CN111028258B - An Adaptive Threshold Extraction Method for Large-Scale Grayscale Images - Google Patents
An Adaptive Threshold Extraction Method for Large-Scale Grayscale Images Download PDFInfo
- Publication number
- CN111028258B CN111028258B CN201911112512.6A CN201911112512A CN111028258B CN 111028258 B CN111028258 B CN 111028258B CN 201911112512 A CN201911112512 A CN 201911112512A CN 111028258 B CN111028258 B CN 111028258B
- Authority
- CN
- China
- Prior art keywords
- value
- gray
- peak
- image
- 1stpk
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
本发明实施例公开了一种大尺度灰度图像自适应阈值提取方法，先矩阵化读入系列大尺度灰度图像，对符合要求的图像依次进行维纳滤波和高斯滤波处理，得到图像样品；然后统计图像样品的灰度值信息，得到图像样品的灰度直方分布图，读取系列图像的灰度图进行直方图的累加；再根据累加的直方分布进行求其梯度，得到图像样品不同的梯度分布数据；接着根据图像样品的灰度直方分布图及其梯度分布数据结果，进行灰度分布特征值求解，并由经验公式计算得到分割阈值；最后计算分割阈值的可调节范围，提供进行手动调节的范围限制。本发明的方法具有良好的自适应能力，全过程自动化进行，极大地提高了识别提取工作效率和识别准确度。
The embodiment of the present invention discloses an adaptive threshold value extraction method for large-scale grayscale images. Firstly, a series of large-scale grayscale images are read in a matrix, and Wiener filtering and Gaussian filtering are sequentially performed on the images that meet the requirements to obtain image samples; Then count the gray value information of the image sample to obtain the gray histogram distribution map of the image sample, read the gray map of a series of images to accumulate the histogram; then calculate the gradient according to the accumulated histogram distribution, and obtain the different values of the image sample Gradient distribution data; then according to the gray histogram of the image sample and its gradient distribution data results, the gray distribution characteristic value is solved, and the segmentation threshold is calculated by the empirical formula; finally, the adjustable range of the segmentation threshold is calculated and provided for manual Adjustment range limit. The method of the invention has good self-adaptive ability, and the whole process is carried out automatically, which greatly improves the work efficiency of recognition and extraction and the recognition accuracy.
Description
Technical Field
The embodiment of the invention relates to the technical field of digital image processing, in particular to a large-scale gray level image self-adaptive threshold value extraction method with obvious four-component characteristics.
Background
The image threshold segmentation technology refers to a technology for distinguishing different components in an image according to different thresholds, the traditional threshold segmentation mainly comprises two types, one type is a manual adjustment method, and the image is subjected to experimental adjustment of different segmentation thresholds through visual identification to obtain an optimal segmentation threshold; the other is an automatic threshold segmentation method represented by the Ojin method, and based on the concept of maximum inter-class difference, two-component segmentation of an image is performed.
The manual adjustment method can sometimes obtain a relatively reliable threshold value, but needs to repeatedly divide the test process, and the process has too much subjective judgment, so that inaccuracy is caused; on the other hand, the segmentation test often takes a long time, and has no economy and operability; meanwhile, the manual adjustment method has operability for small-scale single images, and for large-scale processing objects with series of pictures, accurate adjustment cannot be performed at all.
The Ojin method based on maximum inter-class difference judgment has relatively good stability and operability, is an important method for performing two-component identification, but is difficult to perform large-scale four-component (abstractly, the four-class macroscopic-sortable characteristic) analysis, and is mainly because as a matrix is increased, the calculation amount is increased in a nonlinear manner, and meanwhile, the threshold value of the four-component characteristic is calculated by the two-component method, the repeated iterative process is brought, the calculation amount is increased further, and certain inaccuracy is brought. Threshold segmentation is only the first step of image recognition and extraction, and the calculation amount brought by the current method even exceeds the recognition extraction itself. It is therefore necessary to design a fast segmentation method for a four-component image.
The large-scale image can have two forms, one is a single image, but the pixel matrix is larger; the other is that the single image pixel matrix is small but has a large number of series of images. The method can achieve quick and quick results for the treatment of the two forms.
Disclosure of Invention
Therefore, the embodiment of the invention provides a large-scale gray scale image self-adaptive threshold extraction method with obvious four-component characteristics, which aims to solve the problems in the prior art.
In order to achieve the above object, the embodiments of the present invention provide the following technical solutions: a large-scale gray level image self-adaptive threshold value calculating method comprises the following steps:
step 500, segmentation threshold prediction: according to the solved characteristic values, calculating to obtain a segmentation threshold value by an empirical formula;
As a preferable scheme of the invention, the image sample is a shale sample, and the characteristic values comprise peak values of organic matters, inorganic matters and pyrite of the shale sample, corresponding gray values, a first-order extreme value, a second-order extreme value and a third-order extreme value on the left side and the right side of the shale sample, and a total of 22 characteristic points of the shale sample.
