An intelligent processing method for ultrasonic images

Abstract

The invention relates to the technical field of image processing, and in particular to an intelligent processing method for ultrasonic images, comprising: obtaining a grayscale image of a liver ultrasonic image; obtaining all filter windows under a plurality of window sizes on the grayscale image; obtaining an optimal initial global filter window according to a global first discrete degree; obtaining a window adjustment parameter according to a foreground area probability, a second discrete degree and an average area of ​​a foreground connected domain; obtaining an adaptive filter window size of each pixel point according to the window adjustment parameter; filtering and denoising the grayscale image according to the adaptive filter window sizes of all pixels to obtain a denoised grayscale image; the invention makes the edge segmentation of the lesion area clearer and improves the efficiency of ultrasonic image segmentation.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to an intelligent processing method for ultrasonic images. Background Art

[0002] Due to the different acoustic impedances of different tissues in the human body, the sound reflected by different tissues will also be different. Ultrasound can reflect the acoustic signals of different tissues and show different brightness. In a two-dimensional sonogram, the intensity of echo reflected by human tissue is usually divided into four levels, namely high echo, equal echo, low echo, and no echo. Compared with X-rays, CT, etc., ultrasound has no ionizing radiation and fast imaging speed, and has a better imaging effect on soft tissues and blood flow. At present, image segmentation is very important in medical diagnosis and can be used for visualization analysis, image-guided radiotherapy and surgery, etc.

[0003] However, ultrasound images have problems such as high speckle noise, low signal-to-noise ratio and uneven intensity. The existing technology handles these problems in the following way: mean filtering denoising. The mean filtering denoising processing method is to calculate the average value of the pixels around each pixel (including the point) as the value of the pixel after filtering. Usually, all pixels in a rectangular window centered on the pixel are used to calculate the average value. When performing mean filtering, the larger the mean filtering window selected, the better the noise reduction effect, but the more serious the loss of image edge details; if the selected window is smaller, the more edge details are preserved, but the noise reduction effect is worse. However, the traditional mean filter uses a fixed-size template to process ultrasound images, which will cause the denoising effect to be unclear or the image edge to be seriously lost. Summary of the invention

[0004] The present invention provides an intelligent processing method for ultrasonic images to solve the existing problems.

[0005] An intelligent processing method for ultrasonic images of the present invention adopts the following technical solution:

[0006] An embodiment of the present invention provides an intelligent processing method for ultrasound images, the method comprising the following steps:

[0007] Obtaining a grayscale image of the liver ultrasound image;

[0008] Obtain all filter windows under several window sizes on the grayscale image; obtain the global first discrete degree under each window size on the grayscale image according to the grayscale value distribution difference of the pixel points in the filter window under each window size on the grayscale image; obtain the optimal initial global filter window according to the global first discrete degree;

[0009] According to the distribution of pixels in the optimal initial global filtering window, the probability of the foreground area in the optimal initial global filtering window, the second discrete degree in the optimal initial global filtering window, and the average area of ​​the foreground connected domain in the optimal initial global filtering window are obtained; the window adjustment parameters are obtained according to the foreground area probability, the second discrete degree, and the average area of ​​the foreground connected domain; the adaptive filtering window size of each pixel is obtained according to the window adjustment parameters;

[0010] The grayscale image is filtered and denoised according to the adaptive filter window size of all pixels to obtain a denoised grayscale image.

[0011] Preferably, the step of obtaining all filter windows of several window sizes on the grayscale image includes the following specific steps:

[0012] Preset a window parameter , record any pixel in the grayscale image as the target pixel, Starting with a step size of 2, the target pixel is used as the center pixel to construct The filter window size is as follows: When the value is greater than half of the side length of the grayscale image, the increment is stopped, and all filter windows under several window sizes on the grayscale image are obtained.

[0013] Preferably, the step of obtaining the global first discrete degree at each window size on the grayscale image according to the grayscale value distribution difference of the pixel points in the filter window at each window size on the grayscale image comprises the following specific steps:

[0014] For a filter window of any window size on the grayscale image, the cumulative sum of the grayscale value variances of all pixels in each filter window of the window size is taken as the global first discrete degree of the window size.

