold.

Any one with reference to first aspect or in the first possible implementation of first aspect, at the 4th kind In possible implementation, according to the decoded pixel information, calculating quantization regulatory factor includes:

According to pixel variance P_varWith pixel mean value P_avgCopula f₁(P_avg)·f₂(P_var) calculate quantization adjust because Son；Wherein, the pixel mean value P_avgBe the luminance component of multiple pixels in the decoded pixel information is averaging it is equal Value, or the mean value that is averaging of the same color space component for the multiple pixel；The pixel variance P_varFor institute The variance of the luminance component of multiple pixels, or the variance of the same color space component for the multiple pixel are stated, it is described Color space component is the corresponding color space component of the first transformation series manifold.

Any one with reference to first aspect or in the first possible implementation of first aspect, at the 5th kind In possible implementation, according to the decoded pixel information, calculating quantization regulatory factor includes:

According to pixel maximum P_maxWith pixel minimum P_minDifference third function f₃(P_max-P_min) calculate quantization tune Save the factor；Wherein, the pixel maximum P_maxFor the maximum of the luminance component of the corresponding multiple pixels of the decoded pixel information The maximum value of value or the same color space component of the multiple pixel；The pixel minimum is the multiple pixel Luminance component minimum value, or be the minimum value of the same spatial component of the multiple pixel, the color space component For the corresponding color space component of the first transformation series manifold.

With reference to first aspect or any one into the 5th kind of possible implementation of first aspect first, It is described that the first transformation series manifold is carried out at inverse quantization according to the quantization regulatory factor in six kinds of possible implementations Reason, obtaining the second transformation series manifold includes:

The corresponding target level scale table of the quantization regulatory factor is determined according to preset first corresponding relationship；Wherein, First corresponding relationship is the corresponding relationship for quantifying regulatory factor and level scale table；

According to the first transformation series manifold described in the target level scale table inverse quantization, second transformation coefficient is obtained Collection.

With reference to first aspect or any one into the 5th kind of possible implementation of first aspect first, It is described that the first transformation series manifold is carried out at inverse quantization according to the quantization regulatory factor in seven kinds of possible implementations Reason, obtaining the second transformation series manifold includes:

The corresponding Target quantization matrix of the quantization regulatory factor is determined according to preset second corresponding relationship；Wherein, institute Stating the second corresponding relationship is the corresponding relationship for quantifying regulatory factor and quantization matrix；

The first transformation series manifold according to the Target quantization matrix inverse quantization, obtains the second transformation series manifold.

With reference to first aspect or any one into the 5th kind of possible implementation of first aspect first, It is described that the first transformation series manifold is carried out at inverse quantization according to the quantization regulatory factor in eight kinds of possible implementations Reason, obtaining the second transformation series manifold includes:

The corresponding first quantization ginseng of the first transformation series manifold is obtained from signal corresponding to the current decoding image Number；

Target quantization parameter offset is determined according to the quantization regulatory factor, by first quantization parameter and the mesh Scalarization parameter shift amount is added to obtain the second quantization parameter；

Level scale is determined according to second quantization parameter, utilizes the first transformation series described in the level scale inverse quantization Manifold obtains the second transformation series manifold.

With reference to first aspect or any one into the 5th kind of possible implementation of first aspect first, It is described that the first transformation series manifold is carried out at inverse quantization according to the quantization regulatory factor in nine kinds of possible implementations Reason, obtaining the second transformation series manifold includes:

The corresponding first quantization step of the first transformation series manifold is obtained from signal corresponding to the current decoding image It is long；

First quantization step is zoomed in and out using the quantization regulatory factor, obtains the second quantization step, is utilized First transformation series manifold described in the second quantization step inverse quantization, obtains the second transformation series manifold.

With reference to first aspect or any one into the 5th kind of possible implementation of first aspect first, It is described that the first transformation series manifold is carried out at inverse quantization according to the quantization regulatory factor in ten kinds of possible implementations Reason, obtaining the second transformation series manifold includes:

Inverse quantization is carried out to the coefficient that first transformation coefficient is concentrated using first quantization step and obtains third change Change the coefficient in coefficient set；

It is zoomed in and out using the coefficient that the quantization regulatory factor being calculated concentrates the third transformation coefficient To the second transformation series manifold.

Second aspect provides adaptive quantification method in a kind of Video coding, this method comprises:

Determine the first transformation series manifold corresponding image-region X in currently decoding image；The first transformation series manifold Including N number of transformation coefficient, the transformation coefficient is the transformation coefficient of any color space component in the current decoding image, In, N is positive integer；

Obtain the time neighborhood territory pixel information of described image region X；

According to the first mean value P_avgWith first variance P_varCalculate quantization regulatory factor, wherein the first mean value P_avgIt is pair The mean value that the luminance component of multiple pixels is averaging in the decoded pixel information, or be the same of the multiple pixel The mean value that a color space component is averaging；The first variance P_varFor the variance of the luminance component of the multiple pixel, Or the variance of the same color space component for the multiple pixel, the color space component are first transformation series The corresponding color space component of manifold；

Inverse quantization processing is carried out to the first transformation series manifold according to the quantization regulatory factor, obtains the second transformation series Manifold.

In conjunction with second aspect, in the first possible implementation, according to the first mean value P_avgWith first variance P_varMeter Calculating quantization regulatory factor includes:

According to the first mean value P_avgWith the first variance P_varCopula f₁(P_avg)·f₂(P_var) calculate quantization Regulatory factor.

In conjunction with any one in the possible implementation of the first of second aspect or second aspect, at second It is described that inverse quantization processing is carried out to the first transformation series manifold according to the quantization regulatory factor in possible implementation, Obtaining the second transformation series manifold includes:

In conjunction with any one in the possible implementation of the first of second aspect or second aspect, at the third It is described that inverse quantization processing is carried out to the first transformation series manifold according to the quantization regulatory factor in possible implementation, Obtaining the second transformation series manifold includes:

In conjunction with any one in the possible implementation of the first of second aspect or second aspect, at the 4th kind It is described that inverse quantization processing is carried out to the first transformation series manifold according to the quantization regulatory factor in possible implementation, Obtaining the second transformation series manifold includes:

In conjunction with any one in the possible implementation of the first of second aspect or second aspect, at the 5th kind It is described that inverse quantization processing is carried out to the first transformation series manifold according to the quantization regulatory factor in possible implementation, Obtaining the second transformation series manifold includes:

In conjunction with any one in the possible implementation of the first of second aspect or second aspect, at the 6th kind It is described that inverse quantization processing is carried out to the first transformation series manifold according to the quantization regulatory factor in possible implementation, Obtaining the second transformation series manifold includes:

The third aspect, provides adaptive inverse quantization device in a kind of Video coding, which includes:

Determination unit, for determining the first transformation series manifold corresponding image-region in currently decoding image；Described One transformation series manifold includes N number of transformation coefficient, and the transformation coefficient is any color space component in the current decoding image Transformation coefficient, wherein N is positive integer；

Acquiring unit, for obtaining the decoded pixel information in described image region, the decoded pixel information includes described The spatial neighborhood Pixel Information of image-region；

Computing unit, for calculating quantization regulatory factor according to the decoded pixel information；

Inverse quantization unit is obtained for carrying out inverse quantization to the first transformation series manifold according to the quantization regulatory factor To the second transformation series manifold.

In conjunction with the third aspect, in the first possible implementation, the acquiring unit is also used to obtain the decoding The time neighborhood territory pixel information in the described image region in Pixel Information and/or the Pixel Information in described image region.

In conjunction with any one in the possible implementation of the first of the third aspect or first aspect, at second In possible implementation, the computing unit is specifically used for according to pixel mean value P_avgFirst function f₁(P_avg) calculate quantization Regulatory factor；Wherein, the pixel mean value P_avgIt is that the luminance components of multiple pixels corresponding to the decoded pixel information asks flat The mean value that the same color space component of the mean value or the multiple pixel that obtain is averaging, the color Spatial component is the corresponding color space component of the first transformation series manifold.

In conjunction with any one in the possible implementation of the first of the third aspect or first aspect, at the third In possible implementation, the computing unit is specifically used for according to pixel variance P_varSecond function f₂(P_var) calculate quantization Regulatory factor；Wherein, the pixel variance P_varFor the side of the luminance component of the corresponding multiple pixels of the decoded pixel information Difference, or the variance of the same color space component for the multiple pixel, the color space component are first change Change the corresponding color space component of coefficient set.