As a preferred embodiment of the present invention, the wiener filtering method includes:
generating a filtering template with a matrix size of 3 multiplied by 3 or 5 multiplied by 5, calculating an average value and a variance of 8 or 24 pixels around a target point, wherein the target point is positioned at the center of the filtering template, creating a pixel matrix wiener filter by using the average value and the variance of the pixel gray scales in the filtering template, and calculating the average value of the gray scales in the pixel matrix of the filtering template by using the wiener filter as the central value of the filtering template matrix, so as to obtain the filtered value of the target point;
the average value is:
the variance is:
the filter template values are:
wherein a is the number (n) 1 ,n 2 ) Pixel gray value, v, of location 2 Is the standard deviation of the noise, or the local deviation value estimated in the template is used, b (n 1 ,n 2 )。
As a preferred embodiment of the present invention, the gaussian filtering method includes:
according to Gaussian distribution characteristics, a filtering template with the size of 3 multiplied by 3 or 5 multiplied by 5 is generated, the original image gray value of the area with the same size as the filtering template, which is close to the center point, is multiplied by the filtering template, the result is used as a value after the center point is filtered, for example, when the value takes 0.8, the generated template is:
and replacing the central value by a Gaussian average value, and sequentially solving the Gaussian average value of the neighborhood for all the pixel points of the image to obtain the filtered image.
As a preferred embodiment of the present invention, the method for extracting the gray scale rectangularity distribution feature of the image in step 200 includes:
sequentially reading in each image in the batch of images, counting the number of pixels with gray values of 0-255, and accumulating the gray distribution data obtained by the batch of images according to the number of gray values to obtain total gray straight distribution data of the sample images; respectively solving a first-order gradient, a second-order gradient and a third-order gradient for the total gray level straight distribution; extracting the total gray level straight distribution and the maximum value and the minimum value of the one-step gradient distribution data, the second-step gradient distribution data and the three-step gradient distribution data, carrying out combined analysis operation, and adaptively obtaining three characteristic thresholds for gray level threshold segmentation, so as to carry out remarkable four-component segmentation, and obtaining the adjustable upper limit value and the adjustable lower limit value of corresponding segmentation gray level according to the characteristics of the image.
As a preferred embodiment of the present invention, the method for extracting the feature distribution value in step 300 includes:
extracting the peak value of gray level straight distribution, finding two maximum value marks on the left side and the right side according to the maximum peak value, obtaining five maximum values altogether, marking the maximum peak value as maxPk, marking the two peak values on the right side as RPk1 and RPk2, and marking the two peak values on the left side as LPk1 and LPk2;
peak value arrangement, namely taking peak value points corresponding to the maximum three peak values from the five extracted peak values, wherein the peak value points are from small to large and are respectively regarded as a first peak value 1stPk, a second peak value 2ndPk and a third peak value 3rdPk; judging that if the minimum peak value is too small and is close to zero, only the maximum two peak values are reserved;
for each peak value, respectively finding the maximum value and the minimum value of the first-order guide corresponding to the left side and the right side, and analyzing; the left and right extreme values of the first peak value are respectively marked as 1stLSp and 1stRSp; the left and right extreme values of the second peak value are respectively marked as 2ndLSp and 2ndRSp, the left and right extreme values of the third peak value are respectively marked as 3rdLSp and 3rdRSp, and if the minimum peak value is zero, the left and right extreme values are also set to be zero;
for each peak, find the maximum value of the second order gradient on the left and right sides respectively as inflection point TF, and TF of each peak is marked as: 2ndrtf,2ndltf,1strtf,1stltf,3rdrtf,3rdltf;
for TF of each point, finding maximum value and minimum value of third-order guide corresponding to left and right sides respectively, analyzing, and marking as: 2ndRKt,2ndLKt,1stRKt,1stLKt,3rdRKt,3rdLKt;
finding a valley value of gray level straight distribution between 1stPk and 2ndPk, and marking the valley value as medvally;
when the third peak value is not present in the gray level image, a virtual peak of the third peak is constructed according to the existing data for the next calculation, and the peak point of the virtual peak and the relevant point are given by an empirical formula:
3rdRPk＝floor(1stPk^(-2)*2+2ndLSp^(-2)*3+2ndPk^0.71*5.79+2ndRSp^(-2)*3)；
3rdLTF＝3rdRPk-14；
3rdLSp＝3rdRPk-10；
3rdRSp＝3rdRPk+10；
3rdRTF＝3rdRPk+14；
when the third peak value is not present in the gray level image, calculating a virtual peak of the third peak according to the existing data for the next calculation, wherein the peak point of the third peak can be given by an empirical formula:
1stPk＝round(2ndLSp^0.13*119.56-2ndPk^(-5.09)*0.31-2ndRSp^(-0.81)*8541.39)；
1stLSp＝round(1stPk^0.54*9.30+medvally^(-13.97)*0.20+2ndLSp^(0.75)*0.99-2ndPk^(-3.91)*2.89-2ndRSp^(-0.04)*86.37)；
1stRSp＝round(1stLSp^(-0.06)*0.50+1stPk^0.99*1.29+medvally^(-0.31)*0.21-2ndLSp^0.24*10.84+2ndPk^(-0.054)*50.83-2ndRSp^(-3.74)*26.14-3rdPk^1.09*0.034)；
1stLTF＝round(1stLSp^0.94*1.26-1stPk^(-6.81)*0.88-1stRSp^0.45*2.56-medvally^0.24*2.02+2ndPk^0.31*1.39+3rdPk^1.31*0.013)；
1stLKt＝round(1stLTF^1.15*0.23-1stLSp^(-0.0048)*0.77+1stPk^0.76*3.58-1stRSp^0.70*3.50+medvally^(-7.62)*0.63+2ndPk^0.85*0.26)；
1stRTF＝floor(1stPk+(1stPk-1stLTF)*(1stRSp-1stPk)^0.6/(1stPk-1stLSp)^0.6)。
as a preferred embodiment of the invention, the five said maxima and the five said peaks each comprise a zero value.