[0015] Preferably, the step of obtaining the optimal initial global filtering window according to the global first discrete degree includes the following specific steps:

[0016] The window size corresponding to the smallest first discrete degree among the global first discrete degrees under all window sizes is recorded as the target window size; and the filtering window under the target window size is taken as the optimal initial global filtering window.

[0017] Preferably, the method for obtaining the foreground area probability within the optimal initial global filtering window is as follows:

[0018] The Otsu threshold method is used to obtain the Otsu threshold K of the grayscale image. For any optimal initial global filtering window, the ratio of the number of pixels with grayscale values ​​greater than K in the optimal initial global filtering window to the number of all pixels in the optimal initial global filtering window is used as the foreground area probability in the optimal initial global filtering window.

[0019] Preferably, the method for obtaining the second discrete degree within the optimal initial global filtering window is as follows:

[0020] For any optimal initial global filtering window, the standard deviation of the grayscale values ​​of all pixels in the optimal initial global filtering window is used as the second discrete degree in the optimal initial global filtering window.

[0021] Preferably, the method for obtaining the average area of ​​the foreground connected domain within the optimal initial global filtering window is as follows:

[0022] For any optimal initial global filtering window, the number of connected domains composed of pixels with grayscale values ​​greater than K in the optimal initial global filtering window is , the calculation expression of the average area of ​​the foreground connected domain within the optimal initial global filtering window is:

[0023]

[0024] In the formula, Represents the average area of ​​the connected domain greater than the threshold K in the optimal initial global filtering window; Indicates the pixel points with grayscale values ​​greater than K in the optimal initial global filtering window. The area of ​​a connected domain.

[0025] Preferably, the step of obtaining the window adjustment parameter according to the foreground area probability, the second discrete degree and the average area of ​​the foreground connected domain comprises the following specific steps:

[0026] For the optimal initial global filtering window of any pixel point in the grayscale image, the product of the global first discrete degree under the window size corresponding to the optimal initial global filtering window, the second discrete degree within the optimal initial global filtering window, and the inverse of the average area of ​​the foreground connected domain within the optimal initial global filtering window is recorded as the window adjustment parameter under the optimal initial global filtering window size of the pixel point.

[0027] Preferably, the step of obtaining the adaptive filtering window size of each pixel point according to the window adjustment parameter comprises the following specific steps:

[0028] Starting with the optimal initial global filtering window, the optimal initial global filtering window is recorded as ,Will Increase in steps of 2 until The size of several global filter windows is obtained when the size of the global filter window is greater than half of the side length of the grayscale image. For any pixel point in the grayscale image, the window adjustment parameters of the pixel point under different global filter window sizes are obtained, and the global filter window size corresponding to the minimum value of the window adjustment parameter is used as the adaptive filter window size of the pixel point.

[0029] Preferably, filtering and denoising the grayscale image according to the adaptive filtering window size of all pixels to obtain the denoised grayscale image includes the following specific steps:

[0030] The adaptive filtering window size of all pixels is input into the mean filtering algorithm, and the grayscale image is filtered to obtain the denoised grayscale image.

[0031] The beneficial effect of the technical solution of the present invention is: the present invention utilizes the global changes of the image to obtain the optimal global filtering window, and adaptively adjusts the window size of each pixel point according to the pixel characteristics of the window to perform noise reduction, so that the edge segmentation of the lesion area is clearer, thereby improving the efficiency of ultrasonic image segmentation. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0033] Figure 1 The present invention is a flowchart of the steps of an intelligent processing method for ultrasonic images. DETAILED DESCRIPTION

[0034] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation, structure, features and effects of an intelligent processing method for ultrasonic images proposed by the present invention in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0035] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0036] The specific scheme of the intelligent processing method for ultrasonic images provided by the present invention is described in detail below with reference to the accompanying drawings.