In conjunction with any one in the possible implementation of the first of the third aspect or first aspect, at the 4th kind In possible implementation, the computing unit is specifically used for according to pixel variance P_varWith pixel mean value P_avgCopula f₁ (P_avg)·f₂(P_var) calculate quantization regulatory factor；Wherein, the pixel mean value P_avgIt is to multiple in the decoded pixel information The mean value that the luminance component of pixel is averaging, or be averaging for the same color space component of the multiple pixel The mean value arrived；The pixel variance P_varIt for the variance of the luminance component of the multiple pixel, or is the same of the multiple pixel The variance of one color space component, the color space component are the corresponding color space point of the first transformation series manifold Amount.

In conjunction with any one in the possible implementation of the first of the third aspect or first aspect, at the 5th kind In possible implementation, the computing unit is specifically used for according to pixel maximum P_maxWith pixel minimum P_minDifference Third function f₃(P_max-P_min) calculate quantization regulatory factor；Wherein, the pixel maximum P_maxFor the decoded pixel information The maximum value of the luminance component of corresponding multiple pixels or the maximum of the same color space component of the multiple pixel Value；The pixel minimum is the minimum value of the luminance component of the multiple pixel, or is the same sky of the multiple pixel Between component minimum value, the color space component be the corresponding color space component of the first transformation series manifold.

In conjunction with any one of the third aspect or first aspect first into the 5th kind of possible implementation, In six kinds of possible implementations, the inverse quantization unit is specifically used for determining the quantization according to preset first corresponding relationship The corresponding target level scale table of regulatory factor；Wherein, first corresponding relationship is quantization regulatory factor and level scale table Corresponding relationship；According to the first transformation series manifold described in the target level scale table inverse quantization, second transformation series is obtained Manifold.

In conjunction with any one of the third aspect or first aspect first into the 5th kind of possible implementation, In seven kinds of possible implementations, the inverse quantization unit is specifically used for determining the quantization according to preset second corresponding relationship The corresponding Target quantization matrix of regulatory factor；Wherein, second corresponding relationship is pair for quantifying regulatory factor and quantization matrix It should be related to；The first transformation series manifold according to the Target quantization matrix inverse quantization, obtains the second transformation series manifold.

In conjunction with any one of the third aspect or first aspect first into the 5th kind of possible implementation, In eight kinds of possible implementations, the inverse quantization unit is specifically used for obtaining from signal corresponding to the current decoding image Corresponding first quantization parameter of the first transformation series manifold；Determine that Target quantization parameter deviates according to the quantization regulatory factor Amount, obtains the second quantization parameter for first quantization parameter and the Target quantization parameter offset addition；According to described Two quantization parameters determine level scale, using the first transformation series manifold described in the level scale inverse quantization, obtain described second Transformation series manifold.

In conjunction with any one of the third aspect or first aspect first into the 5th kind of possible implementation, In nine kinds of possible implementations, the inverse quantization unit is specifically used for obtaining from signal corresponding to the current decoding image Corresponding first quantization step of the first transformation series manifold；Using the quantization regulatory factor to first quantization step into Row scaling, obtains the second quantization step, using the first transformation series manifold described in the second quantization step inverse quantization, obtains described Second transformation series manifold.

In conjunction with any one of the third aspect or first aspect first into the 5th kind of possible implementation, In ten kinds of possible implementations, the inverse quantization unit is specifically used for obtaining from signal corresponding to the current decoding image Corresponding first quantization step of the first transformation series manifold；Using first quantization step to the first transformation series manifold In coefficient carry out inverse quantization obtain third transformation coefficient concentration coefficient；Utilize the quantization regulatory factor pair being calculated The coefficient that the third transformation coefficient is concentrated zooms in and out to obtain the second transformation series manifold.

Fourth aspect, provides adaptive inverse quantization device in a kind of Video coding, which includes:

First determination unit, for determining the first transformation series manifold corresponding image-region X in currently decoding image；Institute Stating the first transformation series manifold includes N number of transformation coefficient, and the transformation coefficient is any color space in the current decoding image The transformation coefficient of component, wherein N is positive integer；

First acquisition unit, for obtaining the time neighborhood territory pixel information of described image region X；

First computing unit, for according to the first mean value P_avgWith first variance P_varCalculate quantization regulatory factor, wherein institute State the first mean value P_avgIt is the mean value being averaging to the luminance component of multiple pixels in the decoded pixel information, Huo Zhewei The mean value that the same color space component of the multiple pixel is averaging；The first variance P_varFor the multiple picture The variance of the luminance component of element, or the variance of the same color space component for the multiple pixel, the color space Component is the corresponding color space component of the first transformation series manifold；

First inverse quantization unit, for carrying out inverse quantization to the first transformation series manifold according to the quantization regulatory factor Processing, obtains the second transformation series manifold.

In conjunction with second aspect, in the first possible implementation, first computing unit is specifically used for according to institute State the first mean value P_avgWith the first variance P_varCopula f₁(P_avg)·f₂(P_var) calculate quantization regulatory factor.

In conjunction with any one in the possible implementation of the first of second aspect or second aspect, at second In possible implementation, first inverse quantization unit is specifically used for determining the quantization according to preset first corresponding relationship The corresponding target level scale table of regulatory factor；Wherein, first corresponding relationship is quantization regulatory factor and level scale table Corresponding relationship；According to the first transformation series manifold described in the target level scale table inverse quantization, second transformation series is obtained Manifold.

In conjunction with any one in the possible implementation of the first of second aspect or second aspect, at the third In possible implementation, first inverse quantization unit is specifically used for determining the quantization according to preset second corresponding relationship The corresponding Target quantization matrix of regulatory factor；Wherein, second corresponding relationship is pair for quantifying regulatory factor and quantization matrix It should be related to；The first transformation series manifold according to the Target quantization matrix inverse quantization, obtains the second transformation series manifold.

In conjunction with any one in the possible implementation of the first of second aspect or second aspect, at the 4th kind In possible implementation, first inverse quantization unit is specifically used for obtaining from signal corresponding to the current decoding image Corresponding first quantization parameter of the first transformation series manifold；Determine that Target quantization parameter deviates according to the quantization regulatory factor Amount, obtains the second quantization parameter for first quantization parameter and the Target quantization parameter offset addition；According to described Two quantization parameters determine level scale, using the first transformation series manifold described in the level scale inverse quantization, obtain described second Transformation series manifold.

In conjunction with any one in the possible implementation of the first of second aspect or second aspect, at the 5th kind In possible implementation, first inverse quantization unit is specifically used for obtaining from signal corresponding to the current decoding image Corresponding first quantization step of the first transformation series manifold；Using the quantization regulatory factor to first quantization step into Row scaling, obtains the second quantization step, using the first transformation series manifold described in the second quantization step inverse quantization, obtains described Second transformation series manifold.

In conjunction with any one in the possible implementation of the first of second aspect or second aspect, at the 6th kind In possible implementation, first inverse quantization unit is specifically used for obtaining from signal corresponding to the current decoding image Corresponding first quantization step of the first transformation series manifold；Using first quantization step to the first transformation series manifold In coefficient carry out inverse quantization obtain third transformation coefficient concentration coefficient；Utilize the quantization regulatory factor pair being calculated The coefficient that the third transformation coefficient is concentrated zooms in and out to obtain the second transformation series manifold.

One or two of above-mentioned technical proposal at least has the following technical effect that

Scheme provided in an embodiment of the present invention utilizes the space neighborhood information of transform block, estimates background locating for Current Transform block The statistical property in region adaptively exports the quantization regulatory factor of inverse quantization, neatly adjusts inverse quantization processing.The present invention is real The scheme adaptive quantizing method for applying example offer makes quantizing distortion intensity more meet human eye visual perception, to improve coding Efficiency.

Detailed description of the invention

Fig. 1 is the flow diagram of adaptive quantification method in a kind of Video coding provided in an embodiment of the present invention；

Fig. 2 is the schematic diagram of the first image-region spatial neighborhood pixel in the embodiment of the present invention；

Fig. 3 is the schematic diagram of second of image-region spatial neighborhood pixel in the embodiment of the present invention；

Fig. 4 is the process signal of adaptive quantification method in another Video coding provided in an embodiment of the present invention Figure；

Fig. 5 is the structural schematic diagram of adaptive inverse quantization device in a kind of Video coding provided in an embodiment of the present invention.