As a preferred embodiment of the present invention, the empirical formula in step 400:
3rdlevel＝floor(-2ndLSp^0.85*3.78+2ndPk^0.76*6.83-2ndRSp^0.81*0.55+2ndRTF^0.82*3.22+3rdLTF^0.93*0.41-2ndRKt^1.00*0.53+3rdLSp^0.77*2.63-3rdPk^0.98*0.88)；
2ndlevel＝floor(1stPk^1.03*1.50-medvally^1.02*0.15+2ndLKt^1.09*0.73-2ndLTF^0.92*0.48-2ndLSp^0.98*3.32+2ndPk^0.94*6.50-2ndRSp^0.94*6.62+2ndRTF^0.98*2.39+2ndRKt^0.91*0.026)；
1stlevel＝floor(-1stLKt^(-1.48)*5.10+1stLTF^(2.37)*0.00069-1stLSp^0.30*54.92+1stPk^0.30*58.25-1stRSp^1.77*0.0032+medvally^0.31*10.41+2ndPk^(-0.40)*0.78)。
as a preferred embodiment of the present invention, the method for calculating the adjustable range of the segmentation threshold in step 500 includes:
the upper and lower limits of the first threshold are respectively marked as u1stlevel and d1stlevel, and the calculation empirical formula is as follows:
u1stlevel＝1stLTF；
u1stlevel(u1stlevel<＝1stlevel)＝1stlevel+5；
d1stlevel＝1stLKt-4；
d1stlevel(d1stlevel>＝1stlevel)＝1stlevel-5；
the upper and lower limiting intervals of the second threshold are respectively marked as u2ndlevel and d2ndlevel, and the calculation empirical formula is as follows:
u2ndlevel＝floor(medvally*0.5+2ndLKt*0.5)；
u2ndlevel(u2ndlevel<＝2ndlevel)＝2ndlevel+5；
d2ndlevel＝floor(0.5*1stRSp+0.5*1stRTF)；
d2ndlevel(d2ndlevel>＝2ndlevel)＝2ndlevel-5；
the up-and-down change space of the third threshold is respectively marked as u3rdlevel and d3rdlevel, and the calculation empirical formula is as follows:
u3rdlevel＝floor(2ndRKt*0.6+3rdLSp*0.4)；
u3rdlevel(u3rdlevel<＝3rdlevel)＝3rdlevel+7；
d3rdlevel＝floor(2ndRSp*0.7+3rdLTF*0.3)；d3rdlevel(d3rdlevel>＝3rdlevel)＝3rdlevel-7。
embodiments of the present invention have the following advantages:
the method has good self-adaptive capability, and can automatically identify the gray distribution characteristics of the image according to the characteristics of the image, so that the characteristic parameters are extracted according to the characteristics of the four-component image, the whole process is automatically carried out, the low efficiency and the uncertainty caused by manual setting are avoided, and the identification and extraction working efficiency and the identification accuracy are greatly improved.
The invention can process and extract batch files, process large-scale images and obtain the common gray threshold value of the series of images (large-scale images), which has higher credibility than the processing result of single images.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below. It will be apparent to those of ordinary skill in the art that the drawings in the following description are exemplary only and that other implementations can be obtained from the extensions of the drawings provided without inventive effort.
The structures, proportions, sizes, etc. shown in the present specification are shown only for the purposes of illustration and description, and are not intended to limit the scope of the invention, which is defined by the claims, so that any structural modifications, changes in proportions, or adjustments of sizes, which do not affect the efficacy or the achievement of the present invention, should fall within the ambit of the technical disclosure.
FIG. 1 is a flow chart of an algorithm implemented by the present invention;
FIG. 2 is a gray scale view of a shale scanning electron microscope in accordance with an embodiment of the present invention;
FIG. 3 is a graph of shale gray scale orthometric distribution in accordance with an embodiment of the present invention;
FIG. 4 is a graph of a first order gradient of shale gray scale orthometric distribution in accordance with an embodiment of the present invention;
FIG. 5 is a graph of a shale gray scale orthometric distribution second order gradient in an embodiment of the invention;
FIG. 6 is a third-order gradient chart of shale gray scale orthometric distribution in an embodiment of the invention;
fig. 7 is a graph of extraction results of various components of shale in accordance with an embodiment of the present invention.
Detailed Description
Other advantages and advantages of the present invention will become apparent to those skilled in the art from the following detailed description, which, by way of illustration, is to be read in connection with certain specific embodiments, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
As shown in fig. 1 to 7, the invention provides a large-scale gray image adaptive threshold calculation method, which is mainly applied to the identification and processing process of image components with obvious four components, and adopts a digital image processing technology, in particular to a gray straight distribution statistical method to automatically analyze and judge image characteristics so as to adaptively obtain the segmentation threshold of an image. The method comprises the following steps:
step 500, segmentation threshold prediction: according to the solved characteristic values, calculating to obtain a segmentation threshold value by an empirical formula;
In this embodiment, the image sample may be a shale sample, and the feature values include peak values of organic matters, inorganic matters and pyrite of the shale sample, and first-order extremum, second-order extremum and third-order extremum on left and right sides of the peak values, and a total of 22 feature points added with valley values between the organic matters and the inorganic matters.
