Medical Image Quality Assessment Method and System Based on Image Processing

Abstract

This invention relates to the field of image quality assessment technology, specifically to a medical image quality assessment method and system based on image processing. The method includes the following steps: extracting grayscale and directional changes to generate a texture layer; analyzing edge slope changes to extract the blur range; calculating texture grayscale contours to generate a quality factor; dividing spatial segments to form a grade map; and identifying paths to generate a quality degradation chain assessment scheme. In this invention, by analyzing the grayscale directional consistency and multi-scale edge slope changes of image segments, combined with differentiated directional texture refinement and blur jump detection, accurate localization of areas with sudden quality changes is achieved. Based on a comprehensive assessment of texture coverage, grayscale continuity, and contour density, regional structural characteristics are extracted, further completing the spatial distribution tracking and classification of quality degradation chains. This enables image quality expression to possess characteristics such as clear partitioning, well-defined paths, and quantifiable indicators, effectively improving the recognition accuracy and completeness of quality changes in structurally complex regions.

Description

Technical Field

[0001] This invention relates to the field of image quality assessment technology, and in particular to a method and system for medical image quality assessment based on image processing. Background Technology

[0002] Image quality assessment technology is an important branch at the intersection of image processing and computer vision. It primarily studies the automated analysis and judgment of image quality through computational methods. Its core aspects include image sharpness assessment, noise level detection, artifact recognition, and contrast and brightness evaluation. This field is widely used in various image systems such as medical imaging, remote sensing images, and security monitoring. Especially in medical imaging scenarios, image quality directly affects the accuracy of doctors' diagnoses and clinical decisions. Therefore, this field places high demands on automated analysis methods for structural information, texture details, gray-level distribution, and signal-to-noise ratio in images. The overall development of image quality assessment technology has evolved from subjective manual assessment to objective computational models based on feature extraction. Current research mainly focuses on quantitative evaluation methods for indicators such as image segmentation accuracy, texture consistency analysis, and structure preservation capability. Traditional image processing-based medical image quality assessment methods refer to methods that judge image quality by extracting and analyzing statistical features of images acquired by medical imaging equipment, such as gray-level distribution characteristics, edge sharpness variations, noise distribution patterns, and local contrast characteristics. The method typically employs image processing techniques such as gradient histogram analysis, local binary mode operation, image entropy analysis, frequency domain texture energy assessment, and edge detection operator response analysis to analyze structural details in medical images. Based on established image quality judgment criteria, a classification or scoring model is established to achieve automatic image quality assessment.

[0003] Existing technologies, based on statistical modeling of overall image features, tend to use grayscale distribution, noise patterns, and edge responses across the entire image for unified evaluation. However, when images have dense detailed structures, significant blurring and changes in boundary regions, or uneven texture distribution, the lack of fine characterization of quality differences between regions within the image and the ability to track regional change paths leads to quality assessment results that cannot accurately reflect the structural stability and sharpness differences between different regions of the image. For example, in medical images involving multi-layered interwoven structures or significant local texture degradation, problems such as unclear identification of abnormal quality areas and blurred boundary judgments can easily occur, affecting the subsequent clinical reference value and quality control effectiveness of the image. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a medical image quality assessment method and system based on image processing.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a medical image quality assessment method based on image processing, comprising the following steps:

[0006] S1: Divide the medical images into structural segments, extract the central gray value and four-way gray value changes for each segment, determine whether the change amplitude and direction are consistent, select segments with stable gray value changes and consistent directions as texture regions, and generate a reliable texture layer.

[0007] S2: Within the area covered by the reliable texture layer, subdivide the image along the differentiation direction, extract the trend of the slope change of the lower edge of the differentiation scaling layer, compare the edge change speed of adjacent layers, mark the change jump area as the local blur range, and generate a boundary blur distribution map.

[0008] S3: Based on the boundary fuzzy distribution map, mark the region, extract the three elements of texture coverage, gray-level change continuity, and contour density, calculate the structural proportion and distribution characteristics respectively, summarize them into regional structure performance items, and generate image region quality factor group;

[0009] S4: Based on the distribution characteristics of elements in the quality factor group, divide the image spatial segments, count the element indicators of each segment, screen out areas with performance below average, classify them as areas with weak structural quality, and generate a spatial structural quality level map.

[0010] As a further embodiment of the present invention, the reliable texture layer includes the spatial distribution results of the texture region, gray-level change direction information, and gray-level stability features; the boundary blur distribution map includes edge jump position, multi-scale edge slope change, and local blur degree indicator; the image region quality factor group includes texture coverage factor, gray-level continuity factor, and contour density factor; and the spatial structure quality level map includes spatial segment number, quality level label, and low-quality region identifier.

[0011] As a further aspect of the present invention, the specific steps of S1 are as follows:

[0012] S101: Acquire medical image content, divide it into structural segments, extract the coordinates and corresponding gray values ​​of the center pixels of the segments, record the relationship between gray values ​​and positions, and generate the gray value distribution of the center pixels.

[0013] S102: Based on the gray value distribution of the center pixel, extract the gray values ​​of the four adjacent pixels of the center pixel of the segment, calculate the directional gray value difference, and determine whether the gray value change direction is consistent to obtain the directional consistency offset.

[0014] S103: Based on the directional consistency offset, filter segments with grayscale difference values ​​lower than the grayscale change threshold and consistent directions, integrate grayscale information to generate a layer, and obtain a reliable texture layer.

[0015] As a further aspect of the present invention, the specific steps of S2 are as follows:

[0016] S201: Based on the area covered by the reliable texture layer, the image is divided into multiple sub-regions along the gray-level difference direction. The coordinate range and boundary direction of the sub-regions in the original image are extracted, the angle change of the direction is marked, and the boundary angle distribution result of the difference direction is obtained.

[0017] S202: Based on the differential direction boundary angle distribution results, call the image boundary of each subdivided region under the differential scaling layer, extract the edge slope data at the same boundary position, calculate the edge slope change rate between adjacent scaling layers, record the change trend within the region, and obtain the edge change rate feature.

[0018] S203: Based on the edge change rate characteristics, identify regions where the rate difference exceeds the edge jump threshold, locate the regions in the original image coordinates and mark them as blurred boundaries, integrate all marking results to construct a layer, and obtain a blurred boundary distribution layer.