Specific embodiment

In the prior art, transmitting quantified parameter information through coding side decision quantization parameter and in code stream obtains decoding end Know that the quantization step of each encoding block adjusts numerical value, to realize that adaptive quantizing step-length is adjusted.But the corresponding side of quantization parameter Information (side information) limits the promotion of code efficiency to a certain extent.So for asking for technology is shown Topic, the embodiment of the present invention provide adaptive quantification method in a kind of Video coding, this method comprises:

Determine the first transformation series manifold corresponding image-region in currently decoding image；First transformation series manifold includes N A transformation coefficient, transformation coefficient are the transformation coefficient of any color space component in the current decoding image, wherein N is positive Integer；

The decoded pixel information of image-region is obtained, decoded pixel information includes the spatial neighborhood pixel in described image region Information；

Quantization regulatory factor is calculated according to decoded pixel information；

Inverse quantization is carried out to the first transformation series manifold according to quantization regulatory factor, obtains the second transformation series manifold.

Scheme provided in an embodiment of the present invention utilizes the space neighborhood information of transform block, estimates background locating for Current Transform block The statistical property in region, adaptively export quantization regulatory factor, neatly adjusts inverse quantization processing.On the one hand, of the invention Adaptive quantizing method makes quantizing distortion intensity more meet human eye visual perception, to improve code efficiency；

On the other hand compared in code stream transmission quantization adjust the existing scheme of information, provided by the embodiment of the present invention Scheme does not need additional overhead bit transmission quantization and adjusts information, therefore further improves code efficiency.

In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is A part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art Every other embodiment obtained without creative efforts, shall fall within the protection scope of the present invention.

Embodiment

As shown in Figure 1, the embodiment of the present invention provides adaptive quantification method in a kind of Video coding, this method comprises:

Step 101, the first transformation series manifold corresponding image-region in currently decoding image is determined；First transformation coefficient Collection includes N number of transformation coefficient, and transformation coefficient is the transformation coefficient of any color space component in the current decoding image, In, N is positive integer；

In embodiments of the present invention, the first transformation series manifold may include N number of transformation coefficient A (i), i=1,2 ..., N, N For positive integer, such as N=1,2,4,16,32,50,64,81,128,256,512 or 1024；Transformation coefficient A (i) is described current Decode the transformation coefficient of any color space component in image, such as the transformation series of the transformation coefficient of luminance component, chromatic component The transformation coefficient of number, any component of RGB (such as R component).

The corresponding image-region of first transformation series manifold is the first transformation series manifold corresponding area in currently decoding image Domain, for example, for H.264/AVC, H.265/HEVC under the corresponding hybrid encoding frame of standard, the first transformation series manifold is corresponding Image-region is the corresponding transform block of the first transformation series manifold (transform block) or converter unit (transform Unit) the image-region covered.In embodiments of the present invention, the first transformation series manifold may include the whole of a transform block and become Change coefficient；Also the DC coefficient in a part of transformation coefficient an of transform block, such as a transform block, Huo Zheyi be may include Group low frequency coefficient perhaps one group of high frequency coefficient perhaps DC coefficient and one group of high frequency coefficient or suitable according to a certain coefficient scanning Preceding several coefficients under sequence etc..

Step 102, the decoded pixel information in described image region is obtained, the decoded pixel information includes described image area The spatial neighborhood Pixel Information in domain；

In embodiments of the present invention, spatial neighborhood Pixel Information refers to described image region on same frame decoding image Pixel information, the position of adjacent pixel and range can be according to the different adjustment for carrying out adaptability of decoding condition.

Wherein spatial neighborhood Pixel Information may include: the reconstruction of at least one pixel in the spatial neighborhood of image-region X At least one intra prediction picture of value and/or the image-region X generated by the Pixel Information in the spatial neighborhood of image-region X Element.

When decoded pixel information is spatial neighborhood Pixel Information, spatial neighborhood Pixel Information may include M pixel, M For positive integer, spatial neighborhood Pixel Information includes at least one of following Pixel Information:

1) the pixel letter in the spatial neighborhood Y1 (several optional examples of spatial neighborhood Y1 are as shown in Figure 2) of image-region X Breath, the reconstructed value including at least one pixel in spatial neighborhood Y1；Wherein the concrete condition of spatial neighborhood Y1 may is that

Spatial neighborhood Y1 is the region adjacent at least one boundary in the upper and lower, left and right boundary of X, such as one and X The adjacent L shape region (a example and b example in such as Fig. 2) of left margin and coboundary or one and the region X coboundary, a left side Boundary or the adjacent rectangular area (c example, d example and e example in such as Fig. 2) of right margin or one include the rectangle of X The partial region (f example in such as Fig. 2) of X is removed in region.

2) at least one intra prediction of the image-region X generated by the Pixel Information in the spatial neighborhood of image-region X Pixel；It specifically includes:

By intra prediction mode, by the spatial neighborhood Y2 of image-region X, (several optional examples of spatial neighborhood Y2 are as schemed 3) the prediction pixel information for the image-region X that the Pixel Information in generates.

Wherein, intra prediction mode for example H.265/HEVC in the intra prediction mode based on direction, DC prediction mode, Planar prediction mode, in another example stencil matching (template matching) method or intra frame motion compensation (Intra Motion Compensation) etc. modes.

The prediction pixel information of image-region X includes at least one intra prediction pixel of image-region X；Spatial neighborhood Y2 It can be a part of spatial neighborhood Y1, also may include the region other than spatial neighborhood Y1.Such as: spatial neighborhood Y2 can be Left or the upper area in the L shape region (a example and b example in such as Fig. 3) or X adjacent with X left margin and coboundary In one with the rectangular area (c example and d example in such as Fig. 3) of X same size or one comprising in the rectangular area of X Remove the partial region (the e example in such as Fig. 3) of X.

Step 103, quantization regulatory factor is calculated according to decoded pixel information；

In embodiments of the present invention, it in order to realize adaptive quantization effect, can be utilized when carrying out inverse quantization The parameter of some characteristics of decoded pixel information is able to reflect to calculate quantization regulatory factor；In the embodiment of the present invention, it can adopt Quantization regulatory factor is calculated with the parameter of reflection decoded pixel statistical property, is specifically included:

The statistical property of decoded pixel information refers to the pixel value to pixels multiple in decoded pixel information by statistical Obtained numerical result is analysed, the parameter with statistical property includes at least pixel mean value P_avg, pixel variance P_var, pixel maximum P_max, pixel minimum P_minOne of.

Step 104, inverse quantization is carried out to the first transformation series manifold according to quantization regulatory factor, obtains the second transformation series Manifold.

In order to obtain more decoded pixel information, statistical property is made to calculate more accurate stabilization, in the embodiment of the present invention Decoded pixel information can also include image-region X time neighborhood territory pixel information, wherein time neighborhood territory pixel information be work as The Pixel Information of the coded image of the preceding a later frame or a few frames of image where preceding decoding image；And time neighborhood territory pixel packet Include at least one of following two:

1) image-region X the reconstructed value of at least one pixel or passes through in the corresponding region on another width encoded image The sub-pixel precision reconstructed value that filtering interpolation obtains；

Corresponding region can be obtained according to the motion vector of image-region X by common motion compensation process.Particularly, It can also be approximately that integer value moves again by the motion vector when the motion vector of image-region X is sub-pixel precision Compensation obtains the corresponding region.

2) image-region X respective pixel at least one position in the corresponding region at least two width encoded images The weighted average of reconstructed value.

Such as: when using two width encoded images, two encoded figures are found by the motion vector of image-region X respectively Corresponding region as in, the pixel in two corresponding regions on same position are respective pixel, and the weighting of one group of respective pixel is flat Mean value is as a pixel in time neighborhood territory pixel information；Same position can be whole location of pixels or sub-pixel location.

In the embodiment of the present invention, believed in decoding end using the spatial neighborhood Pixel Information or spatial neighborhood pixel rebuild The combination of breath and time neighborhood territory pixel information, carrys out the corresponding original pixels information of approximate current decoding block, export quantization adjust because Son, and then the quantization step of image different zones residual transform coefficient is adaptively adjusted, realize the promotion of coding efficiency.

Scheme provided by the embodiment of the present invention does not need the parameter information relevant to quantization is adjusted of the transmission in code stream, because This saves code rate.In addition, the embodiment of the present invention can export quantization regulatory factor merely with spatial neighborhood Pixel Information, because This is suitable for the coding of I picture.Method provided by the embodiment of the present invention can equally be well applied to the coding of inter frame image.