In step 100, the sample image is filtered, first using Wiener (Wiener) adaptive filtering, and then using Gaussian (Gaussian) filtering. Wiener filtering is an optimal filtering system that can be used to extract signals contaminated with stationary noise. The wiener filtering method comprises the following steps:
generating a filtering template with a matrix size of 3 multiplied by 3 or 5 multiplied by 5, calculating an average value and a variance of 8 or 24 pixels around a target point, wherein the target point is positioned at the center of the filtering template, creating a pixel matrix wiener filter by using the average value and the variance of the pixel gray scales in the filtering template, and calculating the average value of the gray scales in the pixel matrix of the filtering template by using the wiener filter as the central value of the filtering template matrix, so as to obtain the filtered value of the target point;
the average value is:
the variance is:
the filter template values are:
wherein a is the number (n) 1 ,n 2 ) Pixel gray value, v, of location 2 Is the standard deviation of the noise, or the local deviation value estimated in the template is used, b (n 1 ,n 2 )。
In this step, the method of gaussian filtering specifically includes:
according to Gaussian distribution characteristics, a filtering template with the size of 3 multiplied by 3 or 5 multiplied by 5 is generated, the original image gray value of the area with the same size as the filtering template, which is close to the center point, is multiplied by the filtering template, the result is used as a value after the center point is filtered, for example, when the value takes 0.8, the generated template is:
and replacing the central value by a Gaussian average value, and sequentially solving the Gaussian average value of the neighborhood for all the pixel points of the image to obtain the filtered image.
In step 200, the specific method for extracting the image gray scale rectangularity distribution feature is as follows:
sequentially reading in each image in the batch of images, counting the number of pixels with gray values of 0-255, and accumulating the gray distribution data obtained by the batch of images according to the number of gray values to obtain total gray straight distribution data of the sample images; respectively solving a first-order gradient, a second-order gradient and a third-order gradient for the total gray level straight distribution; extracting the total gray level straight distribution and the maximum value and the minimum value of the one-step gradient distribution data, the second-step gradient distribution data and the three-step gradient distribution data, carrying out combined analysis operation, and adaptively obtaining three characteristic thresholds for gray level threshold segmentation, so as to carry out remarkable four-component segmentation, and obtaining the adjustable upper limit value and the adjustable lower limit value of corresponding segmentation gray level according to the characteristics of the image.
The method for extracting the feature distribution value in step 300 specifically includes:
extracting peaks of gray level straight distribution, finding two maximum value marks on the left side and the right side according to the maximum peak value, obtaining five maximum values (which can be zero values), marking the maximum peak value as maxPk, marking the two peaks on the right side as RPk and RPk2, and marking the two peaks on the left side as LPk1 and LPk2;
peak value arrangement, namely taking peak value points corresponding to the maximum three peak values from five peak values (which can be zero values) obtained through extraction, wherein the peak value points are from small to large and are respectively regarded as a first peak value 1stPk, a second peak value 2ndPk and a third peak value 3rdPk; judging that if the minimum peak value is too small and is close to zero, only the maximum two peak values are reserved;
for each peak value, respectively finding the maximum value and the minimum value of the first-order guide corresponding to the left side and the right side, and analyzing; the left and right extreme values of the first peak value are respectively marked as 1stLSp and 1stRSp; the left and right extreme values of the second peak value are respectively marked as 2ndLSp and 2ndRSp, the left and right extreme values of the third peak value are respectively marked as 3rdLSp and 3rdRSp, and if the minimum peak value is zero, the left and right extreme values are also set to be zero;
for each peak, find the maximum value of the second order gradient on the left and right sides respectively as inflection point TF, and TF of each peak is marked as: 2ndrtf,2ndltf,1strtf,1stltf,3rdrtf,3rdltf;
for TF of each point, finding maximum value and minimum value of third-order guide corresponding to left and right sides respectively, analyzing, and marking as: 2ndRKt,2ndLKt,1stRKt,1stLKt,3rdRKt,3rdLKt;
finding a valley value of gray level straight distribution between 1stPk and 2ndPk, and marking the valley value as medvally;
