CN110033458A - It is a kind of based on pixel gradient distribution image threshold determine method - Google Patents

Abstract

The invention discloses an image threshold determination method based on pixel gradient distribution. For an image with a unimodal distribution of grayscale distribution histogram, calculate the gradient of each pixel in the image, perform statistical analysis on the obtained gradient value, draw the pixel gradient distribution histogram of the image, and based on the pixel gradient distribution histogram The morphology determines the optimal segmentation threshold. The invention can accurately and quickly determine the segmentation threshold of the image whose gray scale distribution histogram is in a unimodal distribution, thereby improving the segmentation effect of the image.

Description

An Image Threshold Determination Method Based on Pixel Gradient Distribution

technical field

The invention belongs to the field of digital image information extraction, and particularly relates to an image threshold determination method based on pixel gradient distribution.

Background technique

Thanks to the rapidly developing digital imaging technology (CT, SEM, FIB/SEM), the microscopic pore structure inside the sample can be visualized. For the image with obvious contrast between the target and the background, that is, its gray distribution histogram shows obvious bimodal or multimodal distribution, the existing algorithm selects the trough point between the two peaks as the optimal threshold, so as to determine the target and background in the image. Segmentation is performed accurately and efficiently. In addition, there are still some images because the target area is smaller than the background area or the grayscale transition between the target and the background is relatively smooth, that is, the grayscale distribution histogram shows an obvious unimodal distribution. Practice has shown that the existing algorithms have large errors in the segmentation of images with a unimodal distribution of gray distribution histograms, and the authenticity and reliability of the segmentation results are low.

SUMMARY OF THE INVENTION

In order to solve the technical problems raised by the above background technology, the present invention proposes an image threshold determination method based on pixel gradient distribution, which can accurately and quickly determine the segmentation threshold of an image whose grayscale distribution histogram exhibits a unimodal distribution, and improve the image segmentation effect. .

In order to realize the above-mentioned technical purpose, the technical scheme of the present invention is:

An image threshold determination method based on pixel gradient distribution. For an image whose grayscale distribution histogram is unimodal, the gradient of each pixel in the image is calculated, and the obtained gradient value is statistically analyzed, and the pixels of the image are drawn. Gradient distribution histogram, and determine the optimal segmentation threshold based on the shape of the pixel gradient distribution histogram.

The preferred solution based on the above-mentioned technical scheme, comprises the following steps:

(1) Select the gradient operator template;

(2) the matrix of the gradient operator template selected in step (1) is moved on the image that needs to carry out gradient calculation, and the gradient value of each pixel point on the image is calculated;

(3) The gradient values of the pixels with different gray values are superimposed and counted, and the gradient distribution histogram is drawn;

(4) The peak area in the histogram of the gradient distribution of the image corresponds to the pixel points in the target and background boundary areas in the image, while the trough area corresponds to the pixel points in the same area category; find out the corresponding peaks and troughs respectively. The gray value of the pixel point, the two gray values have the greatest possibility to be located at the boundary and the background, respectively, and the segmentation threshold is determined accordingly.

Based on the preferred solution of the above technical solution, in step (1), the gradient operator template selects a 3×3 Sobel operator.

Based on the preferred solution of the above technical solution, in step (3), the gradient value under each gray value is averaged to obtain its average gradient value, and a gradient distribution histogram is drawn according to the average gradient value.

Based on the preferred solution of the above technical solution, in step (4), find out the gray value of the pixel point corresponding to the peak and the trough, and take the average value of the two as the optimal segmentation threshold.

The beneficial effects brought by the above technical solutions:

Compared with the existing digital image threshold determination method, the present invention can more accurately and effectively segment the image whose grayscale histogram presents a unimodal distribution by taking into account the gradient characteristics of the pixel neighborhood, which makes up for the existing method. The disadvantages of large technical error and low precision. At the same time, the algorithm involved in the present invention is simple, the calculation amount is small, and the calculation speed is fast.

Description of drawings

Fig. 1 is the method flow chart of the present invention;

Fig. 2 is the test chart provided in the embodiment;

Fig. 3 is a grayscale histogram corresponding to Fig. 2;

Fig. 4 utilizes Otsu algorithm to process the segmentation result diagram of Fig. 2;

Fig. 5 utilizes MaxEntropy algorithm to process the segmentation result diagram of Fig. 2;

Fig. 6 utilizes Valley-Emphasis algorithm to process the segmentation result diagram of Fig. 2;

Fig. 7 is the gradient map that utilizes the present invention to obtain;

Fig. 8 is the gray distribution histogram that utilizes the present invention to obtain;

Fig. 9 is the threshold value diagram that utilizes the present invention to obtain;

FIG. 10 is a graph of segmentation results obtained by the present invention.

Detailed ways

The technical solutions of the present invention will be described in detail below with reference to the accompanying drawings.