[0037] See also Figure 1 , which shows a flowchart of a method for intelligent processing of ultrasound images provided by an embodiment of the present invention, the method comprising the following steps:

[0038] Step S001: Acquire a grayscale image of a liver ultrasound image.

[0039] Specifically, an ultrasonic device is used to scan the liver region of a human body to acquire a liver ultrasonic image. The acquired image is grayed out using a weighted average method to obtain a grayscale image of the liver ultrasonic image. In this embodiment, the size of the liver ultrasonic image is 1024×1024 as an example.

[0040] Step S002: Obtain all filter windows under several window sizes on the grayscale image; obtain the global first discrete degree under each window size on the grayscale image according to the grayscale value distribution difference of the pixel points in the filter window under each window size on the grayscale image; obtain the optimal initial global filter window according to the global first discrete degree.

[0041] It should be noted that one of the reasons for whether image segmentation can achieve a good segmentation effect depends on the amount of noise. The less noise, the better the segmentation effect, and the more noise, the worse the segmentation effect. Therefore, this embodiment analyzes the pixel features of the image to achieve an adaptive mean filter window size for each element point of the image to achieve the best noise reduction effect.

[0042] However, before adjusting the filter window size, it is necessary to first select an optimal global filter window as the initial window size, and then adjust the window size according to the difference characteristics of the gray value distribution of pixels in the window.

[0043] It should be further explained that by customizing a global mean filter window, the size is , calculate the first discrete degree of the pixel value in the window. The higher the first discrete degree, the greater the difference in the grayscale value distribution of the pixels in the window, indicating that the grayscale values ​​of the pixels in the window are less similar; the lower the first discrete degree, the smaller the difference in the grayscale value distribution of the pixels in the window, indicating that the grayscale values ​​of the pixels in the window are more similar. That is, the overall grayscale value distribution is relatively concentrated, and the image attributes in this area may be relatively simple, may not belong to the junction of the foreground and background, and may not have some important grayscale change information. Therefore, for the selection of a global filter window, the lower the first discrete degree, the better the filtering effect of the window of this size.

[0044] Preset a window parameter , wherein this embodiment is based on This example is described as an example, and this embodiment is not specifically limited. It depends on the specific implementation situation.

[0045] Specifically, any pixel in the grayscale image is recorded as the target pixel. Starting with a step size of 2, the target pixel is used as the center pixel to construct The filter window size is as follows: When the value is greater than half of the side length of the grayscale image, the increment is stopped, and all filter windows under several window sizes on the grayscale image are obtained.

[0046] It should be noted that, by comparing the sum of the variances of the pixel values ​​in each window in the image, when the sum of the first discrete degrees of the pixel values ​​in each window is the smallest, it means that the window has a high probability of placing pixels with similar pixel features in the window, indicating that the global filtering window at this time is the best, and then the window size corresponding to the minimum value of the sum of the first discrete degrees of all windows can be selected as the optimal global filtering window.

[0047] The specific method to obtain the optimal initial global filtering window is:

[0048] For a filter window of any window size on the grayscale image, the cumulative sum of the grayscale value variances of all pixels in each filter window of the window size is taken as the global first discrete degree of the window size.

[0049] At this point, the global first discrete degree under all window sizes is obtained.

[0050] Specifically, the window size corresponding to the smallest first discrete degree among the global first discrete degrees under all window sizes is recorded as the target window size; and the filtering window under the target window size is used as the optimal initial global filtering window.

[0051] At this point, the optimal initial global filtering window of the grayscale image of the liver ultrasound image is obtained.

[0052] Step S003: Obtain the foreground area probability within the optimal initial global filtering window, the second discrete degree within the optimal initial global filtering window, and the average area of ​​the foreground connected domain within the optimal initial global filtering window; obtain window adjustment parameters based on the foreground area probability, the second discrete degree, and the average area of ​​the foreground connected domain; obtain the adaptive filtering window size of each pixel point based on the window adjustment parameters.

[0053] 1. Obtain the probability of the foreground area within the optimal initial global filter window.