Because the decoded pixel information in the embodiment of the present invention can be the spatial neighborhood pixel letter of image-region X Breath, is also possible to the combination of time the neighborhood territory pixel information and spatial neighborhood Pixel Information of image-region X, because decoded pixel is believed The different compositions of breath, so in embodiments of the present invention, can be accomplished in several ways according to decoded pixel information calculation amount Change regulatory factor；Specific implementation includes:

Firstly, in embodiments of the present invention can be according to the corresponding parameter of decoded pixel information: pixel mean value P_avg, pixel Variance P_var, pixel maximum P_max, pixel minimum P_min, calculate quantization regulatory factor.

Based on above-mentioned parameter (pixel mean value P_avg, pixel variance P_var, pixel maximum P_max, pixel minimum P_min) calculate Quantify regulatory factor QC, specific method can in the following ways in any one:

Method one: according to pixel mean value P_avgFirst function f₁(P_avg) calculate quantization regulatory factor；Wherein, the pixel Mean value P_avgIt is the mean value or institute that the luminance component of multiple pixels corresponding to the decoded pixel information is averaging The mean value that the same color space component of multiple pixels is averaging is stated, the color space component is first transformation The corresponding color space component of coefficient set.

When decoded pixel includes only spatial neighborhood Pixel Information, pixel mean value P_avgSpecifically:

Pixel mean value P_avg, pixel mean value P_avgFor the luminance component of multiple pixels in the spatial neighborhood Pixel Information Mean value, or be the mean value of the same signal component of the multiple pixels of spatial neighborhood Pixel Information, the signal component is institute State the corresponding signal component of the first transformation series manifold；

When decoded pixel is the combination of spatial neighborhood Pixel Information and time neighborhood territory pixel information, pixel mean value P_avgTool Body are as follows:

Pixel mean value P_avg, pixel mean value P_avgIt is more in the spatial neighborhood Pixel Information and time neighborhood territory pixel information The mean value of the luminance component of a pixel, or be multiple pixels in the spatial neighborhood Pixel Information and time neighborhood territory pixel information The mean value of the same signal component, the signal component are the corresponding signal component of the first transformation series manifold；

In specific example, first function can be QC=f₁(P_avg)^β, wherein β > 0；Such as β=1 or 0.5.f₁ (P_avg) can be about P_avgU-shaped function, f₁(P_avg) meet and work as P_avgF when less than threshold value T1₁(P_avg) first derivative be less than 0, work as P_avgF when greater than threshold value T2₁(P_avg) first derivative be greater than 0, P_avgF when between threshold value T1 and T2₁(P_avg) it is equal to constant C0；Wherein, T1 >=0, T2 >=0, T2 >=T1, T1 can be 0,60,64 or 128；T2 can be 0,80,128 or 170；C0 is positive Real number can be 0.5,0.75,1,1.5,8,16,256 or 1024.More specifically, f₁(P_avg) function may is that

Wherein: η₁For positive real number, such as η₁=150 or 200.8；η₂For positive real number, such as η₂=425 or 485.5.f₁ (P_avg) function in another example

Wherein η₃For positive real number, such as η₃=10240 or 8135.6；η₄For positive real number, such as η₄=256 or 135.1.

Method two: according to pixel variance P_varSecond function f₂(P_var) calculate quantization regulatory factor；Wherein, the pixel Variance P_varIt for the variance of the luminance component of the corresponding multiple pixels of the decoded pixel information, or is the multiple pixel The variance of the same color space component, the color space component are the corresponding color space point of the first transformation series manifold Amount.

When decoded pixel includes only spatial neighborhood Pixel Information, pixel variance P_varSpecifically:

Pixel variance P_var, pixel variance P_varFor the side of the luminance component of multiple pixels in the spatial neighborhood Pixel Information Difference, or be the variance of the same signal component of the multiple pixels of spatial neighborhood Pixel Information, the signal component is described The corresponding signal component of first transformation series manifold；

When decoded pixel is the combination of spatial neighborhood Pixel Information and time neighborhood territory pixel information, pixel variance P_varTool Body are as follows:

Pixel variance P_var, pixel variance P_varFor multiple pixels in spatial neighborhood Pixel Information and time neighborhood territory pixel information Luminance component variance, or be multiple same signals of pixel in spatial neighborhood Pixel Information and time neighborhood territory pixel information The variance of component, the signal component are the corresponding signal component of the first transformation series manifold；

In specific example, according to pixel variance P_varSecond function QC=f₂(P_var)^γCalculate quantization regulatory factor QC； Wherein, γ > 0；γ such as 1 or 0.8.f₂(P_var) can be about P_varMonotonic function, f₂(P_var) meet as (P_var)^αIt is small When threshold value T3, f₂(P_var) it is constant C3, as (P_var)^αWhen more than or equal to threshold value T3, f₂(P_var) first derivative be greater than 0.Its In, T3 >=0, T3 such as 0,3,5 or 10；α > 0, such as α=1/2 or 1；C3 is positive real number, for example, 0.5,0.8,1,16,32 or 256.More specifically, f₂(P_var) function may is that

Wherein, η₅For positive real number, such as η₅=10,20,35.5,80 or 100.

Method three: according to pixel maximum P_maxWith pixel minimum P_minDifference third function f₃(P_max-P_min) meter Calculate quantization regulatory factor；Wherein, the pixel maximum P_maxFor the brightness point of the corresponding multiple pixels of the decoded pixel information The maximum value of the maximum value of amount or the same color space component of the multiple pixel；The pixel minimum is described The minimum value of the luminance component of multiple pixels, or be the minimum value of the same spatial component of the multiple pixel, the color Spatial component is the corresponding color space component of the first transformation series manifold.

According to pixel mean value P_avgFirst function QC=f₁(P_avg)^βWith pixel variance P_varSecond function QC=f₂ (P_var)^γCopula QC=f₁(P_avg)^β·f₂(P_var)^γCalculate quantization regulatory factor QC, wherein beta, gamma > 0, such as β=1, γ=1 or β=0.5, γ=1.5 or β=2, γ=1.

Method four: according to pixel maximum P_maxWith pixel minimum P_minDifference third function f₃(P_max-P_min) meter Calculate quantization regulatory factor；Wherein, the pixel maximum P_maxFor the brightness point of the corresponding multiple pixels of the decoded pixel information The maximum value of the maximum value of amount or the same color space component of the multiple pixel；The pixel minimum is described The minimum value of the luminance component of multiple pixels, or be the minimum value of the same spatial component of the multiple pixel, the color Spatial component is the corresponding color space component of the first transformation series manifold.

When decoded pixel includes only spatial neighborhood Pixel Information, pixel maximum P_maxWith pixel minimum P_minSpecifically Are as follows:

Pixel maximum P_max, pixel maximum P_maxThe luminance component of multiple pixels in the spatial neighborhood Pixel Information Maximum value, or be the maximum value of the same signal component of the multiple pixels of spatial neighborhood Pixel Information, the signal point Amount is the corresponding signal component of the first transformation series manifold；

Pixel minimum P_min, pixel minimum P_minFor the brightness point of multiple pixels in the spatial neighborhood Pixel Information The minimum value of amount, or be the minimum value of the same signal component of the multiple pixels of spatial neighborhood Pixel Information, the signal Component is the corresponding signal component of the first transformation series manifold.

When decoded pixel is the combination of spatial neighborhood Pixel Information and time neighborhood territory pixel information, pixel maximum P_max With pixel minimum P_minSpecifically:

Pixel maximum P_max, pixel maximum P_maxFor the spatial neighborhood Pixel Information and time neighborhood territory pixel information In multiple pixels luminance component maximum value, or be more in the spatial neighborhood Pixel Information and time neighborhood territory pixel information The maximum value of a same signal component of pixel, the signal component are the corresponding signal component of the first transformation series manifold；

Pixel minimum P_min, pixel minimum P_minFor the spatial neighborhood Pixel Information and time neighborhood territory pixel information In multiple pixels luminance component minimum value, or be more in the spatial neighborhood Pixel Information and time neighborhood territory pixel information The minimum value of a same signal component of pixel, the signal component are the corresponding signal component of the first transformation series manifold.

In specific example, according to pixel maximum P_maxWith pixel minimum P_minDifference third function QC=f₃ (P_max-P_min) calculate quantization regulatory factor QC.The third function can be monotonically increasing function, such as:

Wherein, C4 is positive real number, such as 0.5,1,16,32 or 256；η₆For positive real number, such as η₆=1 or 1/2；η₇It is positive Real number, such as η₇=20 or η₇=C4/2.