For the image whose gray distribution histogram shows a unimodal distribution, a certain property of the pixel neighborhood is introduced to determine the threshold. The present invention selects the gradient of the pixel point as the research object, and research shows that the size of the pixel gradient reflects the size of the difference between different regions. By calculating the gradient of each pixel in the original image and performing statistical analysis on the obtained gradient value, the pixel gradient distribution histogram of the image is drawn, and the optimal segmentation threshold is determined based on the shape of the pixel gradient distribution histogram. As shown in Figure 1, the specific steps are as follows:

Step 1: Selection of gradient operator template

The Sobel operator assigns different weights to the pixels at different positions, so that the obtained gradient values of the pixels are more accurate. Therefore, in this embodiment, a 3×3 Sobel template is selected to perform subsequent pixel point gradient value calculation.

Step 2: Calculation of pixel gradient values

The small matrix of the Sobel operator template is moved on the image that needs to be gradient calculated, and the convolution operation is performed at each pixel position to calculate the gradient value of the pixel.

Step 3: Drawing the gradient distribution histogram

The gradient values of pixels with different gray values are superimposed and counted. Considering that the number of boundary pixels between different regions is small, even if the gradient values are superimposed and calculated, the prominent effect of the boundary points in the histogram is very limited. Therefore, in the present invention, the gradient value under each gray value is averaged to obtain the average gradient value, thereby eliminating the influence of the number of pixels.

Step 4: Determination of the optimal segmentation threshold

The peak area in the image gradient distribution histogram corresponds to the pixels in the boundary area of the target and background in the image, while the trough area corresponds to the pixels in the same area category. The grayscale values of the pixels corresponding to the peaks and troughs are respectively found, and the two grayscale values have the highest probability of being located at the boundary and the background, respectively. In this embodiment, the average value of the two is taken as the determined optimal threshold.

In this embodiment, the method of the present invention is compared with three existing threshold segmentation algorithms--Otsu, MaxEntropy, Valley-Emphasis, so as to verify the effectiveness of the present invention.

Select the 4 pictures (a), (b), (c), (d) shown in Figure 2 as test pictures, and the corresponding grayscale histograms are shown in Figure 3 (the abscissa is the gray value, and the ordinate is the gray value). is the frequency). Figure 4-6 shows the segmentation images of the four test images processed by the Otsu, MaxEntropy, and Valley-Emphasis algorithms in turn. It can be seen that the threshold determined by Otsu and MaxEntropy is too large, resulting in a large number of pixel points being misclassified, while the Valley-Emphasis algorithm is not stable enough, and the segmentation effect of some images is acceptable, and other images exist with Otsu and MaxEntropy algorithm. The same problem means that the robustness of the Valley-Emphasis algorithm is poor.

Then use the present invention to segment the same test image, and the process is as follows.

First, the gradient map of the image is obtained. Brighter pixels represent high gradients, and darker pixels represent low gradients, as shown in Figure 7.

Next, obtain the gradient distribution histogram of the image, as shown in Figure 8 (the abscissa is the gray value, and the ordinate is the average gradient).

Again, determine the position of the highest peak and lowest valley to obtain the optimal threshold, as shown in Figure 9.

Finally, the image is segmented using the determined threshold, and the result is shown in Figure 10.

It can be clearly seen that the segmentation performance of the present invention is obviously better than the segmentation results of other commonly used threshold algorithms, which proves that the present invention is accurate and effective.

The embodiment is only to illustrate the technical idea of the present invention, and cannot limit the protection scope of the present invention. Any changes made on the basis of the technical solution according to the technical idea proposed by the present invention all fall within the protection scope of the present invention. .

Claims (5)

1. a kind of image threshold based on pixel gradient distribution determines method, which is characterized in that be in for intensity profile histogram The image of Unimodal Distribution calculates the gradient of each pixel in image, and for statistical analysis to gradient value obtained, draws The pixel gradient distribution histogram of image, and optimal segmentation threshold is determined based on the form of pixel gradient distribution histogram.

2. determining method based on the image threshold of pixel gradient distribution according to claim 1, which is characterized in that including following Step:

(1) gradient operator template is selected；

(2) matrix of the gradient operator template of step (1) selection is moved on needing the image for carrying out gradient calculating, calculates figure As the gradient value of upper each pixel；

(3) statistics is overlapped to the gradient value of the pixel of different gray values, draws gradient distribution histogram；

(4) peak regions in the gradient distribution histogram of image are corresponding to pixel in target and background borderline region in image Point, and valley regions then correspond to the pixel in the same area classification；Pixel corresponding with wave crest and trough is found out respectively There is a possibility that maximum to be located at boundary and background, determine therefrom that segmentation threshold for the gray value of point, the two gray values.

3. determining method based on the image threshold of pixel gradient distribution according to claim 1, which is characterized in that in step (1) in, the Sobel operator of the gradient operator stencil-chosen 3 × 3.

4. determining method based on the image threshold of pixel gradient distribution according to claim 1, which is characterized in that in step (3) in, the gradient value under each gray value is averaging processing, its average gradient value is obtained, is drawn according to average gradient value Gradient distribution histogram.

5. determining method based on the image threshold of pixel gradient distribution according to claim 1, which is characterized in that in step (4) in, the gray value of pixel corresponding with wave crest and trough is found out, takes the mean value of the two as optimal segmenting threshold.