when the third peak value is not present in the gray level image, a virtual peak of the third peak is constructed according to the existing data for the next calculation, and the peak point of the virtual peak and the relevant point are given by an empirical formula:
3rdRPk＝floor(1stPk^(-2)*2+2ndLSp^(-2)*3+2ndPk^0.71*5.79+2ndRSp^(-2)*3)；
3rdLTF＝3rdRPk-14；
3rdLSp＝3rdRPk-10；
3rdRSp＝3rdRPk+10；
3rdRTF＝3rdRPk+14；
when the third peak value is not present in the gray level image, calculating a virtual peak of the third peak according to the existing data for the next calculation, wherein the peak point of the third peak can be given by an empirical formula:
1stPk＝round(2ndLSp^0.13*119.56-2ndPk^(-5.09)*0.31-2ndRSp^(-0.81)*8541.39)；
1stLSp＝round(1stPk^0.54*9.30+medvally^(-13.97)*0.20+2ndLSp^(0.75)*0.99-2ndPk^(-3.91)*2.89-2ndRSp^(-0.04)*86.37)；
1stRSp＝round(1stLSp^(-0.06)*0.50+1stPk^0.99*1.29+medvally^(-0.31)*0.21-2ndLSp^0.24*10.84+2ndPk^(-0.054)*50.83-2ndRSp^(-3.74)*26.14-3rdPk^1.09*0.034)；
1stLTF＝round(1stLSp^0.94*1.26-1stPk^(-6.81)*0.88-1stRSp^0.45*2.56-medvally^0.24*2.02+2ndPk^0.31*1.39+3rdPk^1.31*0.013)；
1stLKt＝round(1stLTF^1.15*0.23-1stLSp^(-0.0048)*0.77+1stPk^0.76*3.58-1stRSp^0.70*3.50+medvally^(-7.62)*0.63+2ndPk^0.85*0.26)；
1stRTF＝floor(1stPk+(1stPk-1stLTF)*(1stRSp-1stPk)^0.6/(1stPk-1stLSp)^0.6)。
in step 400, since 22 basic feature values required for calculating the gray segmentation threshold have been obtained, image gray value prediction is performed by using a certain empirical formula, and the empirical formula is obtained from the images of the 3 classes of different samples, and has a larger representativeness, and the specific empirical formula is as follows:
3rdlevel＝floor(-2ndLSp^0.85*3.78+2ndPk^0.76*6.83-2ndRSp^0.81*0.55+2ndRTF^0.82*3.22+3rdLTF^0.93*0.41-2ndRKt^1.00*0.53+3rdLSp^0.77*2.63-3rdPk^0.98*0.88)；
2ndlevel＝floor(1stPk^1.03*1.50-medvally^1.02*0.15+2ndLKt^1.09*0.73-2ndLTF^0.92*0.48-2ndLSp^0.98*3.32+2ndPk^0.94*6.50-2ndRSp^0.94*6.62+2ndRTF^0.98*2.39+2ndRKt^0.91*0.026)；
1stlevel＝floor(-1stLKt^(-1.48)*5.10+1stLTF^(2.37)*0.00069-1stLSp^0.30*54.92+1stPk^0.30*58.25-1stRSp^1.77*0.0032+medvally^0.31*10.41+2ndPk^(-0.40)*0.78)。
in step 500, according to the basic segmentation threshold obtained in the previous step, an adjustable range of the segmentation threshold, that is, a recommended accuracy space, is calculated, and the adjustable range is obtained according to an empirical formula, specifically as follows:
the upper and lower limits of the first threshold are respectively marked as u1stlevel and d1stlevel, and the calculation empirical formula is as follows:
u1stlevel＝1stLTF；
u1stlevel(u1stlevel<＝1stlevel)＝1stlevel+5；
d1stlevel＝1stLKt-4；
d1stlevel(d1stlevel>＝1stlevel)＝1stlevel-5；
the upper and lower limiting intervals of the second threshold are respectively marked as u2ndlevel and d2ndlevel, and the calculation empirical formula is as follows:
u2ndlevel＝floor(medvally*0.5+2ndLKt*0.5)；
u2ndlevel(u2ndlevel<＝2ndlevel)＝2ndlevel+5；
d2ndlevel＝floor(0.5*1stRSp+0.5*1stRTF)；
d2ndlevel(d2ndlevel>＝2ndlevel)＝2ndlevel-5；
the up-and-down change space of the third threshold is respectively marked as u3rdlevel and d3rdlevel, and the calculation empirical formula is as follows:
u3rdlevel＝floor(2ndRKt*0.6+3rdLSp*0.4)；
u3rdlevel(u3rdlevel<＝3rdlevel)＝3rdlevel+7；
d3rdlevel＝floor(2ndRSp*0.7+3rdLTF*0.3)；d3rdlevel(d3rdlevel>＝3rdlevel)＝3rdlevel-7。
in this embodiment, the software for image processing may be MATLAB, or may be implemented by other software.
The method has good self-adaptive capability, and can automatically identify the gray distribution characteristics of the image according to the characteristics of the image, so that the characteristic parameters are extracted according to the characteristics of the four-component image, the whole process is automatically carried out, the low efficiency and the uncertainty caused by manual setting are avoided, and the identification and extraction working efficiency and the identification accuracy are greatly improved. The method has good universality and can be used for extracting the segmentation threshold value of the typical image sample with four-component characteristics such as rock scanning electron microscope images.
On the other hand, the invention can process and extract batch files, process large-scale images and obtain the common gray threshold value of the series of images (large-scale images), which has higher credibility than the processing result of single images.
The method also has extremely high calculation speed, and in statistics and calculation based on MatLab, the main time consumption is that the gray scale of the extracted image is distributed in a straight direction, and the identification time is about 2.7 seconds for the samples of 50 images in total; the recognition extraction time is about 6.8 seconds for a total of 150 image samples, and about 13.5 seconds for a total of 305 image samples.
The method has better robustness, can obtain three segmentation thresholds more accurately for the image with clearer and visible four-component distribution, and provides the threshold range with adjustable results.