After obtaining quantization regulatory factor QC by one of above-mentioned four kinds of methods, it is also an option that further to the quantization Regulatory factor QC carries out numerical value limitation processing, and QC is corresponded within a preset numberical range.Such as:

QC is clamped down within a preset numberical range, i.e.,

Wherein Val1 and Val2 is positive real number, such as Val1=0.5, Val2=2 or Val1=0.6, Val2=1.5；

In another example by QC by the way that within mapping function to a certain numberical range, the mapping function is for example

Wherein Val3, Val4, Val5, Val6 are real number, such as Val3=1.0, Val4=0.5, Val5=1.5, Val6 =1.0；Or Val3=1.1, Val4=0.4, Val5=1.2, Val6=1.15.

Wherein, above-mentioned parameter T1, T2, T3, C0, C3, C4, η₁、η₂、η₃、η₄、η₅、η₆、η₇、Val1、Val2、Val3、 Val4, Val5, Val6 can be preset constant, can also be obtained according to the statistical property adaptive polo placement of video image It arrives, can also extract and obtain from video code flow.

In the embodiment of the present invention, can be accomplished in several ways step 104 inverse quantization processing, it is presented below it is several can Implementation method:

Method one: the corresponding first quantization step of the first transformation series manifold is obtained from signal corresponding to current decoding image It is long；The first quantization step is zoomed in and out using quantization regulatory factor, obtains the second quantization step, it is anti-using the second quantization step Quantify the first transformation series manifold, obtains the second transformation series manifold.Specific implementation may is that

The corresponding first quantization step of the first transformation series manifold is obtained from signal corresponding to the current decoding image Long Qs (i)；

The first quantization step Qs (i) is zoomed in and out according to quantization regulatory factor, is scaled Qs (i) QC, inverse quantization the The transformation coefficient A (i) that one variation coefficient is concentrated obtains the transformation coefficient R (i) of the second transformation coefficient concentration, that is, has:

R (i)=sign { A (i) } round { A (i) Qs (i) QC+o2 (i) }

Wherein o2 (i) is round-off constant, such as o2 (i)=0.5.When inverse quantization is realized using addition of integer and displacement, It can use:

R (i)=sign { A (i) } (A (i) Qs'(i) QC'+ (1<<(bdshift-1+delta)))>> (bdshift+delta)

Wherein Qs'(i) it is the integer of the first quantization step as a result, bdshift is signal corresponding to current decoding image In shift amount；QC' obtains QC' as a result, QC can for example be multiplied with 2 integral number power and be rounded for the integer of QC, or Person is directly rounded QC and obtains QC', or obtains the integer result of QC according to pixel mean value, pixel variance with look-up table； Delta is the shift amount being additionally required.

Method two: the corresponding target level scale of the quantization regulatory factor is determined according to preset first corresponding relationship Table；Wherein, first corresponding relationship is the corresponding relationship for quantifying regulatory factor and level scale table；According to the target level First transformation series manifold described in scale table inverse quantization, obtains the second transformation series manifold.

The first of a quantization regulatory factor QC and multiple level scale tables can be preset in embodiments of the present invention Corresponding relationship.

More specifically, the value range of QC is divided into H1 in advance (H1 is positive integer greater than 1, such as H1=3,5 or 6) Section section, each section of section correspond to a preset level scale table；Retrieve what calculated quantization regulatory factor QC belonged to Value interval is chosen level scale table corresponding to the value interval that quantization regulatory factor QC belongs to and is converted as inverse quantization first The level scale table of coefficient set.Such as: it provides in the mapping table (table 1) of QC and level scale table for comprising 6 elements Level scale table, according to QC select a level scale table example.

	Level scale table
		QC≥1.5	{60,68,77,86,96,108}

1.2<QC<1.5	{54,61,69,77,86,97}
		0.8≤QC≤1.2	{40,45,51,57,64,72}
0.6<QC<0.8	{28,32,36,40,45,50}
		QC≤0.6	{24,27,31,34,38,43}

Table 1

According to example given by table 1, when QC is 0.9, the value interval that retrieval QC belongs to is 0.8≤QC≤1.2, choosing Level scale table { 40,45,51,57,64,72 } corresponding to value interval 0.8≤QC≤1.2 is taken to convert as inverse quantization first The level scale table of coefficient set.

Method three: the corresponding Target quantization matrix of the quantization regulatory factor is determined according to preset second corresponding relationship； Wherein, second corresponding relationship is the corresponding relationship for quantifying regulatory factor and quantization matrix；According to the Target quantization matrix First transformation series manifold described in inverse quantization obtains the second transformation series manifold.

Wherein, the value range of QC is divided into H2 in advance (H2 is positive integer greater than 1, such as H2=3,5 or 6) section area Between, each section of section corresponds to a preset quantization matrix；The value interval that retrieval QC belongs to, chooses corresponding quantization matrix Quantization matrix as inverse quantization the first transformation series manifold.Such as: it is provided in table 2 for 4 × 4 transform blocks

Table 2

According to the example that table 2 provides, if calculated QC value is 18, the parameter given by the table 2 can determine that QC belongs to In value interval be 16≤QC≤64, then the first transformation series manifold described in corresponding inverse quantization quantization matrix 4 × 4 quantify square Battle array is 4 × 4 quantization matrixes shown in 2 third of table column.

Method four: the first transformation series manifold corresponding first is obtained from signal corresponding to the current decoding image Quantization parameter；Target quantization parameter offset is determined according to the quantization regulatory factor, by first quantization parameter and described Target quantization parameter offset addition obtains the second quantization parameter；Level scale is determined according to second quantization parameter, is utilized First transformation series manifold described in the level scale inverse quantization, obtains the second transformation series manifold.

Wherein, using level scale l (i) to the first transformation parameter collection carry out inverse quantization specific implementation can be by with Lower formula is realized:

DQP=round { 6log₂(QC)+o3 },

QP₂=QP+dQP,

Wherein o3 is rounding-off biasing.

Method five: the first transformation series manifold corresponding first is obtained from signal corresponding to the current decoding image Quantization step；Inverse quantization is carried out to the coefficient that first transformation coefficient is concentrated using first quantization step and obtains third change Change the coefficient in coefficient set；The coefficient that the third transformation coefficient is concentrated using the quantization regulatory factor being calculated into Row scaling obtains the second transformation series manifold.

Specific implementation includes: first to carry out inverse quantization to the first transformation series manifold with the first quantization step Qs (i) to obtain third Transformation series manifold B (i) has B (i)=sign { A (i) } round { A (i) Qs (i)+o4 (i) }, then contracts to B (i) It puts and is rounded to obtain the second transformation coefficient R (i), that is, have R (i)=sign { B (i) } round { B (i) QC+o5 (i) }, wherein O4 and o5 is rounding-off biasing.

Wherein, the first quantization parameter QP is as where the first variation coefficient collection for currently decoding signal designation corresponding to image The default quantization parameter of transform block, such as frame level quantization parameter.

Scheme provided by the embodiment of the present invention utilizes the spatial neighborhood Pixel Information or space rebuild in decoding end The combination of neighborhood territory pixel information and time neighborhood territory pixel information comes the corresponding original pixels information of approximate current decoding block, export Quantify regulatory factor, and then the quantization step of image different zones residual transform coefficient is adaptively adjusted, realizes coding efficiency Promotion.

Embodiment

As shown in figure 4, the embodiment of the present invention also provides adaptive quantification method, this method in another Video coding Include:

Step 401, the first transformation series manifold corresponding image-region X in currently decoding image is determined；First transformation series Manifold includes N number of transformation coefficient, and transformation coefficient is the transformation coefficient of any color space component in current decoding image, wherein N For positive integer；

In this embodiment, the first transformation series manifold may include all ac coefficients of a luminance component transform block, Or all coefficients in a luminance component transform block.

Step 402, the time neighborhood territory pixel information of described image region X is obtained；

Wherein, time neighborhood territory pixel information include: image-region X in the corresponding region on another width encoded image extremely The reconstructed value of a few pixel or the sub-pixel precision reconstructed value obtained by filtering interpolation；And/or image-region X is at least In corresponding region on two width encoded images at least one position the reconstructed value of respective pixel weighted average.

Wherein, when the corresponding region as image-region X on another width encoded image is sub-pixel location, time neighborhood Pixel Information is the sub-pixel precision reconstructed value obtained in sub-pixel location by filtering interpolation.