[0054] It should be noted that, because the denoised image needs to be segmented, analyzing the threshold segmented image that needs denoising can achieve better denoising effect. The filter window adjustment parameters are obtained by analyzing the pixel distribution and number of pixels in the optimal initial global filter window, so as to select the appropriate filter window for denoising.

[0055] It should be further explained that ultrasound images utilize the characteristics of ultrasound propagation, reflection and scattering in tissues. The reflection and scattering characteristics of different tissues are different. The lesion can be regarded as a kind of body tissue to some extent, but there may be other tissues or water accumulation in the scanning area that affect the reflection and scattering of ultrasound. This is reflected in the grayscale image, which means that other parts of the tissue with lower or higher grayscale values ​​may be misclassified as foreground or background. Therefore, the Otsu method can be used to roughly distinguish the image into foreground and background. The Otsu method may also be used to segment the image later. Therefore, when denoising, the threshold of the Otsu method can be used as the initial condition to denoise the image foreground and background, so as to better achieve the segmentation effect.

[0056] By using the Otsu method to process the liver ultrasound grayscale image, we can obtain the Otsu threshold K and the processed image, which can roughly divide the lesion area. Then, the pixels in the image with grayscale values ​​greater than K are most likely classified as background pixels, while the pixels with grayscale values ​​less than K are most likely classified as foreground areas.

[0057] Because all pixels within the optimal initial global filter window are divided into foreground pixels and background pixels according to the Otsu threshold method, the window adjustment parameters can be obtained according to the number of pixels that may be divided into foreground pixels and the number of pixels that may be divided into background pixels.

[0058] It should be noted that for any optimal initial global filtering window, the window adjustment parameters are obtained according to the size and number of pixels in the window. The more pixels greater than K in the filtering window, the greater the probability that the window is in the background area; and the more pixels less than K in the window, the higher the probability that the window is in the foreground area.

[0059] Specifically, the Otsu threshold method is used to obtain the Otsu threshold K of the grayscale image. For any optimal initial global filtering window, the ratio of the number of pixels with grayscale values ​​greater than K in the optimal initial global filtering window to the number of all pixels in the optimal initial global filtering window is used as the probability of the foreground area in the optimal initial global filtering window.

[0060] The larger the proportion in the filter window, the greater the probability of suspected background or noise area in the window, so the filter window should be increased. The probability of the foreground area in the optimal initial global filtering window is recorded as the window adjustment influencing parameter .

[0061] At this point, the probability of the foreground area within the optimal initial global filtering window is obtained.

[0062] 2. Obtain the second discrete degree within the optimal initial global filtering window.

[0063] It should be noted that the number of pixels greater than the threshold K in the window cannot fully determine the amount of noise. Therefore, in the optimal initial global filtering window, the window with a large pixel change rate has a greater probability of noise. Therefore, the size of the filtering window can be adjusted according to the pixel change rate.

[0064] It should be further explained that the standard deviation is a statistic used to measure the degree of data dispersion in a data set. The standard deviation can be used to determine the distribution of pixel grayscale values ​​in the optimal initial global filter window. If the standard deviation is larger, it means that the pixel distribution difference in the optimal initial global filter window is greater, and the probability of it being noise is greater, so the window should be enlarged for noise reduction. In the optimal initial global filter window, the greater the pixel change rate, the greater the probability of noise in the window, that is, the greater the second discrete degree of the pixel value in the window, the greater the probability of noise. On the contrary, the smaller the second discrete degree, the smaller the probability of noise.

[0065] Specifically, for any optimal initial global filtering window, the standard deviation of the grayscale values ​​of all pixels in the optimal initial global filtering window is used as the second discrete degree in the optimal initial global filtering window.

[0066] At this point, the second discrete degree within the optimal initial global filtering window is obtained.

[0067] 3. Obtain the average area of ​​the foreground connected domain within the optimal initial global filtering window.

[0068] It should be noted that the large difference in pixel changes may also be caused by the window being at the edge of the foreground image, but the connected domain formed by noise is generally small, so the window size can be adjusted by analyzing the size of the connected domain that is greater than the Otsu threshold.