While the invention has been described in detail in the foregoing general description and specific examples, it will be apparent to those skilled in the art that modifications and improvements can be made thereto. Accordingly, such modifications or improvements may be made without departing from the spirit of the invention and are intended to be within the scope of the invention as claimed.
Claims (7)
1. The adaptive threshold value calculating method for the large-scale gray level image is characterized by comprising the following steps of:
step 100, image preprocessing: matrixing and reading in a series of large-scale images, cutting out images which do not meet the requirements, and then sequentially carrying out wiener filtering and Gaussian filtering on the images which meet the requirements to obtain an image sample;
step 200, extracting image gray level straight side distribution: counting gray value information of the image samples to obtain gray straight distribution data of the image samples, reading series of image samples in batches, and calculating an accumulation result of the gray straight distribution data;
step 300, obtaining data of one-step gradient distribution, second-order gradient distribution and three-step gradient distribution corresponding to the image sample by a data gradient solving method according to the accumulated square distribution data;
the method for extracting the feature distribution value in the step 300 includes:
extracting the peak value of gray level straight distribution, finding two maximum value marks on the left side and the right side according to the maximum peak value, obtaining five maximum values altogether, marking the maximum peak value as maxPk, marking the two peak values on the right side as RPk1 and RPk2, and marking the two peak values on the left side as LPk1 and LPk2;
peak value arrangement, namely taking peak value points corresponding to the maximum three peak values from the five extracted peak values, wherein the peak value points are from small to large and are respectively regarded as a first peak value 1stPk, a second peak value 2ndPk and a third peak value 3rdPk; judging that if the minimum peak value is too small and is close to zero, only the maximum two peak values are reserved;
for each peak value, respectively finding the maximum value and the minimum value of the first-order guide corresponding to the left side and the right side, and analyzing; the left and right extreme values of the first peak value are respectively marked as 1stLSp and 1stRSp; the left and right extreme values of the second peak value are respectively marked as 2ndLSp and 2ndRSp, the left and right extreme values of the third peak value are respectively marked as 3rdLSp and 3rdRSp, and if the minimum peak value is zero, the left and right extreme values are also set to be zero;
for each peak, find the maximum value of the second order gradient on the left and right sides respectively as inflection point TF, and TF of each peak is marked as: 2ndrtf,2ndltf,1strtf,1stltf,3rdrtf,3rdLTF;
for TF of each point, finding maximum value and minimum value of third-order guide corresponding to left and right sides respectively, analyzing, and marking as: 2ndRKt,2ndLKt,1stRKt,1stLKt,3rdRKt,3rdLKt;
finding a valley value of gray level straight distribution between 1stPk and 2ndPk, and marking the valley value as medvally;
when the third peak value is not present in the gray level image, a virtual peak of the third peak is constructed according to the existing data for the next calculation, and the peak point of the virtual peak and the relevant point are given by an empirical formula:
3rdRPk=floor(1stPk^(-2)*2+2ndLSp^(-2)*3+2ndPk^0.71*5.79+2ndRSp^(-2)*3)；
3rdLTF=3rdRPk-14；
3rdLSp=3rdRPk-10；
3rdRSp=3rdRPk+10；
3rdRTF=3rdRPk+14；
1stPk=round(2ndLSp^0.13*119.56-2ndPk^(-5.09)*0.31-2ndRSp^(-0.81)*8541.39)；
1stLSp=round(1stPk^0.54*9.30+medvally^(-13.97)*0.20+2ndLSp^(0.75)*0.99-2ndPk^(-3.91)*2.89-2ndRSp^(-0.04)*86.37)；
1stRSp=round(1stLSp^(-0.06)*0.50+1stPk^0.99*1.29+medvally^(-0.31)*0.21-2ndLSp^0.24*10.84+2ndPk^(-0.054)*50.83-2ndRSp^(-3.74)*26.14-3rdPk^1.09*0.034)；
1stLTF=round(1stLSp^0.94*1.26-1stPk^(-6.81)*0.88-1stRSp^0.45*2.56-medvally^0.24*2.02+2ndPk^0.31*1.39+3rdPk^1.31*0.013);
1stLKt=round(1stLTF^1.15*0.23-1stLSp^(-0.0048)*0.77+1stPk^0.76*3.58-1stRSp^0.70*3.50+medvally^(-7.62)*0.63+2ndPk^0.85*0.26);
1stRTF=floor(1stPk+(1stPk-1stLTF)*(1stRSp-1stPk)^0.6/(1stPk-1stLSp)^0.6)；
step 400, extracting feature points: carrying out eigenvalue solving according to gray level straight distribution data and one-step gradient distribution, second-order gradient distribution and three-step gradient distribution data of the image sample;
step 500, segmentation threshold prediction: according to the solved characteristic values, calculating to obtain a segmentation threshold value by an empirical formula;
the empirical formula in step 500:
3rdlevel=floor(-2ndLSp^0.85*3.78+2ndPk^0.76*6.83-2ndRSp^0.81*0.55+2ndRTF^0.82*3.22+3rdLTF^0.93*0.41-2ndRKt^1.00*0.53+3rdLSp^0.77*2.63-3rdPk^0.98*0.88);
2ndlevel=floor(1stPk^1.03*1.50-medvally^1.02*0.15+2ndLKt^1.09*0.73-2ndLTF^0.92*0.48-2ndLSp^0.98*3.32+2ndPk^0.94*6.50-2ndRSp^0.94*6.62+2ndRTF^0.98*2.39+2ndRKt^0.91*0.026);
1stlevel=floor(-1stLKt^(-1.48)*5.10+1stLTF^(2.37)*0.00069-1stLSp^0.30*54.92+1stPk^0.30*58.25-1stRSp^1.77*0.0032+medvally^0.31*10.41+2ndPk^(-0.40)*0.78)；
step 600, calculating an adjustable range of the segmentation threshold, providing a range limit for manual adjustment.