Step 403, according to the first mean value P_avgWith first variance P_varCalculate quantization regulatory factor, wherein described first Value P_avgIt is the mean value being averaging to the luminance component of multiple pixels in the decoded pixel information, or is the multiple The mean value that the same color space component of pixel is averaging；The first variance P_varFor the brightness of the multiple pixel The variance of component, or the variance of the same color space component for the multiple pixel, the color space component are institute State the corresponding color space component of the first transformation series manifold；

According to the first mean value P_avgFirst function QC=f₁(P_avg)^βWith first variance P_varSecond function QC=f₂ (P_var)^γCopula QC=f₁(P_avg)^β·f₂(P_var)^γQuantization regulatory factor QC is calculated, wherein such as β=0.4, γ= 0.6 or β=0.5, γ=0.5.Wherein, first function is for example:

Second function is for example:

It in embodiments of the present invention, can be with after calculating quantization regulatory factor using the first function and second function To quantifying, regulatory factor QC is further to be handled for selection, comprising:

Numerical value limitation processing is carried out to quantization regulatory factor QC, QC is corresponded within a preset numberical range.

Wherein, the optional way for carrying out the specific implementation of numerical value limitation processing includes: to clamp down on QC in a preset numerical value Within the scope of；Alternatively, QC is passed through within mapping function to a certain numberical range.

Step 404, inverse quantization processing is carried out to the first transformation series manifold according to the quantization regulatory factor, obtains the Two transformation series manifolds.

In embodiments of the present invention, it can be accomplished in several ways the inverse quantization transformation of the first transformation series manifold, below Several implemented methods are provided:

R (i)=sign { A (i) } round { A (i) Qs (i) QC+o2 (i) }

	Level scale table
		QC≥1.5	{60,68,77,86,96,108}
1.2<QC<1.5	{54,61,69,77,86,97}
		0.8≤QC≤1.2	{40,45,51,57,64,72}
0.6<QC<0.8	{28,32,36,40,45,50}
		QC≤0.6	{24,27,31,34,38,43}

Table 1

Table 2

DQP=round { 6log₂(QC)+o3 },

QP₂=QP+dQP,

Wherein o3 is rounding-off biasing.

Embodiment

It is described in detail below in conjunction with scheme of the specific parameter to the embodiment of the present invention, specific implementation includes:

First transformation series manifold includes all coefficients of a luminance component transform block.Image-region X is one 8 × 8 big Small region (i.e. one rectangular area being made of 8 row, 8 column pixel), the coordinate of the top left corner pixel of image-region X is denoted as (x₀,y₀)；The coordinate system of usual image is defined as from left to right for the direction x, is from the top down the direction y.

Spatial neighborhood Pixel Information includes following two parts:

1) Pixel Information in the spatial neighborhood Y1 of image-region X, spatial neighborhood Pixel Information include in spatial neighborhood Y1 At least one pixel reconstructed value；If the spatial neighborhood Y1 includes that pixel coordinate (x, y) meets condition:

x₀-J1≤x≤x₀+J2,y₀-J3≤y≤y₀- J4, or

y₀-J5≤y≤y₀+J6,x₀-J7≤x≤x₀The pixel of-J8, J1 to J8 are positive integer, such as J1=J2=J3= J5=J6=J7=8, J4=J8=1 or J1=J3=12, J2=4, J4=1, J5=J7=12, J6=4, J8=1, or Person J1=5, J2=8, J3=6, J4=2, J5=10, J6=11, J7=12, J8=1；

2) by intra prediction mode, the image-region X generated by the Pixel Information in the spatial neighborhood Y2 of image-region X Prediction pixel information, Pixel Information includes at least one intra prediction pixel of image-region X；Spatial neighborhood Y2 includes pixel Coordinate (x, y) meets condition:

X=x₀-1,y₀-1≤y≤y+S1；Or

Y=y₀-1,x₀- 1≤x≤x+S2 pixel, S1, S2 are positive integer, such as S1=S2=15 or S1=7, S2 =15；The intra prediction mode can be one of 33 kinds of directional intra prediction modes in H.265/HEVC, DC prediction mould Formula or Planar prediction mode.

The quantization regulatory factor QC is by pixel mean value P_avgFirst function f₁(P_avg) and about the pixel variance P_var Second function f₂(P_var) joint decision, using QC=f₁(P_avg)^1/2·f₂(P_var)^1/2, wherein the variance P of luminance component_var With mean value P_avgFor the variance and mean value of all pixels in spatial neighborhood Pixel Information.First function and second function specifically:

According to quantization regulatory factor, the inverse quantization of the transformation coefficient of the first transformation series manifold is adjusted, inverse quantization first converts Coefficient set obtains the second transformation series manifold, wherein the method for adjusting the inverse quantization of the transformation coefficient of the first transformation series manifold are as follows: root Original quantization step Qs (i) of the first transformation series manifold is zoomed in and out according to quantization regulatory factor, by the quantization step after scaling Quantization step as inverse quantization the first transformation series manifold.

Embodiment

As shown in figure 5, the embodiment of the present invention provides adaptive inverse quantization device in a kind of Video coding, which includes:

Determination unit 501, for determining the first transformation series manifold corresponding image-region in currently decoding image；It is described First transformation series manifold includes N number of transformation coefficient, and the transformation coefficient is any color space point in the current decoding image The transformation coefficient of amount, wherein N is positive integer；

Acquiring unit 502, for obtaining the decoded pixel information in described image region, the decoded pixel information includes institute State the spatial neighborhood Pixel Information of image-region；

In embodiments of the present invention, acquiring unit is also used to obtain the described image region in the decoded pixel information Time neighborhood territory pixel information and/or the Pixel Information in described image region.

Computing unit 503, for calculating quantization regulatory factor according to the decoded pixel information；

Inverse quantization unit 504, for carrying out inverse quantization to the first transformation series manifold according to the quantization regulatory factor, Obtain the second transformation series manifold.

In embodiments of the present invention, computing unit 503 calculates the specific reality of quantization regulatory factor according to decoded pixel information Existing mode includes a variety of, several achievable modes presented below:

Mode one, according to the pixel mean value P of decoded pixel information_avgQuantization regulatory factor is calculated, then computing unit 503 has Body is used for:

Mode two, according to the pixel variance P of decoded pixel information_varQuantization regulatory factor is calculated, then computing unit 503 has Body is used for:

Mode three, according to the pixel variance P of decoded pixel information_varWith pixel mean value P_avgQuantization regulatory factor is calculated, then Computing unit 503 is specifically used for:

Mode three, according to the pixel maximum P of decoded pixel information_maxWith pixel minimum P_minCalculate quantization adjust because Son, then computing unit 503 is specifically used for:

In embodiments of the present invention, inverse quantization unit 504 is according to the quantization regulatory factor to first transformation coefficient Collection carries out inverse quantization, and the specific implementation for obtaining the second transformation series manifold includes a variety of, several achievable modes presented below:

Mode one, using level scale table inverse quantization the first transformation series manifold, then inverse quantization unit 504 is specifically used for:

The corresponding target level scale table of the quantization regulatory factor is determined according to preset first corresponding relationship；Wherein, First corresponding relationship is the corresponding relationship for quantifying regulatory factor and level scale table；It is anti-according to the target level scale table Quantify the first transformation series manifold, obtains the second transformation series manifold.

Mode two, using quantization matrix inverse quantization the first transformation series manifold, then inverse quantization unit 504 is specifically used for:

The corresponding Target quantization matrix of the quantization regulatory factor is determined according to preset second corresponding relationship；Wherein, institute Stating the second corresponding relationship is the corresponding relationship for quantifying regulatory factor and quantization matrix；According to Target quantization matrix inverse quantization institute The first transformation series manifold is stated, the second transformation series manifold is obtained.

Mode three utilizes the quantization parameter carried in signal corresponding to current decoding image and the adaptive quantity being calculated Change difference inverse quantization the first transformation series manifold between parameter, then inverse quantization unit 504 is specifically used for:

The corresponding first quantization ginseng of the first transformation series manifold is obtained from signal corresponding to the current decoding image Number；Target quantization parameter offset is determined according to the quantization regulatory factor, by first quantization parameter and the aim parameter Change parameter shift amount to be added to obtain the second quantization parameter；Level scale is determined according to second quantization parameter, utilizes the electricity First transformation series manifold described in leveling ruler degree inverse quantization, obtains the second transformation series manifold.

Mode four utilizes the first transformation coefficient of quantization step inverse quantization carried in signal corresponding to current decoding image Collection, then inverse quantization unit 504 is specifically used for:

The corresponding first quantization step of the first transformation series manifold is obtained from signal corresponding to the current decoding image It is long；First quantization step is zoomed in and out using the quantization regulatory factor, obtains the second quantization step, utilizes described First transformation series manifold described in two quantization step inverse quantizations, obtains the second transformation series manifold.

Mode five utilizes the first transformation coefficient of quantization step inverse quantization carried in signal corresponding to current decoding image Collection, then inverse quantization unit 504 is specifically used for:

The corresponding first quantization step of the first transformation series manifold is obtained from signal corresponding to the current decoding image It is long；Inverse quantization is carried out to the coefficient that first transformation coefficient is concentrated using first quantization step and obtains third transformation coefficient The coefficient of concentration；It is zoomed in and out using the coefficient that the quantization regulatory factor being calculated concentrates the third transformation coefficient Obtain the second transformation series manifold.

Said one or multiple technical solutions in the embodiment of the present application, at least have the following technical effect that:

Scheme provided in an embodiment of the present invention utilizes space neighborhood information, time neighborhood information or the space of transform block Neighborhood information and time neighborhood information estimate the statistical property of background area locating for Current Transform block, adaptively export inverse The quantization regulatory factor of change neatly adjusts inverse quantization processing.On the one hand, adaptive quantizing method of the invention loses quantization True intensity more meets human eye visual perception, to improve code efficiency；On the other hand compared to the transmission quantization tune in code stream The existing scheme of information is saved, the present invention does not need to provide additional overhead bit transmission quantization adjusting information, therefore further mentions High code efficiency.

In addition, the present invention is suitable for the coding of I picture (I frame) and inter frame image (P frame or B frame), so the present invention is real Technical solution provided by example is applied with the wider scope of application.

Method of the present invention is not limited to embodiment described in specific embodiment, those skilled in the art according to Technical solution of the present invention obtains other embodiments, also belongs to the scope of the technical innovation of the present invention.

Obviously, various changes and modifications can be made to the invention without departing from essence of the invention by those skilled in the art Mind and range.In this way, if these modifications and changes of the present invention belongs to the range of the claims in the present invention and its equivalent technologies Within, then the present invention is also intended to include these modifications and variations.

Claims

1. an adaptive inverse quantization method in video coding, is characterized in that, the method comprises:

Determine the image area corresponding to the first transform coefficient set in the current decoded image; the first transform coefficient set includes N transform coefficients, and the transform coefficients are transform coefficients of any color space component in the currently decoded image, wherein , N is a positive integer;

obtaining decoded pixel information of the image area, where the decoded pixel information includes spatial neighborhood pixel information of the image area;

Calculate a quantization adjustment factor according to the decoded pixel information;

Performing inverse quantization on the first transform coefficient set according to the quantization adjustment factor to obtain a second transform coefficient set includes: obtaining a first quantization step corresponding to the first transform coefficient set from a signal corresponding to the current decoded image using the quantization adjustment factor to scale the first quantization step size to obtain a second quantization step size, and use the second quantization step size to inversely quantize the first transform coefficient set to obtain the second transform coefficient set;

According to the decoded pixel information, calculating the quantization adjustment factor includes:

The quantization adjustment factor is calculated according to the first function f ₁ (P _avg ) of the pixel mean value P _avg ; wherein the pixel mean value P _avg is the mean value obtained by averaging the luminance components of the plurality of pixels corresponding to the decoded pixel information, or is the mean value obtained by averaging the same color space component of the plurality of pixels, and the color space component is the color space component corresponding to the first transformation coefficient set; f ₁ (P _avg ) may be U about P _avg shape function, f ₁ (P _avg ) satisfies that the first derivative of f ₁ (P _avg ) is less than 0 when P _avg is less than the threshold T1, and the first derivative of f ₁ (P _avg ) is greater than 0 when P _avg is greater than the threshold T2, f ₁ (P _avg ) equals constant C0 when P _avg is between thresholds T1 and T2; where T1 ≥ 0, T2 ≥ 0, and T2 ≥ T1.

2. The method of claim 1, wherein the decoded pixel information comprises temporal neighborhood pixel information of the image area and/or pixel information of the image area.

3. A method for adaptive inverse quantization in video coding, characterized in that the method comprises:

The quantization adjustment factor is calculated according to the second function f ₂ (P _var ) of the pixel variance P _var ; wherein, the pixel variance P _var is the variance of the luminance components of the plurality of pixels corresponding to the decoded pixel information, or the The variance of the same color space component of pixels, the color space component is the color space component corresponding to the first transform coefficient set; f ₂ (P _var ) is a monotonic function about P _var , f ₂ (P _var ) Satisfy that when (P _var ) ^α is less than the threshold value T3, f ₂ (P _var ) is a constant C3, and when (P _var ) ^α is greater than or equal to the threshold value T3, the first derivative of f ₂ (P _var ) is greater than 0.

4. An adaptive inverse quantization method in video coding, characterized in that the method comprises:

Performing inverse quantization on the first transform coefficient set according to the quantization adjustment factor to obtain a second transform coefficient set includes: obtaining a first quantization step corresponding to the first transform coefficient set from a signal corresponding to the current decoded image using the quantization adjustment factor to scale the first quantization step size to obtain a second quantization step size, and use the second quantization step size to inversely quantize the first transform coefficient set to obtain the second transform coefficient set; according to the decoded pixel information, calculating the quantization adjustment factor includes:

The quantization adjustment factor is calculated according to the joint function f ₁ (P _avg )·f ₂ (P _var ) of the pixel variance P _var and the pixel mean value P _avg ; wherein the pixel mean value P _avg is a The mean value obtained by averaging the luminance components of , or the mean value obtained by averaging the same color space components of the plurality of pixels; the pixel variance P _var is the variance of the luminance components of the plurality of pixels, or the The variance of the same color space component of multiple pixels, the color space component is the color space component corresponding to the first transformation coefficient set; f ₁ (P _avg ) is a U-shaped function about P _avg , f ₁ (P avg ) _avg ) satisfies that the first derivative of f ₁ (P _avg ) is less than 0 when P _avg is less than the threshold T1, and the first derivative of f ₁ (P _avg ) is greater than 0 when P _avg is greater than the threshold T2, and P _avg is between the thresholds T1 and T2 When f ₁ (P _avg ) is equal to the constant C0, f ₂ (P _var ) is a monotonic function of P _var , and f ₂ (P _var ) satisfies that when (P _var ) ^α is less than the threshold T3, f ₂ (P _var ) ) is a constant C3, and when (P _var ) ^α is greater than or equal to the threshold value T3, the first derivative of f ₂ (P _var ) is greater than 0.

5. An adaptive inverse quantization method in video coding, characterized in that the method comprises:

The quantization adjustment factor is calculated according to the third function f ₃ (P _max -P _min ) of the difference between the pixel maximum value P _max and the pixel minimum value P _min ; wherein, the pixel maximum value P _max is the value corresponding to the decoded pixel information The maximum value of the luminance components of multiple pixels, or the maximum value of the same color space component of the multiple pixels; the minimum pixel value is the minimum value of the luminance components of the multiple pixels, or the multiple The minimum value of the same space component of a pixel, where the color space component is the color space component corresponding to the first transform coefficient set.

6. The method according to any one of claims 1-5, wherein the performing inverse quantization processing on the first transform coefficient set according to the quantization adjustment factor to obtain the second transform coefficient set comprises:

Determine the target level scale table corresponding to the quantization adjustment factor according to a preset first correspondence; wherein, the first correspondence is the correspondence between the quantization adjustment factor and the level scale table;

The first transform coefficient set is inversely quantized according to the target level scale table to obtain the second transform coefficient set.

7. The method according to any one of claims 1-5, wherein the performing inverse quantization processing on the first transform coefficient set according to the quantization adjustment factor to obtain the second transform coefficient set comprises:

Determine the target quantization matrix corresponding to the quantization adjustment factor according to a preset second correspondence; wherein, the second correspondence is the correspondence between the quantization adjustment factor and the quantization matrix;

The first transform coefficient set is inversely quantized according to the target quantization matrix to obtain the second transform coefficient set.

8. The method according to any one of claims 1-5, wherein the performing inverse quantization processing on the first transform coefficient set according to the quantization adjustment factor to obtain the second transform coefficient set comprises:

Obtain the first quantization parameter corresponding to the first transform coefficient set from the signal corresponding to the current decoded image;

Determine a target quantization parameter offset according to the quantization adjustment factor, and add the first quantization parameter and the target quantization parameter offset to obtain a second quantization parameter;

A level scale is determined according to the second quantization parameter, and the first transform coefficient set is inversely quantized by using the level scale to obtain the second transform coefficient set.

9. The method according to any one of claims 1-5, wherein the performing inverse quantization processing on the first transform coefficient set according to the quantization adjustment factor to obtain the second transform coefficient set comprises:

Obtain the first quantization step size corresponding to the first transform coefficient set from the signal corresponding to the current decoded image;

Perform inverse quantization on the coefficients in the first transform coefficient set by using the first quantization step size to obtain the coefficients in the third transform coefficient set;

The coefficients in the third transform coefficient set are scaled by using the calculated quantization adjustment factor to obtain a second transform coefficient set.

10. A method for adaptive inverse quantization in video coding, characterized in that the method comprises:

Determine the image area X corresponding to the first transform coefficient set in the current decoded image; the first transform coefficient set includes N transform coefficients, and the transform coefficients are transform coefficients of any color space component in the current decoded image, Among them, N is a positive integer;

obtaining the temporal neighborhood pixel information of the image area X;

The quantization adjustment factor is calculated according to the first average value P _avg and the first variance P _var , where the first average value P _avg is an average value obtained by averaging the luminance components of a plurality of pixels in the temporal neighborhood pixel information, or The mean value obtained by averaging the same color space components of the plurality of pixels; the first variance P _var is the variance of the luminance components of the plurality of pixels, or the same color space of the plurality of pixels The variance of the component, the color space component is the color space component corresponding to the first transform coefficient set;

Calculating the quantitative adjustment factor according to the first mean value P _avg and the first variance P _var includes:

Calculate the quantization adjustment factor according to the joint function f ₁ (P _avg )·f ₂ (P _var ) of the first mean value P _avg and the first variance P _var ;

Performing inverse quantization processing on the first transform coefficient set according to the quantization adjustment factor to obtain a second transform coefficient set includes:

The first quantization step size is scaled by the quantization adjustment factor to obtain a second quantization step size, and the first transform coefficient set is inversely quantized by using the second quantization step size to obtain the second transform coefficient set .

11. The method according to claim 10, wherein the performing inverse quantization processing on the first transform coefficient set according to the quantization adjustment factor to obtain the second transform coefficient set comprises:

12. The method according to claim 10, wherein the performing inverse quantization processing on the first transform coefficient set according to the quantization adjustment factor to obtain the second transform coefficient set comprises:

13. The method according to claim 10, wherein the performing inverse quantization processing on the first transform coefficient set according to the quantization adjustment factor to obtain the second transform coefficient set comprises:

14. The method according to claim 10, wherein the performing inverse quantization processing on the first transform coefficient set according to the quantization adjustment factor to obtain the second transform coefficient set comprises:

15. A device for adaptive inverse quantization in video coding, characterized in that the device comprises:

A determination unit, configured to determine an image area corresponding to the first transformation coefficient set in the current decoded image; the first transformation coefficient set includes N transformation coefficients, and the transformation coefficient is any color space component in the currently decoded image , where N is a positive integer;

an acquisition unit, configured to acquire decoded pixel information of the image area, where the decoded pixel information includes spatial neighborhood pixel information of the image area;

a calculation unit, configured to calculate a quantization adjustment factor according to the decoded pixel information;

an inverse quantization unit, configured to obtain a first quantization step size corresponding to the first transform coefficient set from a signal corresponding to the current decoded image; scaling the first quantization step size by using the quantization adjustment factor, obtaining a second quantization step size, and using the second quantization step size to inversely quantize the first transform coefficient set to obtain the second transform coefficient set;

The calculation unit is specifically configured to calculate the quantization adjustment factor according to the first function f ₁ (P _avg ) of the pixel mean value P _avg ; wherein, the pixel mean value P _avg is the luminance component of a plurality of pixels corresponding to the decoded pixel information The mean value obtained by averaging, or the mean value obtained by averaging the same color space component of the plurality of pixels, and the color space component is the color space component corresponding to the first transformation coefficient set;

f ₁ (P _avg ) may be a U-shaped function with respect to P _avg , and f ₁ (P _avg ) satisfies that the first derivative of f ₁ (P _avg ) is less than 0 when P _avg is less than the threshold T1, and when P _avg is greater than the threshold T2 The first derivative of f ₁ (P _avg ) is greater than 0, and f ₁ (P _avg ) is equal to a constant C0 when P _avg is between the thresholds T1 and T2; wherein T1≥0, T2≥0, T2≥T1.

16. The apparatus according to claim 15, wherein the obtaining unit is further configured to obtain the temporal neighborhood pixel information of the image area and/or the pixel information of the image area in the decoded pixel information .

17. A device for adaptive inverse quantization in video coding, characterized in that the device comprises:

The calculation unit is specifically configured to calculate the quantization adjustment factor according to the second function f ₂ (P _var ) of the pixel variance P _var ; wherein, the pixel variance P _var is the difference between the luminance components of the plurality of pixels corresponding to the decoded pixel information. variance, or the variance of the same color space component of the plurality of pixels, and the color space component is the color space component corresponding to the first transformation coefficient set;

f ₂ (P _var ) is a monotonic function about P _var , f ₂ (P _var ) satisfies that when (P _var ) ^α is less than the threshold T3, f ₂ (P _var ) is a constant C3, and when (P _var ) ^α is greater than or equal to the threshold value T3 At threshold T3, the first derivative of f ₂ (P _var ) is greater than zero.

18. A device for adaptive inverse quantization in video coding, characterized in that the device comprises:

The calculation unit is specifically configured to calculate the quantization adjustment factor according to the joint function f ₁ (P _avg )·f ₂ (P _var ) of the pixel variance P _var and the pixel mean value P _avg _; The mean value obtained by averaging the luminance components of multiple pixels in the decoded pixel information, or the mean value obtained by averaging the same color space component of the multiple pixels; the pixel variance P _var is the luminance component of the multiple pixels , or the variance of the same color space component of the plurality of pixels, where the color space component is the color space component corresponding to the first transformation coefficient set.

19. A device for adaptive inverse quantization in video coding, characterized in that the device comprises:

The calculation unit is specifically configured to calculate the quantization adjustment factor according to the third function f ₃ (P _max -P _min ) of the difference between the pixel maximum value P _max and the pixel minimum value P _min ; wherein the pixel maximum value P _max is The maximum value of the luminance components of the plurality of pixels corresponding to the decoded pixel information, or the maximum value of the same color space component of the plurality of pixels; the minimum value of the pixel is the minimum value of the luminance components of the plurality of pixels , or the minimum value of the same space component of the plurality of pixels, where the color space component is the color space component corresponding to the first transform coefficient set.

20. The apparatus according to any one of claims 15-19, wherein the inverse quantization unit is specifically configured to determine a target level scale table corresponding to the quantization adjustment factor according to a preset first correspondence; wherein , the first correspondence is the correspondence between the quantization adjustment factor and the level scale table; the first transform coefficient set is inversely quantized according to the target level scale table to obtain the second transform coefficient set.

21. The apparatus according to any one of claims 15-19, wherein the inverse quantization unit is specifically configured to determine a target quantization matrix corresponding to the quantization adjustment factor according to a preset second correspondence; wherein, the The second corresponding relationship is the corresponding relationship between the quantization adjustment factor and the quantization matrix; the first transformation coefficient set is inversely quantized according to the target quantization matrix to obtain the second transformation coefficient set.

22. The apparatus according to any one of claims 15-19, wherein the inverse quantization unit is specifically configured to obtain the first quantization corresponding to the first transform coefficient set from the signal corresponding to the current decoded image parameter; determine a target quantization parameter offset according to the quantization adjustment factor, add the first quantization parameter and the target quantization parameter offset to obtain a second quantization parameter; determine a level according to the second quantization parameter scale, and inversely quantizes the first transform coefficient set by using the level scale to obtain the second transform coefficient set.

23. The apparatus according to any one of claims 15-19, wherein the inverse quantization unit is specifically configured to obtain the first quantization corresponding to the first transform coefficient set from the signal corresponding to the current decoded image step size; use the first quantization step size to inverse quantize the coefficients in the first transform coefficient set to obtain the coefficients in the third transform coefficient set; use the calculated quantization adjustment factor to quantify the third transform coefficient set The coefficients are scaled to obtain a second set of transform coefficients.