[0069] Specifically, for any optimal initial global filtering window, the number of connected domains composed of pixels with grayscale values ​​greater than K in the optimal initial global filtering window is recorded as , the calculation expression of the average area of ​​the foreground connected domain within the optimal initial global filtering window is:

[0070]

[0071] In the formula, Represents the average area of ​​the connected domain greater than the threshold K in the optimal initial global filtering window; Indicates the pixel points with grayscale values ​​greater than K in the optimal initial global filtering window. The area of ​​a connected domain.

[0072] It should be noted that for the grayscale image of the liver ultrasound image, the noise points are small and the distribution form is small and numerous. If the average area of ​​the connected domain greater than the threshold K in the optimal initial global filter window is smaller, the probability of a large amount of noise in the window is greater; if the average area of ​​the connected domain greater than the threshold K in the optimal initial global filter window is larger, there may be important edge information in the window. That is, the smaller the average area of ​​the connected domain greater than the threshold K in the optimal initial global filter window, the greater the probability of a large amount of noise, and the larger the window needs to be adjusted; so take the inverse; and then obtain the window adjustment influencing parameter based on the average area of ​​the connected domain greater than the threshold K in the optimal initial global filter window. .

[0073] At this point, the average area of ​​the foreground connected domain in the optimal initial global filter window is obtained, and the window adjustment influencing parameter is obtained according to the average area of ​​the foreground connected domain in the optimal initial global filter window. .

[0074] 4. Adjust the influencing parameters according to the window , and Get window adjustment parameters.

[0075] Specifically, the window adjustment parameter is obtained according to the window adjustment influencing parameter, and the calculation expression of the window adjustment parameter is:

[0076]

[0077] In the formula, is the window size adjustment parameter, , and are the influencing parameters obtained above respectively.

[0078] So far, as long as a window size is determined, the adjustment parameters of the window size can be obtained using the above method.

[0079] It should be noted that this embodiment uses the global optimal filter window as the standard window to process pixels, and a window size restriction is given to avoid unlimited window expansion. In order to limit the window from unlimited expansion, the maximum value of the global filter window size depends on the image size and the computer memory size. Generally speaking, for most computers and images, the maximum value of the global filter window size can be half of the image size or less.

[0080] Specifically, starting from the optimal initial global filtering window, the optimal initial global filtering window is recorded as ,Will Increase in steps of 2 until The size of several global filter windows is obtained when the size of the global filter window is greater than half of the side length of the grayscale image. For any pixel point in the grayscale image, the window adjustment parameters of the pixel point under different global filter window sizes are obtained, and the global filter window size corresponding to the minimum value of the window adjustment parameter is used as the adaptive filter window size of the pixel point.

[0081] So far, the adaptive size of each window has been calculated through the above, and the adaptive sizes of all windows in the filtering process can be obtained similarly.

[0082] Step S004: filtering and denoising the grayscale image according to the adaptive filtering window size of all pixels to obtain a denoised grayscale image.

[0083] The adaptive filtering window size of all pixels is input into the mean filtering algorithm, and the grayscale image is filtered to obtain the denoised grayscale image. The denoised grayscale image is segmented using the OTSU algorithm, and the area greater than the Otsu threshold K is marked as 1; the area less than the threshold K is marked as 0, thereby obtaining a segmented liver ultrasound image.