2. The method for extracting the self-adaptive threshold value of the large-scale gray level image according to claim 1, wherein the image sample is a shale sample, and the characteristic values comprise peak corresponding gray values of organic matters, inorganic matters and pyrite of the shale sample, and a first-order extreme value, a second-order extreme value and a third-order extreme value on the left side and the right side of the peak corresponding gray values, and a valley value between the organic matters and the inorganic matters is added for 22 total characteristic points.
3. The method for adaptively extracting a threshold value from a large-scale gray scale image according to claim 1, wherein said wiener filtering method comprises:
generating a filtering template with a matrix size of 3 multiplied by 3 or 5 multiplied by 5, calculating an average value and a variance of 8 or 24 pixels around a target point, wherein the target point is positioned at the center of the filtering template, creating a pixel matrix wiener filter by using the average value and the variance of the pixel gray scales in the filtering template, and calculating the average value of the gray scales in the pixel matrix of the filtering template by using the wiener filter as the central value of the filtering template matrix, so as to obtain the filtered value of the target point;
the average value is:
；
the variance is:
；
the filter template values are:
；
wherein the method comprises the steps of
Is +.>
Pixel gray value of position +.>
Is the standard deviation of the noise, or adopts the local deviation value estimated in the template, +.>
。
4. The method for adaptively extracting a threshold value from a large-scale gray scale image according to claim 1, wherein said gaussian filtering method comprises:
according to Gaussian distribution characteristics, a filtering template with the size of 3 multiplied by 3 or 5 multiplied by 5 is generated, the same as wiener filtering, the original image gray value of the area with the same size as the filtering template and adjacent to the center point is multiplied by the filtering template, the result is taken as the value after the center point is filtered, and when the value is 0.8, the generated template is as follows:
；
and replacing the central value by a Gaussian average value, and sequentially solving the Gaussian average value of the neighborhood for all the pixel points of the image to obtain the filtered image.
5. The method for extracting an automatic threshold value of a large-scale gray scale image according to claim 1, wherein said step 200 of extracting image gray scale straight distribution features comprises:
sequentially reading in each image in the batch of images, counting the number of pixels with gray values of 0-255, and accumulating the gray distribution data obtained by the batch of images according to the number of gray values to obtain total gray straight distribution data of the sample images; respectively solving a first-order gradient, a second-order gradient and a third-order gradient for the total gray level straight distribution; extracting the total gray level straight distribution and the maximum value and the minimum value of the one-step gradient distribution data, the second-step gradient distribution data and the three-step gradient distribution data, carrying out combined analysis operation, and adaptively obtaining three characteristic thresholds for gray level threshold segmentation, so as to carry out remarkable four-component segmentation, and obtaining the adjustable upper limit value and the adjustable lower limit value of corresponding segmentation gray level according to the characteristics of the image.
6. The method of claim 1, wherein five of said maxima and five of said peaks each comprise a zero value.
7. The method for adaptively extracting a threshold value from a large-scale gray-scale image according to claim 2, wherein the method for calculating the adjustable range of the segmentation threshold value in step 600 comprises:
the upper and lower limits of the first threshold are respectively marked as u1stlevel and d1stlevel, and the calculation empirical formula is as follows:
u1stlevel=1stLTF；
u1stlevel(u1stlevel<=1stlevel)=1stlevel+5；
d1stlevel=1stLKt-4；
d1stlevel(d1stlevel>=1stlevel)=1stlevel-5；
the upper and lower limiting intervals of the second threshold are respectively marked as u2ndlevel and d2ndlevel, and the calculation empirical formula is as follows:
u2ndlevel=floor(medvally*0.5+2ndLKt*0.5)；
u2ndlevel(u2ndlevel<=2ndlevel)=2ndlevel+5；
d2ndlevel=floor(0.5*1stRSp+0.5*1stRTF)；
d2ndlevel(d2ndlevel>=2ndlevel)=2ndlevel-5；
the up-and-down change space of the third threshold is respectively marked as u3rdlevel and d3rdlevel, and the calculation empirical formula is as follows:
u3rdlevel=floor(2ndRKt*0.6+3rdLSp*0.4)；
u3rdlevel(u3rdlevel<=3rdlevel)=3rdlevel+7；
d3rdlevel=floor(2ndRSp*0.7+3rdLTF*0.3)； d3rdlevel(d3rdlevel>=3rdlevel)=3rdlevel-7。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911112512.6A CN111028258B (en) | 2019-11-14 | 2019-11-14 | An Adaptive Threshold Extraction Method for Large-Scale Grayscale Images |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911112512.6A CN111028258B (en) | 2019-11-14 | 2019-11-14 | An Adaptive Threshold Extraction Method for Large-Scale Grayscale Images |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111028258A CN111028258A (en) | 2020-04-17 |