[0019] As a further aspect of the present invention, the specific steps of S3 are as follows:

[0020] S301: Based on the marked areas in the boundary fuzzy distribution layer, extract the texture grayscale change layer in each area, count the number of texture blocks and the range of their arrangement directions within a unit area in the layer, and calculate the structure proportion and the degree of arrangement concentration by combining the area of ​​the region to obtain the texture coverage feature distribution result;

[0021] S302: Based on the texture coverage feature distribution results, extract the gray-level change sequence of continuous pixels in the region, collect and statistically analyze the interval length and amplitude of adjacent gray-level abrupt changes in the sequence, calculate the distribution range and continuity of abrupt changes in the region, and obtain the gray-level abrupt change continuity feature.

[0022] S303: Call the gray-scale abrupt change continuous feature, extract the number and average spacing of the corresponding region boundary line segments, calculate the region contour boundary density distribution and boundary direction clustering, integrate the three types of structural performance results in the region, and uniformly summarize them into a structural factor vector group to obtain the image region quality factor group.

[0023] As a further aspect of the present invention, the specific steps of S4 are as follows:

[0024] S401: Based on the distribution characteristics of the elements in the image region quality factor group, the entire image is divided into coordinate grids, the image is divided into multiple spatial segments at fixed intervals, the quality factor value set of each segment is extracted, and the index mapping of the segment is established to obtain the spatial segment index set.

[0025] S402: Based on the set of spatial segment indicators, calculate the average performance value of the quality factor per unit area in each spatial segment, call the mean of the same indicator of all segments to calculate the overall average performance standard, compare the difference between the indicator performance in a single segment and the standard, and obtain the distribution result of the segment quality deviation degree.

[0026] S403: Based on the distribution results of the degree of deviation of the segment quality, select spatial segments whose performance of all indicators is lower than the average standard, mark the segments in the image coordinates and divide them into grade regions, integrate all coordinate information according to grade, and obtain a spatial structure quality grade map.

[0027] As a further aspect of the present invention, the method further includes:

[0028] S5: Based on the weak areas in the quality grade map, extract the continuous directional path on the image, combine the consistency between the texture sparse trend of adjacent areas and the change of boundary distortion direction, determine whether there is a continuous quality degradation chain extending horizontally and vertically, and generate a medical image quality assessment scheme.

[0029] As a further aspect of the present invention, the medical image quality assessment scheme includes a continuous path, a directional consistency trend, and a weakened chain coverage area.

[0030] As a further aspect of the present invention, the specific steps of S5 are as follows:

[0031] S501: Based on the spatial structure quality level map, detect the set of adjacent coordinates that are continuously distributed in the image coordinates, construct spatial path segments according to the coordinate connection order, and count the distribution of path direction angles and the number of line segments to obtain the distribution result of continuous path direction;

[0032] S502: Based on the distribution results of continuous path direction, call the corresponding texture block density and boundary pixel offset distance in the area covered by the path segment, calculate the gradient value of texture density change according to the path direction, analyze the change direction and the boundary distortion offset direction, determine the consistency of the two changes in the horizontal and vertical directions, and obtain the trend characteristics of the direction consistency.

[0033] S503: Based on the aforementioned consistent trend characteristics, select path regions with directional continuity and quality element consistency characteristics, integrate the coordinate, direction and indicator structure characteristics of the regions, establish an evaluation structure set with path chain properties, and obtain a medical image quality evaluation scheme.

[0034] Image processing-based medical image quality assessment systems include:

[0035] The image texture construction module acquires the image, divides it into structural segments, extracts the center and four-way gray values, determines whether the gray value difference is stable and consistent in direction, filters it as a texture region, and generates a reliable texture layer.

[0036] The region edge difference module subdivides the coverage area based on the reliable texture layer, collects the edge slope at multiple scales, compares the rate of change of adjacent scales, filters slope jump segments, and generates a boundary fuzzy distribution map.

[0037] The local blur determination module calls the blurred region in the boundary blur distribution map, extracts its texture coverage, gray-level abrupt change density, and boundary density, calculates the three ratios, summarizes them as the region representation, and generates the image region quality factor group.

[0038] The spatial segment comparison module divides the spatial segments according to the image region quality factor group, calculates the average of three performance factors for each segment, compares it with the overall average, filters out low-performance segments, and generates a spatial structure quality level map.

[0039] The continuous degradation identification module extracts the texture direction changes and boundary undulation directions based on the continuous distribution areas in the spatial structure quality level map, judges the consistency of direction, filters continuous degradation paths, and generates a medical image quality assessment scheme.

[0040] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0041] In this invention, by analyzing the consistency of gray-level direction and the changes in edge slope at multiple scales in image segments, and combining differentiated directional texture refinement and blur jump detection, the precise location of local quality abrupt change regions is achieved. Based on a comprehensive evaluation of texture coverage, gray-level continuity, and contour density, regional structural performance features are extracted, and the spatial distribution tracking and classification of quality degradation chains are further completed. This enables image quality expression to have characteristics such as clear partitioning, clear path, and quantifiable indicators, effectively improving the recognition accuracy and expression completeness of quality changes in structurally complex regions. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 This is a schematic diagram of the steps of the present invention;

[0044] Figure 2 This is a flowchart of steps S1 of the present invention;

[0045] Figure 3 This is a flowchart of steps S2 of the present invention;

[0046] Figure 4 This is a flowchart of steps S3 of the present invention;

[0047] Figure 5 This is a flowchart of step S4 of the present invention;

[0048] Figure 6 This is a flowchart of steps S5 of the present invention;

[0049] Figure 7 This is a system module diagram of the present invention. Detailed Implementation

[0050] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0051] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0052] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, their intended meanings are consistent. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, their intended meanings are consistent.

[0053] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0054] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0055] Please see Figure 1 A medical image quality assessment method based on image processing includes the following steps:

[0056] S1: Obtain medical image content, divide it into structural segments, extract the gray value of the center position and the gray value changes in the four adjacent directions for each segment, judge the consistency of the change amplitude and direction in the four-directional change trend, and select segments with stable gray value changes and consistent directions as texture regions to generate a reliable texture layer.

[0057] S2: Based on the area covered by the reliable texture layer, subdivide along the differential direction in the image, extract the edge slope change trend under multi-layer image scaling for each subdivided region, compare the edge change speed in adjacent layers, mark the edge change jump area as the local blur range, and generate a boundary blur distribution map.

[0058] S3: Based on the marked regions in the boundary fuzzy distribution map, extract the texture coverage, gray-level abrupt change continuity and contour density in the regions, calculate the structural proportion and distribution characteristics of the three elements respectively, summarize them into regional structural performance items, and generate image region quality factor groups.

[0059] S4: Based on the distribution characteristics of elements in the image region quality factor group, multiple spatial segments are segmented in the whole image. The element indicators in each spatial segment are aggregated and statistically analyzed. Regions with indicator performance lower than the average performance of all segments are screened and classified as regions with weak structural quality, generating a spatial structural quality level map.

[0060] S5: Based on the weak quality regions already classified in the spatial structure quality level map, extract the continuous distribution path on the image, and combine the trend of sparse texture and boundary distortion amplitude in adjacent regions to determine whether there is a continuous quality degradation chain extending laterally and longitudinally, and generate a medical image quality assessment scheme.

[0061] The reliable texture layer includes the spatial distribution results of the texture region, gray-level change direction information, and gray-level stability characteristics. The boundary blur distribution map includes the edge jump position, multi-scale edge slope change, and local blur degree indicator. The image region quality factor group includes texture coverage factor, gray-level continuity factor, and contour density factor. The spatial structure quality level map includes spatial segment number, quality level label, and low-quality region identifier. The medical image quality assessment scheme includes continuous path, directional consistency trend, and weakened chain coverage area.

[0062] Please see Figure 2 The specific steps of S1 are as follows:

[0063] S101: Acquire medical image content, divide it into structural segments, extract the coordinates and corresponding gray values ​​of the center pixels of the segments, record the relationship between gray values ​​and positions, and generate the gray value distribution of the center pixels.

[0064] Acquiring medical image content refers to importing CT or MRI image frames from an image data file. The image reading module converts the raw data into a two-dimensional grayscale image matrix. For example, if an image has a resolution of 512×512 pixels, it is then divided into structural segments using a sliding window. A fixed window size of 32×32 pixels is used, sliding simultaneously in both the horizontal and vertical directions at 16-pixel increments. This process yields 961 overlapping structural segments, each with a central pixel whose coordinates are obtained by adding (15, 15) to the coordinates of the top-left corner of the window. For instance, when the window's starting point is (0, 0), the central pixel coordinates are (15, 15). 5, 15), when the starting point is (16, 0), the center coordinates are (31, 15), and so on. For each center pixel, its gray value is read. If it is an 8-bit image, the gray value range is an integer between 0 and 255. Then, the gray values ​​of these center pixels are recorded and mapped to their corresponding two-dimensional image coordinates to form a gray position correspondence table. Then, the frequency of the gray values ​​of all center pixels is counted to form the gray value distribution. For example, a pixel with a gray value of 100 appears 48 times, a pixel with a gray value of 120 appears 130 times, a pixel with a gray value of 150 appears 89 times, etc. All statistical data are arranged to generate a complete gray distribution data list for subsequent analysis.

[0065] S102: Based on the gray value distribution of the center pixel, extract the gray values ​​of the four adjacent pixels of the center pixel of the segment, calculate the directional gray value difference, and determine whether the gray value change direction is consistent to obtain the directional consistency offset.

[0066] Based on the aforementioned grayscale distribution, for the center pixel of each structural segment, the grayscale values ​​of its four adjacent pixels in the top, bottom, left, and right directions are read, and the grayscale difference between them and the center pixel is calculated. For example, if the grayscale of the center pixel is 128, and the grayscale values ​​of its adjacent pixels in the top, bottom, left, and right directions are 124, 130, 126, and 132 respectively, then the grayscale difference is 4, 2, 2, and 4. Comparing these differences, it is determined whether the direction of grayscale change is consistent. The criterion is whether the grayscale value of the adjacent pixels increases or decreases compared to the center pixel. Here, the top direction decreases, the bottom direction increases, the left direction decreases, and the right direction increases, with the top and left directions being the same. If the trend of change is the same both downward and to the right, it indicates that the direction of change is inconsistent, and the directional consistency of this segment is 50%. In another example, if the center pixel is 100 and the adjacent pixels are 98, 96, 97, and 95, then all four directions decrease, and the directional consistency is 100%. This consistency is used as an important basis for evaluating the local texture features of an image. The directional consistency offset can be obtained by calculating the ratio of the number of consistent directions to the total number of directions. In order to ensure data stability, it is recommended to retain only segments with no less than 3 consistent directions, that is, the offset should be no less than 0.75. This indicator will be used to filter the local structure of the image in subsequent steps.

[0067] S103: Based on the directional consistency offset, filter segments with grayscale difference values ​​lower than the grayscale change threshold and consistent directions, integrate grayscale information to generate a layer, and obtain a reliable texture layer;

[0068] The first step in filtering structural segments based on directional consistency offset is to set a threshold for grayscale difference. This threshold is used to filter segments with minimal grayscale changes. The grayscale change threshold can be set based on the overall average grayscale value of the center pixels. For example, if the average grayscale value obtained from all center pixels is 126, and considering that regions with small local differences in medical images have higher analytical value, a tolerance range of 10% of the average can be set, resulting in a grayscale change threshold of 13. In subsequent filtering, only segments where the grayscale difference between the center pixel and its four adjacent pixels does not exceed 13 are retained. Further filtering is performed using directional consistency offset, retaining only structural fragments with 3 or 4 consistent directions, i.e., offsets of 0.75 or 1. These structural fragments represent regions with stable structural textures in the image. The selected fragments are then arranged and combined according to their original spatial positions in the image. The grayscale information of each fragment is spliced ​​with its neighboring grayscale blocks based on the center, forming a new layer. For example, if a total of 800 qualified fragments are selected from all fragments, they are sequentially merged to generate a new image data with a layer matrix size of 256×256 pixels, which is output as the final reliable texture layer.

[0069] Please see Figure 3 The specific steps of S2 are as follows:

[0070] S201: Based on the area covered by the reliable texture layer, divide the image into multiple sub-regions along the gray-level difference direction, extract the coordinate range and boundary direction of the sub-region in the original image, mark the angle change of the direction, and obtain the boundary angle distribution result of the difference direction.

[0071] Based on the area covered by the reliable texture layer, the image coordinates corresponding to all non-empty areas in the layer are first identified. By performing horizontal and vertical scanning operations on the grayscale value change trend of each pixel, the grayscale difference between adjacent pixels is extracted for each row and column. It is determined whether the grayscale difference is outside a set threshold range. If the grayscale difference between two adjacent pixels exceeds 15, its position is recorded as a change point. The arrangement direction of consecutive change points is then statistically analyzed. By comparing the linear relationship between consecutively changing pixels, the direction of the line segments they form in the image coordinate system is calculated. Further clustering is performed based on direction similarity, grouping line segments with similar direction changes (e.g., angle difference less than 10 degrees) into the same region, thus dividing multiple sub-regions. For example, in the reliable texture region extracted from a 512×512 CT image, 16 sub-regions with different main grayscale change directions are formed. The boundary of each sub-region is extracted using an enclosing rectangle. Extract the coordinates of the top, bottom, left, and right edges and record their start and end positions. For example, the top left corner of a certain region's boundary is (100, 80), the bottom right corner is (140, 120), its horizontal boundary is from x=100 to x=140, and its vertical boundary is from y=80 to y=120. Then, calculate the direction angle of each boundary direction. Specifically, use the three-point method to take three consecutive pixel coordinate points on the boundary line and calculate the direction angle of that line segment. For example, take points (101, 80), (102, 81), and (103, 83) and calculate that the boundary direction is approximately 45 degrees. Then, summarize the direction angles of the four boundaries of the subdivided region and statistically analyze the direction angles of all regions to form a distribution set of boundary direction angles. For example, in the above 16 subdivided regions, there are 6 regions with main boundary direction angles between 40 and 50 degrees, and 4 regions with boundary direction angles between 80 and 90 degrees, thus obtaining the distribution of differential direction boundary angles.

[0072] S202: Based on the results of the differential direction boundary angle distribution, the image boundary of each subdivided region under the differential scaling layer is called, the edge slope data at the same boundary position is extracted, the edge slope change rate between adjacent scaling layers is calculated, and the change trend within the region is recorded to obtain the edge change rate feature.

[0073] Formula for the rate of change of edge slope between adjacent scaling layers:

[0074]

[0075] in, The edge slope change rate characteristic value within the j-th subdivision region represents the rate of change of edge slope between adjacent scaling layers. This represents the edge slope value at the k-th sampling point in the j-th region of scaling layer i. This represents the edge slope value at the corresponding position in scaling layer number m. This refers to the edge angle value of the k-th point in the j-th region of scaling layer i. is the edge angle value of the corresponding point in the scaling layer number m, and n is the total number of sampling points participating in the calculation in the j-th region.

[0076] The data acquisition and parameter assignment are explained below:

[0077] Using actual medical images as a background, between the i=1 and m=2 layers, a subdivided region numbered j=3 was selected, and n=6 edge sampling points were selected within the region for measurement. The obtained slope values ​​and direction angle data are shown in the table.

[0078] Table 1. Slope and Direction Angle Values ​​of Edge Sampling Points (Unit: None)

[0079]

[0080] According to Table 1:

[0081] Each product term:

[0082] Item 1: |6.5-5.8|×|40-35|=0.7×5=3.5;

[0083] Item 2: |7.1-6.4|×|42-37|=0.7×5=3.5;

[0084] Item 3: |6.8-6.0|×|38-34|=0.8×4=3.2;

[0085] Item 4: |7.4-6.6|×|45-40|=0.8×5=4.0;

[0086] Item 5: |6.9-6.2|×|41-36|=0.7×5=3.5;

[0087] Item 6: |7.0-6.5|×|43-38|=0.5×5=2.5;

[0088] The sum of the numerators is: 3.5 + 3.5 + 3.2 + 4.0 + 3.5 + 2.5 = 20.2; the denominator is:

[0089] |6.5×40|=260;

[0090] |7.1×42|=298.2;

[0091] |6.8×38|=258.4;

[0092] |7.4×45|=333.0;

[0093] |6.9 × 41| = 282.9;

[0094] |7.0×43|=301.0;

[0095] Sum: 260 + 298.2 + 258.4 + 333.0 + 282.9 + 301.0 = 1733.5;

[0096] Square root:

[0097] Multiply by n = 6: 6 × 41.63 ≈ 249.78;

[0098] Final calculation:

[0099]

[0100] The results show that, between scaling layer 1 and layer 2, the characteristic value of the edge slope change rate of the third subdivision region is 0.0808, indicating that the slope and direction of this region change to a certain extent at both scales, which is at a medium rate level and corresponds to the basis for determining the edge change rate characteristics in this scheme.

[0101] The advantage of the formula is that by incorporating the product of slope change and directional angle change into the molecule and normalizing it with the overall structural features of the region, it can not only enhance the detection capability of regions where the edge morphology simultaneously undergoes angular shift and intensity fluctuation, but also reduce the interference of single-directional change on the overall evaluation, thus more stably reflecting the change range of structural contour between scaling levels.

[0102] S203: Based on the edge change rate characteristics, identify regions where the rate difference exceeds the edge jump threshold, locate the regions in the original image coordinates and mark them as blurred boundaries, integrate all marking results to construct a layer, and obtain the blurred boundary distribution layer;

[0103] Based on the edge change rate characteristics, each subdivided region is individually assessed for edge abrupt changes. An edge abrupt change threshold is set to identify regions with excessively large rate differences. This threshold is set with reference to the mean μ and standard deviation σ of the edge change rates of all regions in the entire image. The abrupt change threshold T is defined as μ plus 1.5 times σ. For example, if μ is 0.35 and σ is 0.12, then T is 0.53. Regions with an average edge change rate greater than 0.53 are selected from all subdivided regions as abrupt change regions. Then, within each abrupt change region, each sampling point is further sampled to screen for local slope rates greater than 0.53. If such a case exists, its corresponding original image coordinates are marked as a blurred boundary. The marking operation involves setting the pixel value of the point to a special flag value on the original image matrix, such as -1. Finally, all marked points are spatially clustered, and neighboring marked points are merged to form blurred boundary regions. A matrix of the same size as the image is then constructed, and all blurred boundary regions are marked in this matrix. Finally, a boundary blur distribution layer corresponding to the original image is generated.

[0104] Please see Figure 4 The specific steps of S3 are as follows:

[0105] S301: Based on the marked areas in the boundary fuzzy distribution layer, extract the texture grayscale change layer in each area, count the number of texture blocks and the range of their arrangement directions within a unit area in the layer, and calculate the structure proportion and the degree of arrangement concentration by combining the area of ​​the region to obtain the texture coverage feature distribution results;

[0106] Based on the marked regions in the boundary blur distribution layer, firstly, the corresponding coordinates of each blurred region in the layer on the original image are extracted, and all grayscale values ​​within that region are obtained. Then, using a fixed unit area as a standard in each region, for example, 8×8 pixels as a unit block, the number of texture blocks in each unit area is counted. A texture block is defined as a sub-block with a grayscale variation of more than 5 and repeated directional texture edges. By traversing the grayscale gradient directions within each unit block and binning the direction values, the number of texture directions is counted for every 30 degrees. Assuming there are 60 unit blocks in a certain region, the number of texture blocks in each unit block ranges from 0 to 6, with the directions mainly concentrated in the 60-degree and 90-degree directions, and the number of blocks corresponding to each direction is 1. The number of texture blocks is 2 and 22, with a total of 235 texture blocks. The maximum directional distribution is concentrated in the 90-degree direction. Then, the structure ratio is calculated based on the total area of ​​the region. The structure ratio is the ratio of the total area of ​​the texture blocks to the total area of ​​the region. If the average area of ​​the texture blocks is 16 pixels, then the structure area is 235 × 16 = 3760 pixels. If the area of ​​the region is 10000 pixels, then the structure ratio is 0.376. The degree of concentration of directional distribution is then evaluated. The ratio of the number of texture blocks in the direction with the highest frequency to the total number of texture blocks is selected as the directional concentration degree. In the example, there are 22 texture blocks in the 90-degree direction, and the concentration degree is 22 / 235, which is approximately 0.094. This value is used to measure the concentration of texture direction. Finally, the texture coverage feature distribution result of the region is obtained.

[0107] S302: Based on the texture coverage feature distribution results, extract the gray-level change sequence of continuous pixels in the region, collect and statistically analyze the interval length and amplitude of adjacent gray-level abrupt changes in the sequence, calculate the distribution range and continuity of abrupt changes in the region, and obtain the gray-level abrupt change continuity feature.

[0108] Based on the texture coverage feature distribution, grayscale value sequences are extracted from each blurred region. The grayscale values ​​of pixels within the region are traversed row by row along the horizontal direction or the main texture direction, recording the grayscale value sequences of consecutive pixels. Abrupt changes in grayscale values ​​between adjacent pixels are detected. An abrupt change is defined as a difference in grayscale values ​​between two adjacent pixels greater than 10. For example, in a grayscale sequence [100, 102, 105, 107, 140, 142, 145, 110], the change from 105 to 107 is 2 (not an abrupt change), while the change from 107 to 140 is 33 (an abrupt change), so the abrupt change amplitude is recorded as 33. The pixel interval between each abrupt change is statistically analyzed, i.e., the abrupt change interval length. In this example, the abrupt change interval is 1 pixel from 107 to 140. Further analysis is performed within the region... All mutation segments within the domain are accumulated, and the mutation amplitude and corresponding interval length are recorded. These values ​​are then classified and statistically analyzed. For example, the mutation amplitude is divided into three categories: 10–20, 21–40, and 41 and above. The corresponding number and average mutation interval length are counted for each category. For example, there are 15 mutation segments with a mutation amplitude of 21–40, and the average interval is 3.2 pixels. Next, the distribution range of mutation intervals within the region is evaluated, and the difference between the maximum and minimum intervals, the total number of mutation segments, and the mutation density per unit area are calculated to determine whether the mutations are concentrated or discrete. Finally, the degree of mutation continuity is obtained by dividing the number of continuous mutation segments per unit area by the total number of pixels in the region. If this value is greater than 0.05, it means that there is at least one mutation segment in every 20 pixels, which is considered to be a strong mutation continuity. Finally, the gray-scale mutation continuity feature of the region is obtained.

[0109] S303: Call the continuous gray-scale abrupt change feature, extract the number and average spacing of the boundary line segments of the corresponding region, calculate the density distribution and boundary direction aggregation of the region contour boundary, integrate the three types of structural performance results in the region, and summarize them into a structural factor vector group to obtain the image region quality factor group.

[0110] The algorithm utilizes continuous grayscale abrupt change features to extract the contour edges of each region boundary. It then extracts continuous edge pixels within the blurred region boundary using grayscale gradient magnitude analysis. The start and end coordinates of each boundary line segment are recorded, and the number of all line segments within the region is counted (e.g., 12 boundary line segments are extracted). The spacing between each pair of segments is calculated, defined as the Euclidean distance between the midpoints of the line segments. All spacings are counted, and their average is calculated. If the average spacing is 6.8 pixels, this value represents the density of the boundary distribution. Next, based on the directional angle value of each line segment, the directional angle distribution is statistically analyzed, categorizing line segments into groups of 15 degrees each. The proportion of concentrated directions is determined, such as the 45-degree direction. If there are 5 line segments, accounting for 41.7%, and the proportion of other directions is relatively low, then the directional clustering is considered to be high. This value is calculated by the proportion of the number of line segments in the largest direction to the total number of line segments. Then, the three types of structural information in the region, texture coverage features, gray-level abrupt change continuity features, and contour boundary density direction features, are combined and organized into a structural factor vector group. The structural factor can be set with 6 items: texture proportion, directional concentration, abrupt change density, average abrupt change amplitude, number of boundary line segments, and boundary direction concentration. Each item is normalized to form a set of vectors. For example, a set of region vectors is [0.376, 0.094, 0.052, 28.6, 12, 0.417], which is used to finally obtain the image region quality factor group.

[0111] Please see Figure 5 The specific steps of S4 are as follows:

[0112] S401: Based on the distribution characteristics of elements in the image region quality factor group, the entire image is divided into coordinate grids, the image is divided into multiple spatial segments at fixed intervals, the quality factor value set of each segment is extracted, and the index mapping of the segment is established to obtain the spatial segment index set.

[0113] Based on the distribution characteristics of elements in the image region quality factor group, a coordinate grid structure is first established according to the overall size of the image. The image is then divided into several spatial segments with a fixed horizontal and vertical pixel spacing. For example, in an image with a size of 512×512 pixels, if the grid cell size is set to 64×64 pixels, the image can be divided into 64 spatial segments. The coordinate range of each segment in the image is determined by its starting point and size. For example, the coordinate range of the first segment is x∈[0, 63], y∈[0, 63]. Subsequently, the image region quality factor value set corresponding to each spatial segment is extracted. This set contains six items of the structure factor vector generated in the previous stage. The data includes texture proportion, orientation concentration, mutation density, average mutation amplitude, number of boundary segments, and boundary orientation concentration. Each factor value can be extracted based on the overlap between the regional quality factor and the spatial location of the fragment. If a fragment region overlaps with three structural factor regions, the factor vectors of these three regions are extracted respectively, and their average value is taken as the index expression of the fragment. For example, if the texture proportions of the three regions corresponding to a fragment are 0.32, 0.29, and 0.31 respectively, then the texture proportion of the fragment is 0.31. After processing in this way, the complete quality factor vector of the fragment is obtained. Finally, the index mapping of all spatial fragments is completed, forming a complete set of spatial fragment indexes.

[0114] S402: Based on the set of spatial segment indicators, calculate the average performance value of the quality factor per unit area in each spatial segment, call the mean of the same indicator of all segments to calculate the overall average performance standard, compare the difference between the indicator performance in a single segment and the standard, and obtain the distribution result of the segment quality deviation degree.

[0115] Based on the spatial fragment index set, the quality factor values ​​of each fragment are averaged per unit area. If the spatial fragment size is 64×64 pixels, or 4096 pixels, then each factor value is divided by 4096 to obtain the performance value per unit area. For example, if there are 12 boundary line segments, then the boundary line segment density per unit area is 0.00293. Then, the same factor items of all spatial fragments are statistically analyzed. For example, the texture ratio of all fragments is collected, and the average value of this factor across all 64 fragments is calculated as the average performance standard for the entire image. Let the average texture ratio of all fragments be 0.305. For individual fragments, if the texture ratio of a fragment is... If the value is 0.25, then the deviation value for this item is 0.305 – 0.25 = 0.055. Continue to perform the same operation on the other five factors, and count and record the deviation values ​​for each. A positive deviation value indicates that the deviation is lower than the overall average value, and a negative deviation value indicates that the deviation is higher than the overall average value. Then, perform vector processing on the six deviation values, for example, it can be recorded as a deviation vector [0.055, 0.012, 0.003, –1.6, 2, 0.031]. Each deviation value should be retained to two decimal places, and its average deviation degree or absolute deviation degree should be calculated for further analysis. Through this processing, the deviation degree distribution results of each spatial segment relative to the overall image at the quality factor level are obtained.

[0116] S403: Based on the distribution results of the degree of deviation of fragment quality, select spatial fragments in which all indicators are lower than the average standard, mark the fragments in the image coordinates and divide them into grade regions, integrate all coordinate information according to grade, and obtain a spatial structure quality grade map.

[0117] Based on the distribution of fragment quality deviation, the performance of each of the six indicators for all spatial fragments is compared with the average standard. Fragments in which all factor values ​​are lower than their corresponding standard averages are selected. The judgment process requires judging each value, such as whether the texture ratio is less than 0.305, the mutation density is less than 0.048, and the boundary orientation concentration is less than 0.3. Fragments that meet all six indicators are considered to meet the selection criteria. The image coordinate starting point and size range of the qualified fragments are recorded and the position is marked. For example, the coordinates of the fragment in the 3rd row and 2nd column are x∈ [64, 127], y∈[128, 191], then calculate the quality level based on the mean of the absolute deviation values ​​of all matching segments. If the average deviation value of a segment is between 0.02 and 0.05, it is defined as Level 1; 0.05–0.08, it is defined as Level 2; and more than 0.08, it is defined as Level 3. Then, the segments of the same level are integrated into the same group, and a quality level matrix layer with the same dimensions as the original image is constructed by assigning color labels or tag values. For example, grayscale values ​​of 80 represent Level 1, 150 represent Level 2, and 220 represent Level 3. Finally, the spatial structure quality level map is obtained.

[0118] Please see Figure 6The specific steps of S5 are as follows:

[0119] S501: Based on the spatial structure quality level map, detect the set of adjacent coordinates that are continuously distributed in the image coordinates, construct spatial path segments according to the coordinate connection order, and count the distribution of path direction angle and number of segments to obtain the distribution result of continuous path direction;

[0120] Based on the marked weak quality regions in the spatial structure quality level map, firstly, pixels with quality level values ​​greater than a set threshold are extracted from the layer. The threshold is set to 150, meaning all pixels with grayscale values ​​greater than 150 are considered weak quality regions. Then, adjacency detection is performed on these pixels using image coordinates to determine if two pixels are adjacent in their four neighboring regions (up, down, left, right). If adjacent, they are considered to be the same continuous segment. All continuous pixels are then combined into a coordinate set through a traversal connection method. For example, if a weak quality region consists of five consecutive pixels (34, 56), (34, 57), (34, 58), (35, 58), and (36, 58), they are connected according to their horizontal and vertical coordinates to form a path. Path segments are constructed according to the coordinate connection order. In the example above, four line segments are generated, connecting the two coordinate points one after the other. Then, the direction angle of each path segment is calculated. The direction angle is obtained by the coordinate difference between the starting point and the ending point. Taking the first segment (34, 56)-(34, 57) as an example, its direction angle is 90 degrees. The direction angle of the second segment (34, 57)-(34, 58) is also 90 degrees. The direction angle of the third segment (34, 58)-(35, 58) is 0 degrees. After counting the angles of all line segments, the direction distribution is divided according to the angle interval. For example, it is grouped into intervals of 15 degrees. The number of line segments in each interval is recorded. Then, the number of all path segments and the frequency of angles in the whole image are counted to finally obtain the distribution of continuous path directions.

[0121] S502: Based on the distribution results of continuous path direction, call the corresponding texture block density and boundary pixel offset distance in the area covered by the path segment, calculate the gradient value of texture density change according to the path direction, analyze the change direction and the boundary distortion offset direction, determine the consistency of the two changes in the horizontal and vertical directions, and obtain the trend characteristics of the direction consistency.

[0122] The specific formula for calculating the gradient value of texture density change is as follows:

[0123]

[0124] in, This represents the gradient value of texture density change along the path direction θ. This represents the density value of the texture block at position p along the path direction θ. This represents the density value of the texture block at position q, which is adjacent to position p. This represents the boundary pixel offset distance of the texture block at position p along the path direction θ. Ω represents the boundary pixel offset distance of the texture block at position q adjacent to position p. xp and xq represent the horizontal coordinates of the center pixel of the texture block at positions p and q, respectively. yp and yq represent the vertical coordinates of the center pixel of the texture block at positions p and q, respectively. Ω represents the set of indices of the center positions of all valid texture blocks along the path direction θ.

[0125] Explanation of Formula Calculation Logic

[0126] The numerator of the formula is the sum of the products of the texture density difference and the spatial distance (Euclidean distance), reflecting the cumulative effect of density changes in spatial distribution, while the denominator is the sum of the absolute values ​​of the differences in boundary offset between adjacent pixels, reflecting the overall magnitude of boundary distortion.

[0127] Parameter Acquisition and Quantization Process Description

[0128] The parameters must be obtained through data acquisition, image processing algorithms, and experimental quantization. The following are the methods for obtaining each parameter and examples:

[0129] Texture density The density is calculated using the local gray-level variance or local entropy of the image. Taking the gray-level variance method as an example, an 8×8 local region is selected, and the corresponding variance is used as the density index.

[0130] For example: On the path θ = 0°, the grayscale value of the texture block at point p = 1 is:

[0131] {120,123,121,119,...,125};

[0132] Variance calculation is The region corresponding to the adjacent point q=2 is calculated as follows

[0133] Boundary pixel offset distance After using an edge detection algorithm (such as Canny), the edge offset (unit: pixels) is measured and calculated based on the alignment difference between the original image and the target image.

[0134] For example: if p = 1, the position is offset by 1.3 pixels; if q = 2, the position is offset by 2.0 pixels.

[0135] Coordinates xp, yp: obtained by the position of the texture block in the image matrix. For example, the center of texture block p=1 is (x1, y1)=(12, 28), and the adjacent point q=2 is (x2, y2)=(13, 28).

[0136] Non-numerical data quantification description

[0137] If the texture density is described in a non-numerical way, such as "high", "medium", or "low", then quantization conversion is required, for example:

[0138] High density → ρ = 7.0;

[0139] Medium density → ρ = 4.0;

[0140] Low density → ρ = 2.0.

[0141] Quantization criteria can be used to classify regions based on the local entropy range (0–8 bits) of a large number of sample image areas. For example, regions with a local entropy of 6.5 or higher are considered high-density regions.

[0142] Threshold / coefficient setting instructions and examples

[0143] In path density analysis, a threshold τ can be introduced to identify regions of significant change. The saliency. The setting method is to apply the saliency to all sampling paths. Mean plus standard deviation:

[0144] For example, 10 paths are calculated as follows:

[0145] {2.1, 1.8, 3.4, 2.5, 1.9, 2.2, 2.8, 2.3, 3.0, 2.0}; with mean μ = 2.4 and standard deviation σ = 0.47, then set:

[0146] τ=μ+σ=2.87;

[0147] like These are then marked as regions with significant changes in texture density.

[0148] Parameter settings and example calculations

[0149] The following table lists example data:

[0150] Table 2 Texture Parameters for Path Direction θ = 0°

[0151]

[0152] As shown in Table 2, the calculations are as follows:

[0153] Molecular part:

[0154]

[0155] Denominator part:

[0156] |1.3-2.0|+|1.5-2.1|+|1.7-2.3|=0.7+0.6+0.6=1.9;

[0157] Substitute into the formula:

[0158]

[0159] Significance of Results and Threshold Comparison

[0160] Calculation results Compared with the threshold τ = 2.87 set earlier, the result is much lower than the threshold, indicating that the texture density change in this path direction is not significant, and it may be a relatively smooth or distortion-free area in the image.

[0161] Explanation of the innovative aspects of the formula

[0162] The advantage of this formula lies in its ability to analyze the sensitivity of texture density changes to local distortions by jointly introducing density difference and spatial distance weights, and using boundary pixel offset distance as a normalization benchmark. This approach is more accurate than traditional density change rate methods, effectively suppressing misjudgments of abnormal edges or weak texture areas, and improving the system's robustness.

[0163] S503: Based on the consistent trend characteristics, select path regions with directional continuity and quality element consistency characteristics, integrate the coordinate, direction and indicator structure characteristics of the region, establish an evaluation structure set with path chain properties, and obtain a medical image quality evaluation scheme.

[0164] Based on the consistent trend characteristics, all path segments were screened. The screening criteria were strong directional continuity and consistent quality factor performance across all path segments. Directional continuity was defined as greater than 90% of adjacent path segments having a directional angle difference of no more than 30 degrees. Quality factor consistency was defined as at least 5 quality factors in the segments covered by the path being lower than the overall map average. The path segment was set to cover 3 spatial segments with factor vectors of [0.33, 0.09, 0.04, 28.5, 10, 0.32], [0.31, 0.11, 0.05, 29.2, 11, 0], and [0.31, 0.11, 0.05, 29.2, 11, 0]. [0.30], [0.32, 0.10, 0.06, 27.9, 9, 0.31], then when the value of each factor is lower than the set standard, it is retained. All coordinate points, main directions (such as the most concentrated range of direction angles, such as 60-75 degrees) of the retained path and the average index value of its covered area are merged to establish a path chain property structure set. Each record in the structure set consists of fields such as path number, starting coordinates, ending coordinates, main direction, average texture density, average mutation intensity, and average boundary concentration. All structural information is integrated according to the path number to finally form a complete medical image quality assessment scheme.

[0165] Please see Figure 7 A medical image quality assessment system based on image processing includes:

[0166] The image texture construction module acquires the image, divides it into structural segments, extracts the center and four-way gray values, determines whether the gray value difference is stable and consistent in direction, filters it as a texture region, and generates a reliable texture layer.

[0167] The regional edge difference module subdivides the coverage area based on the reliable texture layer, collects the edge slope at multiple scales, compares the rate of change of adjacent scales, filters slope jump segments, and generates a boundary fuzzy distribution map.

[0168] The local blur determination module calls the blurred region in the boundary blur distribution map, extracts its texture coverage, gray-level abrupt density, and boundary density, calculates the three ratios, summarizes them as the region representation, and generates the image region quality factor group.

[0169] The spatial segment comparison module divides the image region into spatial segments based on the image region quality factor group, calculates the average of three performance factors for each segment, compares it with the overall average, filters out low-performance segments, and generates a spatial structure quality level map.

[0170] The continuous degradation identification module extracts the changes in texture direction and the direction of boundary undulations based on the continuous distribution areas in the spatial structure quality level map, judges the consistency of direction, filters continuous degradation paths, and generates a medical image quality assessment scheme.

[0171] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for medical image quality assessment based on image processing, characterized in that, The method comprises the following steps: S1: dividing structure segments from medical images, extracting central gray value and four-way gray value change of each segment, judging whether the change amplitude and direction are consistent, screening segments with stable gray value change and consistent direction as texture regions, and generating a reliable texture layer; S2: within the coverage area of the reliable texture layer, subdividing the image along the differentiated direction, extracting the trend of the slope change of the lower edge of the differentiated scaling layer, comparing the edge change speed of adjacent layers, marking the change jump region as a local blur range, and generating a boundary blur distribution map; S3: according to the boundary blur distribution map, extracting three elements of texture coverage, gray mutation continuity and contour density, calculating the structure proportion and distribution characteristics respectively, and generating an image region quality factor group; S4: based on the element distribution characteristics in the quality factor group, dividing the image space segments, counting the element indicators of each segment, screening the regions below the average performance, classifying them as weak structure quality regions, and generating a spatial structure quality level map; The reliable texture layer comprises spatial distribution results of texture regions, gray value change direction information and gray value stability characteristics, the boundary blur distribution map comprises edge jump position, multi-scale edge slope change and local blur degree identification, the image region quality factor group comprises texture coverage factor, gray continuity factor and contour density factor, and the spatial structure quality level map comprises spatial segment number, quality level label and low-quality region identification; The specific steps of S2 are: S201: according to the area covered by the reliable texture layer, dividing into multiple subdivision regions along the gray difference direction in the image, extracting the corresponding coordinate range and boundary direction of the subdivision region in the original image, marking the direction corresponding angle change, and obtaining the difference direction boundary angle distribution result; S202: based on the difference direction boundary angle distribution result, calling the image boundary of each subdivision region under the differentiated scaling layer, extracting the edge slope data of the same boundary position, calculating the edge slope change rate between adjacent scaling layers, recording the change trend in the region, and obtaining the edge change rate characteristics; S203: according to the edge change rate characteristics, identifying the regions with rate difference exceeding the edge jump threshold, positioning and marking the regions in the original image coordinates as fuzzy boundaries, integrating all the marking results to construct a layer, and obtaining a boundary blur distribution map layer.

2. The image processing based medical image quality assessment method according to claim 1, characterized in that, The specific steps of S1 are: S101: obtaining the image content of the medical image, dividing into structure segments, extracting the coordinates and corresponding gray values of the center pixels of the segments, recording the gray value and position relationship, and generating a center pixel gray value distribution amount; S102: based on the center pixel gray value distribution amount, extracting the four-way adjacent pixel gray values of the center pixels of the segments, calculating the direction gray value difference, and judging whether the gray value change direction is consistent to obtain the direction consistency offset degree; S103: according to the direction consistency offset degree, screening the segments with gray value difference below the gray value change threshold and consistent direction, integrating the gray value information to generate a layer, and obtaining a reliable texture layer. 3.The image processing based medical image quality assessment method according to claim 1, characterized in that, The specific steps of S3 are: S301: Based on the marked area in the boundary blur distribution layer, the texture gray scale change layer in each area is extracted, the number of texture blocks in unit area and the arrangement direction range in the layer are counted, the structure proportion and arrangement concentration are calculated combined with the area, and the texture coverage characteristic distribution result is obtained; S302: According to the texture coverage characteristic distribution result, the gray scale change sequence of continuous pixels in the area is extracted, the interval length and mutation amplitude between adjacent gray scale mutations in the sequence are collected and counted, the mutation interval distribution range and mutation continuity in the area are calculated, and the gray scale mutation continuity characteristic is obtained; S303: The gray scale mutation continuity characteristic is called to extract the number and average spacing of the corresponding area boundary line segment, the area contour boundary density distribution and boundary direction aggregation are calculated, the three types of structure performance results in the area are integrated, and are uniformly induced into a structure factor vector group, and the image area quality factor group is obtained. 4.The image processing based medical image quality assessment method according to claim 1, characterized in that, The specific steps of S4 are: S401: Based on the distribution characteristics of the elements in the image area quality factor group, the whole image is divided into a coordinate grid, the image is cut into multiple spatial segments at a fixed interval, the quality factor value set in the area covered by each segment is extracted, and the index mapping of the segment is established, and the spatial segment index set is obtained; S402: According to the spatial segment index set, the average performance value of the quality factor in unit area in each spatial segment is counted, the same index average of all segments is called to calculate the overall average performance standard, the difference between the index performance in a single segment and the standard is compared, and the segment quality deviation degree distribution result is obtained; S403: According to the segment quality deviation degree distribution result, the spatial segments with all index performances lower than the average standard are screened, the segments are marked and divided into level areas in the image coordinates, all coordinate information is integrated according to the level, and the spatial structure quality level map is obtained. 5.The image processing based medical image quality assessment method according to claim 1, characterized in that, The method further comprises: S5: According to the weak area in the quality level map, a continuous trend path on the image is extracted, the consistency of the texture sparseness trend and the boundary distortion direction change of adjacent areas is combined, whether there is a continuous quality decline chain extending in the horizontal and vertical directions is judged, and a medical image quality evaluation scheme is generated. 6.The image processing based medical image quality assessment method according to claim 1, characterized in that, The medical image quality evaluation scheme comprises a continuous trend path, a direction consistency trend, and a weakening chain coverage area. 7.The image processing based medical image quality assessment method according to claim 1, characterized in that, The specific steps of S5 are: S501: According to the spatial structure quality level map, a set of adjacent coordinates continuously distributed in the image coordinates is detected, a spatial path line segment is constructed according to the coordinate connection order, and the path direction angle and line segment number distribution are counted, and the continuous path trend distribution result is obtained; S502: Based on the continuous path trend distribution result, the corresponding texture block density and boundary pixel offset distance in the path line segment coverage area are called, the texture density change gradient value is calculated according to the path direction, the change direction and the boundary distortion offset direction are analyzed, the consistency of the two in the horizontal and vertical directions is judged, and the trend consistency trend characteristic is obtained; S503: According to the trend of the consistency of the direction, the path area with the characteristics of the direction continuity and the quality element consistency is screened, the coordinates, the direction and the index structure characteristics of the area are integrated, the evaluation structure set of the path chain property is established, and the medical image quality evaluation scheme is obtained.

8. A medical image quality assessment system based on image processing, characterized in that, The image processing-based medical image quality evaluation method according to any one of claims 1-7, wherein the system comprises: The image texture construction module acquires the image, divides it into structure segments, extracts the center and four-way gray values, judges whether the gray difference values are stable and consistent in direction, screens the texture area, and generates a reliable texture layer; The area edge difference module subdivides the coverage area according to the reliable texture layer, collects the edge slope under multiple scales, compares the change speed of adjacent scales, screens the slope jump segments, and generates a boundary blur distribution map; The local blur judgment module calls the blur area in the boundary blur distribution map, extracts the texture coverage, the gray mutation density and the boundary density, calculates the three ratios, summarizes the area performance, and generates an image area quality factor group; The space segment comparison module divides the space segment according to the image area quality factor group, calculates the average value of the three performances of each segment, compares it with the overall average, screens the low-performance segments, and generates a space structure quality level map; The continuous deterioration recognition module extracts the texture direction change and the boundary fluctuation trend according to the continuous distribution area in the space structure quality level map, judges the direction consistency, screens the continuous deterioration path, and generates a medical image quality evaluation scheme.

Images