[0084] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for intelligent segmentation of ultrasonic images, characterized in that: The method comprises the following steps: Obtaining a grayscale image of the liver ultrasound image; Obtain all filter windows under several window sizes on the grayscale image; obtain the global first discrete degree under each window size on the grayscale image according to the grayscale value distribution difference of the pixel points in the filter window under each window size on the grayscale image; obtain the optimal initial global filter window according to the global first discrete degree; According to the distribution of pixels in the optimal initial global filtering window, the probability of the foreground area in the optimal initial global filtering window, the second discrete degree in the optimal initial global filtering window, and the average area of ​​the foreground connected domain in the optimal initial global filtering window are obtained; the window adjustment parameters are obtained according to the foreground area probability, the second discrete degree, and the average area of ​​the foreground connected domain; the adaptive filtering window size of each pixel is obtained according to the window adjustment parameters; Filter and denoise the grayscale image according to the adaptive filter window size of all pixels to obtain a denoised grayscale image; The step of obtaining the global first discrete degree at each window size on the grayscale image according to the grayscale value distribution difference of the pixel points in the filter window at each window size on the grayscale image includes the following specific steps: For a filter window of any window size on the grayscale image, the cumulative sum of the grayscale value variances of all pixels in each filter window of the window size is taken as the global first discrete degree of the window size; The specific steps of obtaining the optimal initial global filtering window according to the global first discrete degree are as follows: The window size corresponding to the smallest first discrete degree among the global first discrete degrees under all window sizes is recorded as the target window size; the filtering window under the target window size is taken as the optimal initial global filtering window; The method for obtaining the second discrete degree within the optimal initial global filtering window is as follows: For any optimal initial global filtering window, the standard deviation of the grayscale values ​​of all pixels in the optimal initial global filtering window is used as the second discrete degree in the optimal initial global filtering window; The specific steps of obtaining the window adjustment parameter according to the foreground area probability, the second discrete degree and the average area of ​​the foreground connected domain are as follows: For an optimal initial global filtering window of any pixel in the grayscale image, the product of the global first discrete degree under the window size corresponding to the optimal initial global filtering window, the second discrete degree within the optimal initial global filtering window, and the reciprocal of the average area of ​​the foreground connected domain within the optimal initial global filtering window is recorded as the window adjustment parameter under the optimal initial global filtering window size of the pixel; The specific steps of obtaining the adaptive filtering window size of each pixel point according to the window adjustment parameter are as follows: Starting with the optimal initial global filtering window, the optimal initial global filtering window is recorded as ,Will Increase in steps of 2 until The size of several global filter windows is obtained when the size of the global filter window is greater than half of the side length of the grayscale image. For any pixel point in the grayscale image, the window adjustment parameters of the pixel point under different global filter window sizes are obtained, and the global filter window size corresponding to the minimum value of the window adjustment parameter is used as the adaptive filter window size of the pixel point.

2. According to claim 1, the method for intelligent segmentation of ultrasonic images is characterized in that: The specific steps of obtaining all filter windows of several window sizes on the grayscale image are as follows: Preset a window parameter , record any pixel in the grayscale image as the target pixel, Starting with a step size of 2, the target pixel is used as the center pixel to construct The filter window size is as follows: When the value is greater than half of the side length of the grayscale image, the increment is stopped, and all filter windows under several window sizes on the grayscale image are obtained.

3. The method for intelligent segmentation of ultrasonic images according to claim 1, characterized in that: The method for obtaining the foreground area probability within the optimal initial global filtering window is as follows: The Otsu threshold method is used to obtain the Otsu threshold K of the grayscale image. For any optimal initial global filtering window, the ratio of the number of pixels with grayscale values ​​greater than K in the optimal initial global filtering window to the number of all pixels in the optimal initial global filtering window is used as the foreground area probability in the optimal initial global filtering window.

4. According to claim 3, the method for intelligent segmentation of ultrasonic images is characterized in that: The method for obtaining the average area of ​​the foreground connected domain within the optimal initial global filtering window is as follows: For any optimal initial global filtering window, the number of connected domains composed of pixels with grayscale values ​​greater than K in the optimal initial global filtering window is , the calculation expression of the average area of ​​the foreground connected domain within the optimal initial global filtering window is: In the formula, Represents the average area of ​​the connected domain greater than the threshold K in the optimal initial global filtering window; Indicates the pixel points with grayscale values ​​greater than K in the optimal initial global filtering window. The area of ​​a connected domain.

5. The method for intelligent segmentation of ultrasonic images according to claim 1, characterized in that: The specific steps of filtering and denoising the grayscale image according to the adaptive filtering window size of all pixels to obtain the denoised grayscale image are as follows: The adaptive filtering window size of all pixels is input into the mean filtering algorithm, and the grayscale image is filtered to obtain the denoised grayscale image.

Images