| CN111028258B true CN111028258B (en) | 2023-05-16 |
Family
ID=70200138
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911112512.6A Active CN111028258B (en) | 2019-11-14 | 2019-11-14 | An Adaptive Threshold Extraction Method for Large-Scale Grayscale Images |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111028258B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115661135B (en) * | 2022-12-09 | 2023-05-05 | 山东第一医科大学附属省立医院(山东省立医院) | A Lesion Region Segmentation Method for Cardiocerebral Angiography |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6335980B1 (en) * | 1997-07-25 | 2002-01-01 | Arch Development Corporation | Method and system for the segmentation of lung regions in lateral chest radiographs |
| CN105261017A (en) * | 2015-10-14 | 2016-01-20 | 长春工业大学 | Method for extracting regions of interest of pedestrian by using image segmentation method on the basis of road restriction |
| CN109509199A (en) * | 2018-10-10 | 2019-03-22 | 华南理工大学 | A kind of medical image Organizational Intelligence dividing method based on three-dimensional reconstruction |
| CN110033458A (en) * | 2019-03-12 | 2019-07-19 | 中国矿业大学 | It is a kind of based on pixel gradient distribution image threshold determine method |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1213592C (en) * | 2001-07-31 | 2005-08-03 | 佳能株式会社 | Adaptive two-valued image processing method and equipment |
| US8861814B2 (en) * | 2010-12-22 | 2014-10-14 | Chevron U.S.A. Inc. | System and method for multi-phase segmentation of density images representing porous media |
-
2019
- 2019-11-14 CN CN201911112512.6A patent/CN111028258B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6335980B1 (en) * | 1997-07-25 | 2002-01-01 | Arch Development Corporation | Method and system for the segmentation of lung regions in lateral chest radiographs |
| CN105261017A (en) * | 2015-10-14 | 2016-01-20 | 长春工业大学 | Method for extracting regions of interest of pedestrian by using image segmentation method on the basis of road restriction |
| CN109509199A (en) * | 2018-10-10 | 2019-03-22 | 华南理工大学 | A kind of medical image Organizational Intelligence dividing method based on three-dimensional reconstruction |
| CN110033458A (en) * | 2019-03-12 | 2019-07-19 | 中国矿业大学 | It is a kind of based on pixel gradient distribution image threshold determine method |
Non-Patent Citations (3)
| Title |
|---|
| 利用多重-大津阈值算法和扫描电镜分割CT图像;高衍武;吴伟;张虔;赵燕红;邵广辉;李国利;毛超杰;;长江大学学报(自然科学版);16(04);全文 * |
| 基于DTM阈值分割法的孔裂隙煤岩体瓦斯渗流数值模拟;王刚;杨鑫祥;张孝强;李文鑫;史林肯;;岩石力学与工程学报;35(01);全文 * |
| 基于图像像素间空间信息的加权模糊阈值分割算法;于勇;郑姣;郭希娟;;计算机应用与软件;30(03);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111028258A (en) | 2020-04-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114332026B (en) | Visual detection method and device for scratch defects on surface of nameplate | |
| CN116630813B (en) | An intelligent detection system for highway pavement construction quality | |
| CN107038416B (en) | A Pedestrian Detection Method Based on Improved HOG Feature of Binary Image | |
| CN112614062A (en) | Bacterial colony counting method and device and computer storage medium | |
| CN102222226A (en) | Priori analysis based iterative method for segmenting characters of licence plate | |
| CN105825169A (en) | Road-image-based pavement crack identification method | |
| CN110687122A (en) | Method and system for detecting surface cracks of ceramic tile | |
| CN111429372A (en) | Method for enhancing edge detection effect of low-contrast image | |
| CN105447489B (en) | A kind of character of picture OCR identifying system and background adhesion noise cancellation method | |
| CN103488986A (en) | Method for segmenting and extracting characters in self-adaptation mode | |
| CN106204617A (en) | Adaptive Image Binarization Method Based on Cyclic Shift of Residual Image Histogram | |
| CN117094975A (en) | Steel surface defect detection methods, devices and electronic equipment | |
| CN115731493A (en) | Rainfall micro physical characteristic parameter extraction and analysis method based on video image recognition | |
| CN109460722A (en) | A kind of license plate intelligent identification Method | |
| CN117557572B (en) | Highway construction quality detection method based on computer vision | |
| CN115187788A (en) | Crop seed automatic counting method based on machine vision | |
| CN111028258B (en) | An Adaptive Threshold Extraction Method for Large-Scale Grayscale Images | |
| CN114627463B (en) | A non-contact power distribution data identification method based on machine identification | |
| CN116433978A (en) | Automatic generation and automatic labeling method and device for high-quality flaw image | |
| CN113643290B (en) | Straw counting method and device based on image processing and storage medium | |
| CN119693359B (en) | Vacuum printed circuit board defect detection system based on machine vision | |
| CN112528741B (en) | State identification method and device for transformer substation pressure plate | |
| CN116863463B (en) | Egg assembly line rapid identification and counting method | |
| CN115937052B (en) | Gel electrophoresis image processing method, device, equipment and medium | |
| CN110717465A (en) | Identification method and system based on calcium oxalate crystallization |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |