An image processing method, apparatus, and storage medium

By acquiring endoscopic image sequences and using a classification model to identify fixed-pattern noise, combined with sharpness evaluation data, the endoscopic imaging level is automatically determined, solving the problem of endoscopic image quality evaluation and improving the efficiency and accuracy of level adjustment.

CN123071807APending Publication Date: 2026-11-10SZ HUGEMED MED TECH DEV CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611250251.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-18
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

In existing technologies, it is difficult to balance the suppression of fixed pattern noise and image quality in endoscopic image quality evaluation, resulting in low efficiency and accuracy of imaging level adjustment, and reliance on manual observation and experience judgment.

Method used

By acquiring image sequences from an endoscope, a classification model is used to identify fixed-pattern noise information, and combined with sharpness evaluation data, the target imaging level is determined, enabling automated and precise level adjustment.

Benefits of technology

It enables accurate and rapid classification of endoscopic images, provides comprehensive data support, improves the efficiency and accuracy of imaging level determination, and solves the problems of low efficiency and low accuracy in existing technologies.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123071807A_ABST
    Figure CN123071807A_ABST
Patent Text Reader

Abstract

This invention discloses an image processing method, apparatus, and storage medium. The method includes: acquiring an image sequence captured by a target endoscope, the image sequence including endoscope images at different imaging settings; if the image sequence is valid, acquiring image quality information corresponding to each endoscope image, and determining sharpness evaluation data for the endoscope images based on the image quality information; classifying each endoscope image using a classification model to obtain classification results for each endoscope image, the classification results including column fixed-pattern noise information in the endoscope images; determining column fixed-pattern noise evaluation data corresponding to the endoscope images based on the classification results; determining quality evaluation data for the endoscope images based on the sharpness evaluation data and the column fixed-pattern noise evaluation data; and determining the target imaging setting among different imaging settings based on the quality evaluation data, thus achieving accurate determination of the endoscope imaging setting.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of medical image processing technology, and in particular to an image processing method, apparatus and storage medium. Background Technology

[0002] Imaging settings are imaging control parameters in an endoscopic imaging system; different imaging settings correspond to different imaging control parameters. Vertical stripe artifacts in endoscopic images are essentially column-fixed pattern noise, and imaging settings have a significant modulating effect on this noise, thus affecting the image quality of the endoscopic image.

[0003] In existing technologies, the image quality of endoscopic images is typically analyzed to obtain quality evaluation data. Based on this data, the imaging setting of the endoscope is adjusted to ensure that the acquired images meet clinical observation requirements. Image quality analysis of endoscopic images is usually based on sharpness indicators such as image clarity, edge strength, or Laplacian variance, or on noise evaluation indicators in the frequency and spatial domains. However, determining image quality solely based on sharpness indicators can lead to the misidentification of false edges caused by fixed-pattern noise as true image details. Determining image quality based on frequency and spatial domain noise evaluation indicators can easily overlook the image quality of the central and focused areas, resulting in the selection of endoscopic images with weak fixed-pattern noise but excessive smoothing or high noise levels. The quality evaluation data determined by existing technologies struggles to simultaneously address both the suppression of fixed-pattern noise and overall image quality, making it difficult to objectively and comprehensively characterize the overall image quality of endoscopic images and thus reliably select the imaging setting with the optimal overall performance. Furthermore, the adjustment of the imaging settings of existing endoscopes usually relies on the manual observation and experience of technicians. Different technicians have different experiences, and the imaging settings determined by different technicians are inconsistent. In addition, it is time-consuming and has the problems of low efficiency and low accuracy in adjusting the imaging settings. Summary of the Invention

[0004] This invention provides an image processing method, apparatus, and storage medium to achieve accurate determination of endoscopic imaging settings.

[0005] According to one aspect of the present invention, an image processing method is provided, comprising: Acquire the image sequence captured by the target endoscope, including endoscope images at different imaging settings; If the image sequence is a valid sequence, obtain the image quality information corresponding to each endoscope image, and determine the clarity evaluation data of the endoscope image based on the image quality information; Each endoscope image is classified based on a classification model to obtain the classification result corresponding to each endoscope image. The classification result includes column fixed pattern noise information in the endoscope image. Based on the classification results, determine the column-fixed pattern noise evaluation data corresponding to the endoscopic images; Quality assessment data for endoscopic images are determined based on sharpness assessment data and column fixed pattern noise assessment data. The target imaging level is determined based on the quality assessment data.

[0006] According to another aspect of the present invention, an image processing apparatus is provided, comprising: The image sequence acquisition module is used to acquire the image sequence captured by the target endoscope. The image sequence includes endoscope images at different imaging settings. The sharpness evaluation data determination module is used to obtain the image quality information corresponding to each endoscope image when the image sequence is a valid sequence, and to determine the sharpness evaluation data of the endoscope image based on the image quality information. The classification result determination module is used to classify each endoscope image based on the classification model to obtain the classification result corresponding to each endoscope image. The classification result includes column fixed pattern noise information in the endoscope image. The column fixed pattern noise evaluation data determination module is used to determine the column fixed pattern noise evaluation data corresponding to the endoscopic image based on the classification results. The quality assessment data determination module is used to determine the quality assessment data of endoscopic images based on sharpness assessment data and column fixed pattern noise assessment data. The target imaging level determination module is used to determine the target imaging level among different imaging levels based on quality evaluation data.

[0007] According to another aspect of the present invention, a computer-readable storage medium is provided, which stores computer instructions for causing a processor to execute and implement the image processing method of any embodiment of the present invention.

[0008] The technical solution of this invention acquires an image sequence from a target endoscope, including endoscope images at different imaging settings, providing comprehensive data support for subsequent analysis and processing. When the image sequence is valid, image quality information corresponding to each endoscope image is obtained. Based on this image quality information, the sharpness evaluation data of the endoscope images is determined, achieving accurate determination of the endoscope image sharpness evaluation data and providing comprehensive data support for subsequent analysis and processing. Each endoscope image is classified using a classification model, yielding classification results for each endoscope image. These classification results include column-fixed pattern noise information in the endoscope images, achieving accurate and rapid classification of endoscope images and providing accurate data support for subsequent analysis and processing. Based on the classification results, the column-fixed pattern noise evaluation data corresponding to the endoscope images is determined. The system achieves accurate determination of fixed-mode noise evaluation data, providing accurate data support for subsequent analysis and processing. Based on both sharpness evaluation data and fixed-mode noise evaluation data, it determines the quality evaluation data for endoscopic images, moving beyond solely relying on sharpness indicators to judge image quality. This approach comprehensively considers both the detail rendering capability of endoscopic images and quantifies image artifact interference caused by fixed-mode noise, balancing the two core dimensions of image sharpness and noise defects. This provides a reliable comprehensive evaluation basis for subsequent analysis and processing. Furthermore, it determines the target imaging level within different imaging levels based on the quality evaluation data. This solves the problems of existing technologies struggling to balance fixed-mode noise suppression and image quality, as well as the low efficiency and accuracy of manual determination of imaging levels. It achieves accurate determination of the target imaging level, improving the efficiency of target imaging level determination.

[0009] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0010] 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.

[0011] Figure 1 This is a flowchart of an image processing method provided in an embodiment of the present invention; Figure 2 This is a flowchart of another image processing method provided in an embodiment of the present invention; Figure 3This is a flowchart of another image processing method provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of an image processing device provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0012] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0013] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0014] Figure 1 This is a flowchart of an image processing method provided in an embodiment of the present invention. This embodiment is applicable to situations where the imaging level of an endoscope is determined based on an image sequence acquired by the endoscope. This method can be executed by an image processing device, which can be implemented in hardware and / or software. The image processing device can be configured in the electronic device provided in this embodiment of the present invention. The electronic device can be a server, computer, or mobile terminal, such as a mobile phone or tablet computer. Figure 1 As shown, the method specifically includes the following steps: S110. Acquire the image sequence captured by the target endoscope, including endoscope images at different imaging settings.

[0015] The target endoscope is the endoscope whose imaging settings are to be adjusted. The image sequence is a set of ordered endoscope images. Optionally, the image sequence includes endoscope images at different imaging settings. The imaging setting refers to the imaging control parameter of the target endoscope. Different imaging control parameters correspond to different imaging settings. It should be noted that all endoscope images have the same pixel scale and resolution; for example, the pixel scale can be 8 bits. Optionally, the imaging settings can include eight imaging settings from 0 to 7. The image sequence can be obtained from an image sequence database. For example, by matching the unique identifier information of the target endoscope in the image sequence database, the image sequence corresponding to the target endoscope can be obtained. The image sequence database stores image sequences corresponding to different endoscopes.

[0016] Specifically, the unique identifier of the target endoscope is matched in the image sequence database to obtain the image sequence corresponding to the target endoscope, providing comprehensive data support for subsequent analysis and processing.

[0017] Based on the above embodiments, after acquiring the image sequence captured by the target endoscope, the method further includes: performing preprocessing operations on each endoscope image in the image sequence; wherein the preprocessing operations include one or more of the following: grayscale conversion, region of interest cropping, and image enhancement processing.

[0018] S120. If the image sequence is a valid sequence, obtain the image quality information corresponding to each endoscope image, and determine the clarity evaluation data of the endoscope image based on the image quality information.

[0019] Image quality information refers to objective attributes characterizing the visual quality of endoscopic images across multiple dimensions. Optionally, image quality information includes at least one of sharpness information, local contrast information, glare information, and high-frequency noise information. Sharpness information is a quantitative feature characterizing the sharpness of an endoscopic image. Local contrast information is a quantitative data representation of the difference in brightness and grayscale within a local region of the endoscopic image. Optionally, local contrast information can be the local contrast information of the region of interest in the endoscopic image. Glare information is a quantitative indicator of overexposed areas in the endoscopic image. High-frequency noise information is a quantitative indicator of the degree of random particle noise in the endoscopic image. Edge density information is a quantitative indicator of the richness of effective contour edges in the endoscopic image. The image quality information corresponding to the endoscopic image can be obtained through an image quality information determination model. For example, the endoscopic image is input into a trained image quality information determination model for processing to obtain the image quality information corresponding to the endoscopic image. The image quality information determination model includes, but is not limited to, a neural network model. The image quality information determination model is selected according to requirements, and this invention is not limited thereto. Sharpness evaluation data is a quantitative score characterizing the imaging sharpness of the endoscopic image. Sharpness evaluation data can be determined based on image quality information. For example, image quality information can be input into a trained sharpness evaluation data determination model for processing to obtain sharpness evaluation data for endoscopic images. The sharpness evaluation data determination model includes, but is not limited to, neural network models. The sharpness evaluation data determination model can be selected according to requirements, and this invention does not impose any limitations.

[0020] Specifically, when the image sequence is valid, the endoscope image is input into the trained image quality information determination model for processing to obtain the image quality information corresponding to the endoscope image; the image quality information is then input into the trained sharpness evaluation data determination model for processing to obtain the sharpness evaluation data of the endoscope image. This achieves accurate determination of the sharpness evaluation data of the endoscope image, providing comprehensive data support for subsequent analysis and processing.

[0021] Based on the above embodiments, after obtaining the image quality information corresponding to each endoscopic image, the method further includes: classifying the sharpness information, local contrast information, glare information, and high-frequency noise information to obtain the data types corresponding to the sharpness information, local contrast information, glare information, and high-frequency noise information, respectively. The data types include positive type and risk type; determining the corresponding normalization rule according to the data type; and normalizing the sharpness information, local contrast information, glare information, and high-frequency noise information according to the corresponding normalization rule to obtain normalized sharpness information, normalized local contrast information, normalized glare information, and normalized high-frequency noise information.

[0022] The image quality information corresponding to the positive type includes sharpness and local contrast information, while the image quality information corresponding to the risk type includes glare and high-frequency noise information. Different data types correspond to different normalization rules.

[0023] For example, the normalization rules for positive types are as follows: ; in, This represents the normalized, positively typed image quality information; This represents the image quality information of the positive type before normalization; This represents the minimum value among multiple positive types of image quality information; The span representing multiple positive types of image quality information can be determined based on the difference between the maximum and minimum values ​​of the positive types of image quality information.

[0024] For example, the normalization rules corresponding to risk types are as follows: ; in, Image quality information representing the normalized risk type; Image quality information representing the risk type before normalization; This represents the minimum value among multiple risk types of image quality information. The span of image quality information representing multiple risk types can be determined based on the difference between the maximum and minimum values ​​of image quality information for each risk type.

[0025] Optionally, the sharpness evaluation data of the endoscopic image is determined based on the image quality information, including: weighting the sharpness information, local contrast information, glare information and high-frequency noise information to obtain the sharpness evaluation data.

[0026] The sharpness evaluation data can be determined based on sharpness information, local contrast information, glare information, high-frequency noise information, and edge density information. For example, the weighted sum of sharpness information, local contrast information, glare information, high-frequency noise information, and edge density information is calculated as the sharpness evaluation data, which enables the accurate determination of the sharpness evaluation data of endoscopic images and provides comprehensive data support for subsequent analysis and processing.

[0027] For example, the formula for calculating sharpness evaluation data is as follows: ; in, This represents the data for clarity evaluation; Indicates sharpness information; The weighting coefficient corresponding to the sharpness information can be set to 0.44; Represents local contrast information; The weighting coefficient corresponding to the local contrast information can be 0.16; Indicates glare information; The weighting coefficient corresponding to the glare information can be set to 0.12; Indicates high-frequency noise information; The weighting coefficient corresponding to the high-frequency noise information can be 0.28.

[0028] Optionally, the method further includes: acquiring edge naturalness information, oversharpness penalty information, and noise excess penalty information corresponding to the endoscopic image; and correcting the sharpness evaluation data corresponding to the endoscopic image based on the edge naturalness information, oversharpness penalty information, and noise excess penalty information to obtain corrected sharpness evaluation data.

[0029] Among them, edge naturalness information is feature data characterizing the smoothness of edge transitions in endoscopic images. Oversharpness penalty information is a penalty coefficient characterizing the severity of edge distortion in endoscopic images. Excess noise penalty information is a penalty coefficient characterizing the degree to which random particle noise in endoscopic images exceeds an acceptable threshold. Edge naturalness information, oversharpness penalty information, and excessive noise penalty information can be used to correct sharpness evaluation data. For example, sharpness evaluation data can be input into a trained sharpness evaluation data correction model for processing to obtain corrected sharpness evaluation data. Sharpness evaluation data correction models include, but are not limited to, neural network models and mathematical models. The choice of sharpness evaluation data correction model is based on requirements, and this invention is not restrictive.

[0030] Specifically, by acquiring edge naturalness information, oversharpness penalty information, and noise excess penalty information corresponding to the endoscopic image, the sharpness evaluation data is input into the trained sharpness evaluation data correction model for processing, resulting in corrected sharpness evaluation data. This process corrects the sharpness evaluation data, making it more closely match the endoscopic image and providing accurate data support for subsequent analysis and processing.

[0031] For example, the formula for calculating the corrected sharpness evaluation data is as follows: ; in, This indicates the corrected sharpness evaluation data; Represents the amplitude limiting function; This represents the data for clarity evaluation; Represents edge naturalness information; The weighting coefficient representing the edge naturalness information can be set to 0.3; This represents the basic safety margin, which can be set to 0.7. This indicates an overly sharp penalty message; The weighting coefficient for overly sharp penalty information can be set to 0.1. This indicates a penalty message for excessive noise. The weighting coefficient representing the excessive noise penalty information can be set to 0.08.

[0032] Optionally, the image quality information includes Tenengrad gradient sharpness features, Brenner gray-level difference contrast features, Laplacian variance sharpness features, gradient magnitude percentile information, and sharpness information. The process for determining the sharpness information is as follows: the Tenengrad gradient sharpness features, Brenner gray-level difference contrast features, Laplacian variance sharpness features, and gradient magnitude percentile information are weighted to obtain the sharpness information.

[0033] Among them, the Tenengrad gradient sharpness feature represents the quantitative characteristic of the sharpness of image details in the endoscopic image. The Brenner gray-level difference contrast feature represents the quantitative characteristic of the gray-level difference in the region of interest in the endoscopic image. The Laplacian variance sharpness feature represents the quantitative characteristic of the sharpness within the focus of the endoscopic image. The gradient magnitude percentile information represents the level of effective high-intensity edges in the endoscopic image. Optionally, the gradient magnitude percentile information may include the 95th percentile of the gradient magnitude.

[0034] Sharpness information can be determined based on Tenengrad gradient sharpness features, Brenner gray-level difference contrast features, Laplacian variance sharpness features, and gradient magnitude percentile information. For example, calculating the weighted sum of Tenengrad gradient sharpness features, Brenner gray-level difference contrast features, Laplacian variance sharpness features, and gradient magnitude percentile information provides accurate sharpness information, enabling precise determination of sharpness information and providing accurate data support for subsequent analysis and processing.

[0035] For example, the formula for calculating sharpness information is as follows: ; in, Indicates sharpness information; This represents the clarity feature of the Tenengrad gradient; This represents the weight coefficient corresponding to the Tenengrad gradient clarity feature, which can be set to 0.4; This represents the Brenner grayscale difference contrast feature; This represents the weight coefficient corresponding to the Tenengrad gradient clarity feature, which can be set to 0.25. This represents the clarity characteristic of the Laplace variance; The weighting coefficient corresponding to the Laplace variance clarity feature can be set to 0.2. This indicates the 95th percentile information of the gradient magnitude; The weighting coefficient corresponding to the Laplace variance clarity feature can be set to 0.15.

[0036] Optionally, the image quality information includes first gray-level percentile information, second gray-level percentile information, and local contrast information; the process for determining the local contrast information is as follows: the local contrast information is determined based on the first gray-level percentile information and the second gray-level percentile information.

[0037] The first gray-level percentile information is the low-percentile gray-level value of the pixel gray-level distribution in the endoscopic image, representing the gray-level baseline of the dark area of ​​the endoscopic image. Optionally, the first gray-level percentile information includes the 10th gray-level percentile information. The second gray-level percentile information is the high-percentile gray-level value of the pixel gray-level distribution in the endoscopic image, representing the gray-level baseline of the bright area of ​​the endoscopic image. Optionally, the second gray-level percentile information includes the 90th gray-level percentile information. Local contrast information can be determined based on the first gray-level percentile information and the second gray-level percentile information. For example, the first gray-level percentile information and the second gray-level percentile information are input into a trained local contrast information determination model for processing to obtain local contrast information. The local contrast information determination model includes, but is not limited to, neural network models and mathematical models. The local contrast information determination model is selected according to the requirements, and this invention is not limited thereto.

[0038] Specifically, by inputting the first and second gray-level percentile information into the trained local contrast information determination model for processing, local contrast information is obtained, achieving accurate determination of local contrast information and providing accurate data support for subsequent analysis and processing.

[0039] For example, the formula for calculating local contrast information is as follows: ; in, Represents local contrast information; This represents an endoscope image that has undergone image enhancement processing; This represents the second grayscale percentile information, which can be the 90th grayscale percentile information; This represents the first grayscale percentile information, which can be the 10th grayscale percentile information.

[0040] Optionally, the image quality information includes the proportion of overexposed pixels, the third grayscale percentile information, and glare information. The process for determining the glare information is as follows: the proportion of overexposed pixels is weighted to obtain the first glare evaluation data; the difference between the third grayscale percentile information and the preset bright pixel threshold is calculated, and the second glare evaluation data is determined based on the difference between the three grayscale percentile information and the preset bright pixel threshold; the first glare evaluation data and the second glare evaluation data are summed to obtain the glare information.

[0041] The overexposed pixel ratio refers to the proportion of brightly saturated regions in the endoscopic image. Optionally, the overexposed pixel ratio can be the proportion of pixels with grayscale information greater than a preset grayscale threshold. The preset grayscale threshold can be set to 245 or 240, depending on the requirements; this invention is not limited by this. Optionally, the overexposed pixel ratio includes the proportion of pixels in the region of interest (ROI) with grayscale information greater than the preset grayscale threshold to the total number of pixels in the ROI. The third grayscale percentile information is the grayscale value corresponding to the high percentile of the pixel grayscale distribution in the endoscopic image, representing the grayscale level of the bright areas in the endoscopic image. Optionally, the third grayscale percentile information includes the 99.5th grayscale percentile information.

[0042] The first glare evaluation data is quantitative data characterizing the severity of glare in the dimension of the proportion of bright saturated areas in the endoscopic image. Optionally, the first glare evaluation data may include a weighted sum of the proportion of pixels with grayscale information greater than a preset grayscale threshold in the region of interest of the endoscopic image. The first glare evaluation data can be determined based on the proportion of overexposed pixels. For example, the product of the weight coefficient corresponding to the proportion of overexposed pixels and the proportion of overexposed pixels can be calculated as the first glare evaluation data. The second glare evaluation data is quantitative data characterizing the severity of glare in the dimension of grayscale saturation depth of the bright areas in the endoscopic image.

[0043] The second glare evaluation data can be determined based on the third gray-level percentile information and a preset bright pixel threshold. For example, the difference between the third gray-level percentile information and the preset bright pixel threshold is calculated, and this difference is input into a trained second glare evaluation data determination model for processing to obtain the second glare evaluation data. The second glare evaluation data determination model includes, but is not limited to, neural network models and mathematical models. The second glare evaluation data determination model is selected according to requirements, and this invention is not limited thereto. Glare information can be determined based on the first glare evaluation data and the second glare evaluation data. For example, the arithmetic sum of the first glare evaluation data and the second glare evaluation data is calculated as the glare information.

[0044] Specifically, the product of the weighting coefficient corresponding to the overexposed pixel ratio and the overexposed pixel ratio is calculated as the first glare evaluation data; the difference between the third gray-level percentile information and the preset bright pixel threshold is calculated, and the difference between the third gray-level percentile information and the preset bright pixel threshold is input into the trained second glare evaluation data determination model for processing to obtain the second glare evaluation data; the arithmetic sum of the first glare evaluation data and the second glare evaluation data is calculated as the glare information, realizing the accurate determination of glare information and providing accurate data support for subsequent analysis and processing.

[0045] For example, the formula for calculating the proportion of overexposed pixels is as follows: ; in, Indicates the percentage of overexposed pixels; 245 represents the region of interest in the endoscopic image; 245 represents the preset grayscale threshold. This indicates the number of pixels with grayscale information greater than 245 in the region of interest of an endoscopic image; This represents the total number of pixels in the region of interest of an endoscopic image.

[0046] For example, the formula for calculating the second glare evaluation data is as follows: ; in, This indicates the second glare evaluation data; Indicates the percentage of overexposed pixels; The weighting coefficient corresponding to the proportion of overexposed pixels can be set to 100. This represents the third gray-level percentile information in the region of interest of the endoscopic image, which can be the 99.5th gray-level percentile. This represents the first glare evaluation data.

[0047] Optionally, the image quality information includes grayscale enhancement information, grayscale smoothing information, and high-frequency noise information at each pixel location. The process for determining the high-frequency noise information is as follows: determine the grayscale residual information corresponding to each pixel location based on the grayscale enhancement information and the corresponding grayscale smoothing information at each pixel location; determine the standard deviation corresponding to the grayscale residual information based on the grayscale residual information at each pixel location; and determine the high-frequency noise information based on the standard deviation.

[0048] The grayscale enhancement information refers to the pixel grayscale values ​​at each pixel location obtained after the endoscope image has undergone contrast equalization processing. Optionally, the grayscale enhancement information includes the pixel grayscale values ​​at each pixel location obtained after the endoscope image has undergone CLAHE contrast-limited adaptive histogram equalization processing. The grayscale smoothing information refers to the pixel grayscale values ​​at each pixel location obtained after smoothing the endoscope image that has undergone contrast equalization processing. Optionally, the grayscale smoothing information includes the pixel grayscale values ​​at each pixel location obtained after Gaussian smoothing the endoscope image that has undergone contrast equalization processing.

[0049] Gray-level residual information represents the fluctuation in gray-level information at a pixel location after removing low-frequency main components from an endoscopic image. Gray-level residual information can be determined based on gray-level enhancement information and the corresponding gray-level smoothing information at the pixel location. For example, the difference between the gray-level enhancement information and the corresponding gray-level smoothing information at the pixel location can be calculated as the gray-level residual information for that pixel location. High-frequency noise information can also be determined based on the standard deviation corresponding to the gray-level residual information. For example, the standard deviation corresponding to the gray-level residual information can be used as the high-frequency noise information.

[0050] Specifically, the difference between the gray-level enhancement information and the gray-level smoothing information at each pixel location is calculated to obtain the gray-level residual information corresponding to each pixel location. The standard deviation is calculated based on the gray-level residual information corresponding to each pixel location, and the standard deviation is used as high-frequency noise information, which realizes the accurate determination of high-frequency noise information and provides accurate data support for subsequent analysis and processing.

[0051] For example, the formula for calculating high-frequency noise information is as follows: ; in, Indicates high-frequency noise information; This represents an endoscope image that has undergone image enhancement processing; Indicates Gaussian smoothing; Indicates the Gaussian kernel standard deviation; The standard deviation of the grayscale residual information corresponding to all pixel locations; This represents grayscale residual information.

[0052] S130. Based on the classification model, classify each endoscope image to obtain the classification result corresponding to each endoscope image. The classification result includes column fixed pattern noise information in the endoscope image.

[0053] The classification model is a structure for classifying endoscopic images. The classification model includes, but is not limited to, deep learning models. The classification result characterizes whether the endoscopic image contains fixed-pattern noise and the corresponding quantization information. Optionally, the classification result includes the distribution information of fixed-pattern noise in the vertical pixel columns of the endoscopic image. Optionally, the fixed-pattern noise information includes the encoding information corresponding to each vertical pixel column in the endoscopic image, and the encoding information characterizes whether fixed-pattern noise exists in the vertical pixel columns. Fixed-pattern noise can be understood as vertical stripes in the endoscopic image. Fixed-pattern noise is controlled by the imaging settings of the target endoscope and forms vertically distributed artifacts in the endoscopic image. Vertical stripes in the endoscopic image can interfere with the observation of mucosal texture, blood vessel direction, lesion boundaries, and low-contrast tissue structures in clinical settings.

[0054] It should be noted that the classification results can be represented by different identifiers, including but not limited to numbers, letters, and text. For example, using the numbers 0 and 1 as binary identifiers for pixel columns, when the classification result is 1, the corresponding vertical pixel column in the endoscopic image contains column fixed pattern noise information; when the classification result is 0, the corresponding vertical pixel column in the endoscopic image does not contain column fixed pattern noise information.

[0055] Specifically, the endoscopic images are input into a classification model for classification processing to obtain the corresponding classification results, thus achieving accurate and rapid classification of endoscopic images and providing accurate data support for subsequent analysis and processing.

[0056] Optionally, the classification model includes a global spatial feature extraction module, a height-direction spatial feature extraction module, a pooling module, and a column-level classification module. The global spatial feature extraction module extracts global spatial features from the endoscopic image to obtain global spatial features. The height-direction spatial feature extraction module extracts features from the global spatial features, downsampling along the height direction corresponding to the global spatial features while maintaining the resolution in the width direction to obtain height-direction spatial features. The pooling module performs pooling processing on the height-direction spatial features to obtain target features. The column-level classification module performs classification processing on the target features to obtain the classification result.

[0057] The classification model includes a global spatial feature extraction module, a height-direction spatial feature extraction module, a pooling module, and a column-level classification module. The global spatial feature extraction module is a structure for extracting global spatial features from endoscopic images. Optionally, the global spatial feature extraction module includes a two-dimensional convolutional layer. The height-direction spatial feature extraction module is a structure for extracting vertical spatial features related to column-fixed pattern noise along the height dimension of the global spatial features. Optionally, the height-direction spatial feature extraction module includes four levels of two-dimensional convolutional units. Each two-dimensional convolutional unit performs downsampling along the height direction while maintaining the width resolution, progressively extracting vertical features to obtain height-direction spatial features at different scales. The pooling module is a structure for pooling the height-direction spatial features. Optionally, the pooling module includes an average pooling unit. The column-level classification module is a structure for classifying the target features. Optionally, the column-level classification module includes a column-level classifier.

[0058] Optionally, the process for determining the column fixed pattern noise risk index is as follows: weighting the column fixed pattern noise ratio, the maximum continuous column fixed pattern noise width, the number of column fixed pattern noise clusters, and the odd / even column bias to obtain the column fixed pattern noise risk index.

[0059] The column-fixed pattern noise risk index can also be determined based on the column-fixed pattern noise ratio, the maximum consecutive column-fixed pattern noise width, the number of column-fixed pattern noise clusters, and the odd / even column bias. For example, the weighted sum of the column-fixed pattern noise ratio, the maximum consecutive column-fixed pattern noise width, the number of column-fixed pattern noise clusters, and the odd / even column bias can be used as the column-fixed pattern noise risk index.

[0060] For example, the formula for calculating the fixed-pattern noise risk index is as follows: ; in, This indicates a fixed-pattern noise risk index. Indicates the noise ratio of the column in fixed mode; The weighting coefficient corresponding to the proportion of noise in the fixed-pattern column can be 0.6; Indicates the maximum consecutive column fixed-pattern noise width; The weighting coefficient corresponding to the maximum continuous column fixed pattern noise width can be 0.25; Indicates the number of noise clusters in the fixed-pattern column; The weighting coefficient corresponding to the number of noise clusters in the fixed column pattern can be 0.1; Indicates parity column offset; This represents the weighting coefficient corresponding to the odd / even column bias, which can be 0.05.

[0061] S140. Based on the classification results, determine the column fixed pattern noise evaluation data corresponding to the endoscopic image.

[0062] The column fixed-pattern noise evaluation data is used to characterize the severity of column fixed-pattern noise in endoscopic images. The column fixed-pattern noise evaluation data can be determined based on classification results. For example, the classification results are input into a trained column fixed-pattern noise evaluation data determination model for processing to obtain the column fixed-pattern noise evaluation data. The column fixed-pattern noise evaluation data determination model includes, but is not limited to, neural network models and mathematical models. The selection of the column fixed-pattern noise evaluation data determination model is based on requirements, and this invention is not restrictive.

[0063] Specifically, by inputting the classification results into the trained column-fixed pattern noise evaluation data determination model, column-fixed pattern noise evaluation data is obtained, thus achieving accurate determination of column-fixed pattern noise evaluation data and providing accurate data support for subsequent analysis and processing.

[0064] Optionally, the column fixed pattern noise evaluation data of the endoscopic image is determined based on the classification results, including: determining the column fixed pattern noise ratio, the maximum continuous column fixed pattern noise width, the number of column fixed pattern noise clusters, the odd-even column bias, and the column fixed pattern noise risk index corresponding to the endoscopic image based on the coding information; and performing weighted processing on the column fixed pattern noise ratio, the maximum continuous column fixed pattern noise width, the number of column fixed pattern noise clusters, the odd-even column bias, and the column fixed pattern noise risk index to obtain the column fixed pattern noise evaluation data.

[0065] The column fixed-pattern noise ratio is the ratio of the total number of vertical pixel columns containing column fixed-pattern noise to the total number of vertical pixel columns in the image. The maximum consecutive column fixed-pattern noise width is the number of the longest consecutive vertical pixel columns containing column fixed-pattern noise in the image. The number of column fixed-pattern noise clusters is the total number of mutually separated and independent column fixed-pattern noise stripe clusters. The odd-even column bias is the difference between the distribution ratio of column fixed-pattern noise in odd-numbered vertical pixel columns and the distribution ratio of column fixed-pattern noise in even-numbered vertical pixel columns. The column fixed-pattern noise risk index characterizes the intensity of column fixed-pattern noise. The column fixed-pattern noise ratio, the maximum consecutive column fixed-pattern noise width, the number of column fixed-pattern noise clusters, the odd-even column bias, and the column fixed-pattern noise risk index can all be determined based on the encoded information. For example, the encoded information can be input into a trained statistical feature determination model for processing to obtain the column fixed-pattern noise ratio, the maximum consecutive column fixed-pattern noise width, the number of column fixed-pattern noise clusters, the odd-even column bias, and the column fixed-pattern noise risk index. The statistical feature determination model includes, but is not limited to, a neural network model. The statistical feature determination model can be selected according to the requirements, and this invention is not limited thereto.

[0066] Column-fixed pattern noise evaluation data can also be determined based on the proportion of column-fixed pattern noise, the maximum width of consecutive column-fixed pattern noise, the number of column-fixed pattern noise clusters, the odd / even column bias, and the column-fixed pattern noise risk index. For example, the weighted sum of the proportion of column-fixed pattern noise, the maximum width of consecutive column-fixed pattern noise, the number of column-fixed pattern noise clusters, the odd / even column bias, and the column-fixed pattern noise risk index can be used as the column-fixed pattern noise evaluation data.

[0067] Specifically, the encoded information is input into the trained statistical feature determination model for processing to obtain the column fixed pattern noise ratio, the maximum continuous column fixed pattern noise width, the number of column fixed pattern noise clusters, the odd / even column bias, and the column fixed pattern noise risk index. The weighted sum of the column fixed pattern noise ratio, the maximum continuous column fixed pattern noise width, the number of column fixed pattern noise clusters, the odd / even column bias, and the column fixed pattern noise risk index is calculated as the column fixed pattern noise evaluation data. This achieves accurate determination of the column fixed pattern noise evaluation data and provides accurate data support for subsequent analysis and processing.

[0068] For example, the calculation formula for fixed-pattern noise evaluation data is as follows: ; in, This indicates the fixed-pattern noise evaluation data. Indicates the noise ratio of the column in fixed mode; The weighting coefficient corresponding to the proportion of noise in the fixed pattern column can be 0.45; Indicates the maximum consecutive column fixed-pattern noise width; The weighting coefficient corresponding to the maximum continuous column fixed pattern noise width can be 0.2; Indicates the number of noise clusters in the fixed-pattern column; The weighting coefficient corresponding to the number of noise clusters in the fixed column pattern can be 0.1; Indicates parity column offset; This represents the weighting coefficient corresponding to the odd / even column bias, which can be 0.1. This indicates a fixed-pattern noise risk index. The weighting coefficient corresponding to the fixed-pattern noise risk index can be 0.15.

[0069] S150. Determine the quality assessment data of the endoscopic image based on the sharpness assessment data and the column fixed pattern noise assessment data.

[0070] The quality assessment data refers to quantitative data used to characterize the overall imaging quality of endoscopic images. The quality assessment data can be determined based on sharpness assessment data and column-fixed pattern noise assessment data. For example, the sharpness assessment data and column-fixed pattern noise assessment data are input into a trained quality assessment data determination model for processing to obtain the quality assessment data. The quality assessment data determination model includes, but is not limited to, neural network models and mathematical models. The model is selected according to requirements, and this invention does not impose any limitations.

[0071] Specifically, the sharpness evaluation data and column fixed pattern noise evaluation data are input into the trained quality evaluation data determination model for processing to obtain quality evaluation data. Instead of relying solely on the sharpness index to judge the quality of images, this model can comprehensively consider the detail rendering capability of endoscopic images and quantify the image artifact interference caused by column fixed pattern noise. It takes into account the two core dimensions of image sharpness and noise defects, providing a reliable comprehensive evaluation basis for subsequent analysis and processing.

[0072] Optionally, the method further includes: determining quality assessment data for the endoscopic image based on the corrected sharpness assessment data and column fixed pattern noise assessment data.

[0073] The quality assessment data can also be determined based on the corrected sharpness assessment data and column fixed-pattern noise assessment data. For example, the corrected sharpness assessment data and column fixed-pattern noise assessment data can be input into a trained quality assessment data determination model for processing to obtain quality assessment data. The quality assessment data determination model includes, but is not limited to, neural network models and mathematical models. The quality assessment data determination model can be selected according to the requirements, and this invention does not impose any limitations.

[0074] Specifically, by using the corrected sharpness evaluation data and column fixed pattern noise evaluation data to determine the quality evaluation data of endoscopic images, the accuracy of the quality evaluation data can be further improved, providing a reliable comprehensive evaluation basis for subsequent analysis and processing.

[0075] S160. Determine the target imaging level among different imaging levels based on quality evaluation data.

[0076] The target imaging level is the optimal imaging level among multiple imaging levels. It is the optimal combination of imaging control parameters that can adapt to the current endoscopic acquisition scenario and balance image clarity with the noise interference of the fixed column pattern. The target imaging level can be determined based on quality evaluation data. For example, the maximum value among multiple quality evaluation data can be determined, and the imaging level corresponding to the quality evaluation data of the maximum value can be used as the target imaging level.

[0077] Specifically, by determining the maximum value among multiple quality evaluation data, the imaging level corresponding to the quality evaluation data with the maximum value is taken as the target imaging level, thus achieving accurate determination of the target imaging level and improving the efficiency of target imaging level determination.

[0078] The technical solution of this embodiment acquires an image sequence from the target endoscope, including endoscopic images at different imaging settings, providing comprehensive data support for subsequent analysis and processing. When the image sequence is valid, image quality information corresponding to each endoscopic image is obtained. Based on this image quality information, the sharpness evaluation data of the endoscopic images is determined, achieving accurate determination of the sharpness evaluation data and providing comprehensive data support for subsequent analysis and processing. Each endoscopic image is classified using a classification model, yielding classification results for each image. These results include column-fixed pattern noise information in the endoscopic images, achieving accurate and rapid classification of the endoscopic images and providing accurate data support for subsequent analysis and processing. The system provides data support; based on the classification results, it determines the column fixed-pattern noise evaluation data corresponding to the endoscopic image, achieving accurate determination of the column fixed-pattern noise evaluation data and providing accurate data support for subsequent analysis and processing; based on the sharpness evaluation data and the column fixed-pattern noise evaluation data, it determines the quality evaluation data of the endoscopic image, no longer relying solely on the sharpness index to judge the image quality, but comprehensively considering both the detail rendering capability of the endoscopic image and the image artifact interference caused by the column fixed-pattern noise, taking into account the two core dimensions of image sharpness and noise defects, providing a reliable comprehensive evaluation basis for subsequent analysis and processing; based on the quality evaluation data, it determines the target imaging level in different imaging levels, achieving accurate determination of the target imaging level and improving the efficiency of target imaging level determination.

[0079] Figure 2 This is a flowchart of another image processing method provided by an embodiment of the present invention. This embodiment is a refinement of the above embodiments. Based on the foregoing embodiments, it provides a detailed explanation of determining the quality evaluation data of endoscopic images based on sharpness evaluation data and column fixed pattern noise evaluation data. For specific implementation methods, please refer to the technical solution of this embodiment. Technical terms that are the same as or corresponding to those in the above embodiments will not be repeated here. Figure 2 As shown, the method specifically includes the following steps: S210. Acquire the image sequence captured by the target endoscope, including endoscope images at different imaging settings.

[0080] S220. If the image sequence is a valid sequence, obtain the image quality information corresponding to each endoscope image, and determine the clarity evaluation data of the endoscope image based on the image quality information.

[0081] Optionally, if the image sequence is invalid, a new image sequence can be acquired based on the target endoscope.

[0082] When an image sequence is invalid, it indicates that there are invalid endoscopic images among multiple endoscopic images in the sequence. To improve the accuracy of the target imaging setting, it is necessary to acquire a new image sequence based on the target endoscope.

[0083] Optionally, the validity detection process of the image sequence is as follows: Channel detection is performed on each endoscope image to obtain a first detection result corresponding to each endoscope image. The first detection result indicates whether the channels of the endoscope image meet the preset channel conditions; the validity detection of the image sequence is performed based on the first detection result to obtain a validity detection result; when the validity detection result is that the channels of each endoscope image meet the channel conditions, the image sequence is a valid sequence; when the validity detection result is that the channels of the endoscope image do not meet the channel conditions, the image sequence is an invalid sequence.

[0084] The first detection result is a determination of whether the channel attributes of the endoscopic image meet preset channel conditions. The first detection result can be obtained by performing channel detection on the endoscopic image. For example, detecting the number and order of channels in the endoscopic image; when both the number and order of channels in the endoscopic image meet the preset channel conditions, the first detection result is that the channels of the endoscopic image meet the channel conditions; when the number and / or order of channels in the endoscopic image do not meet the preset channel conditions, the first detection result is that the channels of the endoscopic image do not meet the channel conditions.

[0085] The validity test result determines whether an image sequence can undergo subsequent image quality assessment. Optionally, the validity test result can be that all channels of each endoscopic image meet the channel conditions, or that at least one channel of an endoscopic image does not meet the channel conditions. The validity test result can be determined based on the first test result. When at least one of the multiple first test results indicates that a channel of an endoscopic image does not meet the channel conditions, the validity test result is that at least one channel of an endoscopic image does not meet the channel conditions, and the corresponding image sequence is an invalid sequence; when multiple first test results indicate that all channels of an endoscopic image meet the channel conditions, the validity test result is that all channels of the endoscopic image meet the channel conditions, and the corresponding image sequence is a valid sequence.

[0086] Specifically, by performing channel detection on each endoscopic image, it is possible to identify endoscopic images with abnormal number of channels or disordered channel order, determine whether the image sequence is a valid sequence, and perform subsequent processing only on valid image sequences, thereby reducing invalid operations and saving computing resources.

[0087] Optionally, the validity detection process of the image sequence is as follows: exposure detection is performed on each endoscope image to obtain a second detection result corresponding to the endoscope image. The second detection result indicates whether the endoscope image is overexposed or underexposed. Based on the second detection result, the validity detection of the image sequence is performed to obtain a validity detection result. When the validity detection result is that none of the endoscope images are overexposed or underexposed, the image sequence is a valid sequence. When the validity detection result is that at least one endoscope image is overexposed or underexposed, the image sequence is an invalid sequence.

[0088] The second detection result is information characterizing the exposure state of the endoscopic image, used to identify whether the endoscopic image is overexposed or underexposed. Optionally, the second detection result can indicate that the endoscopic image is overexposed and / or underexposed, or it can indicate that the endoscopic image is neither overexposed nor underexposed. The second detection result can be obtained by performing exposure detection on the endoscopic image. For example, the endoscopic image can be input into a trained exposure detection model for processing to obtain the second detection result. The exposure detection model includes, but is not limited to, a neural network model. The exposure detection model can be selected according to requirements, and this invention is not limited thereto.

[0089] Optionally, the validity detection result can also be that none of the endoscopic images are overexposed or underexposed, or that at least one endoscopic image is overexposed or underexposed. The validity detection result can also be determined based on a second detection result. When at least one of the multiple second detection results indicates that an endoscopic image is overexposed and / or underexposed, the validity detection result is that at least one endoscopic image is overexposed and underexposed, and the corresponding image sequence is an invalid sequence; when all the multiple second detection results indicate that none of the endoscopic images are overexposed or underexposed, the validity detection result is that none of the endoscopic images are overexposed or underexposed, and the corresponding image sequence is a valid sequence.

[0090] Specifically, by detecting whether each endoscopic image is overexposed or underexposed, endoscopic images with abnormal exposure can be filtered out, and it can be determined whether the image sequence is a valid sequence. Only image sequences with valid sequences are processed, reducing invalid operations and saving computing resources.

[0091] Optionally, the validity detection process of the image sequence is as follows: Channel detection is performed on each endoscope image to obtain a first detection result corresponding to each endoscope image. The first detection result indicates whether the channels of the endoscope image meet the preset channel conditions. Exposure detection is performed on each endoscope image to obtain a second detection result corresponding to the endoscope image. The second detection result indicates whether the endoscope image is overexposed or underexposed. Based on the first and second detection results, the image sequence is valid to obtain a validity detection result. When the validity detection result indicates that the channels of each endoscope image meet the channel conditions and there is no overexposure or underexposure in any endoscope image, the image sequence is a valid sequence. When the validity detection result indicates that at least one endoscope image's channel does not meet the channel conditions and / or at least one endoscope image is overexposed or underexposed, the image sequence is an invalid sequence.

[0092] The validity detection result can also be that the channels of each endoscopic image meet the channel conditions and that none of the endoscopic images are overexposed or underexposed, or that at least one endoscopic image channel does not meet the channel conditions and / or at least one endoscopic image is overexposed or underexposed. The validity detection result can also be determined based on a first detection result and a second detection result. When the first detection result is that the channels of each endoscopic image meet the channel conditions and the second detection result is that none of the endoscopic images are overexposed or underexposed, the validity detection result is that the channels of each endoscopic image meet the channel conditions and none of the endoscopic images are overexposed or underexposed, and the corresponding image sequence is a valid sequence; when the first detection result is that the channels of each endoscopic image meet the channel conditions and the second detection result is that at least one endoscopic image is overexposed or underexposed, the validity detection result is that the channels of each endoscopic image meet the channel conditions and at least one endoscopic image is overexposed or underexposed, and the corresponding image sequence is an invalid sequence; when the first detection result is... If at least one channel of an endoscope image does not meet the channel condition and the second detection result is that none of the endoscope images are overexposed or underexposed, the validity detection result is that at least one channel of an endoscope image does not meet the channel condition and none of the endoscope images are overexposed or underexposed, and the corresponding image sequence is an invalid sequence; when the first detection result is that at least one channel of an endoscope image does not meet the channel condition and the second detection result is that at least one endoscope image is overexposed or underexposed, the validity detection result is that at least one channel of an endoscope image does not meet the channel condition and at least one endoscope image is overexposed or underexposed, and the corresponding image sequence is an invalid sequence.

[0093] Specifically, by performing channel detection and exposure detection on each endoscopic image, it is possible to identify endoscopic images with abnormal number of channels, disordered channel order, and abnormal exposure. This allows for the determination of whether the image sequence is valid. Subsequent processing is only performed on valid image sequences, reducing invalid computations and saving computing resources.

[0094] S230. Based on the classification model, classify each endoscope image to obtain the classification result corresponding to each endoscope image. The classification result includes column fixed pattern noise information in the endoscope image.

[0095] S240. Based on the classification results, determine the column fixed pattern noise evaluation data corresponding to the endoscopic image.

[0096] Optionally, after determining the column fixed pattern noise evaluation data of the endoscopic image based on the classification results, the method further includes: obtaining the column fixed pattern noise evaluation data threshold; filtering the column fixed pattern noise evaluation data based on the column fixed pattern noise evaluation data threshold to obtain the filtered column fixed pattern noise evaluation data.

[0097] The threshold for column-fixed pattern noise evaluation data can be preset. Filtering column-fixed pattern noise evaluation data based on this threshold can involve removing data below the threshold and retaining only data above or equal to the threshold. This data above or equal to the threshold is then used as the filtered column-fixed pattern noise evaluation data.

[0098] Specifically, by filtering the fixed-pattern noise evaluation data by using a threshold, endoscopic images with different levels of noise pollution can be quickly distinguished, reducing computational overhead and thus improving the accuracy of target imaging levels.

[0099] Optionally, after determining the column-fixed pattern noise evaluation data of the endoscope images based on the classification results, the method further includes: performing span detection on the column-fixed pattern noise evaluation data corresponding to each endoscope image to obtain a span detection result, wherein the span detection result characterizes whether the span of the column-fixed pattern noise evaluation data corresponding to each endoscope image is consistent; if the span detection result indicates that the span of the column-fixed pattern noise evaluation data corresponding to each endoscope image is consistent, updating the column-fixed pattern noise evaluation data corresponding to each endoscope image to obtain updated column-fixed pattern noise evaluation data.

[0100] The span detection result is a determination of the differences in the column-fixed pattern noise evaluation data corresponding to each endoscopic image in the image sequence. The span detection result can be determined based on the column-fixed pattern noise evaluation data corresponding to each endoscopic image. For example, the difference between multiple column-fixed pattern noise evaluation data is calculated as span information, which characterizes the deviation between the column-fixed pattern noise evaluation data. Multiple span information values ​​are compared with a preset span threshold. When all span information values ​​are less than or equal to the span threshold, the span detection result indicates that the span of the column-fixed pattern noise evaluation data corresponding to each endoscopic image is consistent. When at least one span information value is greater than the span threshold, the span detection result indicates that the span of the column-fixed pattern noise evaluation data corresponding to each endoscopic image is inconsistent.

[0101] When the span detection results show that the span of the column fixed-pattern noise evaluation data corresponding to each endoscopic image is consistent, it indicates that the severity of column fixed-pattern noise in images acquired at different imaging settings is relatively small. Therefore, it is difficult to distinguish the differences in noise suppression effectiveness between different imaging settings based on the existing column fixed-pattern noise evaluation data. Thus, it is necessary to update the column fixed-pattern noise evaluation data corresponding to each endoscopic image. For example, the column fixed-pattern noise evaluation data corresponding to each endoscopic image can be replaced with preset desired column fixed-pattern noise evaluation data.

[0102] Specifically, span detection is performed on the column fixed-pattern noise evaluation data corresponding to each endoscopic image. When the span detection result shows that the span of the column fixed-pattern noise evaluation data corresponding to each endoscopic image is consistent, the column fixed-pattern noise evaluation data corresponding to each endoscopic image is replaced with the preset expected column fixed-pattern noise evaluation data. The expected column fixed-pattern noise evaluation data is used as the updated column fixed-pattern noise evaluation data. This achieves the update when the span of the column fixed-pattern noise evaluation data is consistent, providing more accurate data support for subsequent analysis and processing.

[0103] Based on the above embodiments, after determining the sharpness evaluation data of the endoscope image based on image quality information, the method further includes: performing span detection on the sharpness evaluation data corresponding to each endoscope image to obtain a span detection result, wherein the span detection result indicates whether the span of the sharpness evaluation data corresponding to each endoscope image is consistent; if the span detection result indicates that the span of the sharpness evaluation data corresponding to each endoscope image is consistent, updating the sharpness evaluation data corresponding to each endoscope image to obtain updated sharpness evaluation data.

[0104] Optionally, the method further includes: when the span detection results show that the span of the column fixed pattern noise evaluation data corresponding to each endoscope image is inconsistent, normalizing the column fixed pattern noise evaluation data corresponding to each endoscope image based on a preset normalization rule to obtain the normalized column fixed pattern noise evaluation data corresponding to each endoscope image.

[0105] The normalization rules include, but are not limited to, the maximum-minimum normalization rule. When the span detection results show inconsistent spans in the column fixed-pattern noise evaluation data corresponding to each endoscopic image, the column fixed-pattern noise evaluation data corresponding to each endoscopic image is normalized according to the preset normalization rules. This results in normalized column fixed-pattern noise evaluation data for each endoscopic image, which unifies the numerical range of the column fixed-pattern noise evaluation data, eliminates the influence of differences in the original value ranges of the column fixed-pattern noise evaluation data, and thus helps improve the accuracy of the quality evaluation data and the accuracy of the target imaging level.

[0106] S250. Weight the sharpness evaluation data and the column fixed pattern noise evaluation data to obtain the quality evaluation data.

[0107] The quality assessment data can also be obtained by weighting the sharpness assessment data and the column fixed-pattern noise assessment data. For example, the weighted sum of the sharpness assessment data and the column fixed-pattern noise assessment data can be used as the quality assessment data.

[0108] Specifically, by calculating the weighted sum of the sharpness evaluation data and the column fixed pattern noise evaluation data as the quality evaluation data, the image quality is no longer judged solely by the sharpness index. This approach can comprehensively assess the detail rendering capability of endoscopic images and quantify the image artifact interference caused by column fixed pattern noise. It takes into account both the two core dimensions of image sharpness and noise defects, providing a reliable comprehensive evaluation basis for subsequent analysis and processing.

[0109] For example, the formula for calculating quality evaluation data is as follows: ; in, This represents quality evaluation data; This represents the data for clarity evaluation; This represents the weighting coefficients corresponding to the clarity evaluation data; This indicates the fixed-pattern noise evaluation data. This represents the weighting coefficients corresponding to the fixed-pattern noise evaluation data.

[0110] S260. Determine the target imaging level among different imaging levels based on quality evaluation data.

[0111] Optionally, the method further includes: sending the target imaging setting to the target endoscope; and the target endoscope adjusting its imaging setting to the target imaging setting.

[0112] The target imaging setting is sent to the target endoscope via a communication interface. This communication interface includes, but is not limited to, wired and wireless communication interfaces. Upon receiving the target imaging setting, the target endoscope triggers the setting adjustment logic to adjust the imaging setting to the target setting.

[0113] Specifically, the target imaging setting is transmitted via a communication link, enabling data interaction between the image processing unit and the target endoscope. This allows the target endoscope to receive the target imaging setting in a timely manner, thereby automatically configuring imaging parameters and eliminating the need for manual input of imaging settings, thus improving the efficiency of endoscopic imaging parameter adjustment. For example, see... Figure 3 , Figure 3 This is a flowchart of another image processing method provided in an embodiment of the present invention.

[0114] The technical solution of this embodiment acquires an image sequence from the target endoscope, including endoscopic images at different imaging settings, providing comprehensive data support for subsequent analysis and processing. When the image sequence is valid, image quality information corresponding to each endoscopic image is obtained. Based on this image quality information, the sharpness evaluation data of the endoscopic images is determined, achieving accurate determination of the sharpness evaluation data and providing comprehensive data support for subsequent analysis and processing. Each endoscopic image is classified using a classification model, yielding classification results for each image. These results include column-fixed pattern noise information in the endoscopic images, achieving accurate and rapid classification of the endoscopic images and providing accurate data support for subsequent analysis and processing. The system provides data support; based on the classification results, it determines the column fixed-pattern noise evaluation data corresponding to the endoscopic image, achieving accurate determination of the column fixed-pattern noise evaluation data and providing accurate data support for subsequent analysis and processing; it performs weighted processing on the sharpness evaluation data and the column fixed-pattern noise evaluation data to obtain quality evaluation data, no longer relying solely on the sharpness index to judge the image quality, but comprehensively considering both the detail rendering capability of the endoscopic image and the image artifact interference caused by the column fixed-pattern noise, taking into account the two core dimensions of image sharpness and noise defects, and providing a reliable comprehensive evaluation basis for subsequent analysis and processing; based on the quality evaluation data, it determines the target imaging level in different imaging levels, achieving accurate determination of the target imaging level and improving the efficiency of target imaging level determination.

[0115] Figure 4 This is a schematic diagram of the structure of an image processing device provided in an embodiment of the present invention. Figure 4As shown, the device includes an image sequence acquisition module 310, a sharpness evaluation data determination module 320, a classification result determination module 330, a column fixed pattern noise evaluation data determination module 340, a quality evaluation data determination module 350, and a target imaging level determination module 360.

[0116] The system includes: an image sequence acquisition module 310 for acquiring image sequences captured by the target endoscope, including endoscope images at different imaging levels; a sharpness evaluation data determination module 320 for acquiring image quality information corresponding to each endoscope image when the image sequence is valid, and determining sharpness evaluation data for the endoscope images based on the image quality information; a classification result determination module 330 for classifying each endoscope image based on a classification model to obtain classification results corresponding to each endoscope image, including column fixed pattern noise information in the endoscope images; a column fixed pattern noise evaluation data determination module 340 for determining column fixed pattern noise evaluation data corresponding to the endoscope images based on the classification results; a quality evaluation data determination module 350 for determining quality evaluation data for the endoscope images based on the sharpness evaluation data and the column fixed pattern noise evaluation data; and a target imaging level determination module 360 ​​for determining the target imaging level among different imaging levels based on the quality evaluation data.

[0117] The technical solution of this embodiment includes an image sequence acquisition module for acquiring image sequences captured by the target endoscope. These image sequences include endoscope images at different imaging settings, providing comprehensive data support for subsequent analysis and processing. A sharpness evaluation data determination module, assuming the image sequence is valid, acquires image quality information corresponding to each endoscope image and determines the sharpness evaluation data based on this information, achieving accurate determination of the endoscope image sharpness evaluation data and providing comprehensive data support for subsequent analysis and processing. A classification result determination module classifies each endoscope image based on a classification model, obtaining classification results for each endoscope image. These classification results include column-fixed pattern noise information in the endoscope images, achieving accurate and rapid classification of endoscope images and providing accurate data support for subsequent analysis and processing. The system includes: a fixed-pattern noise evaluation data determination module, used to determine the fixed-pattern noise evaluation data corresponding to the endoscopic image based on the classification results, achieving accurate determination of the fixed-pattern noise evaluation data and providing accurate data support for subsequent analysis and processing; a quality evaluation data determination module, used to determine the quality evaluation data of the endoscopic image based on the sharpness evaluation data and the fixed-pattern noise evaluation data, no longer relying solely on the sharpness index to judge the image quality, but comprehensively considering both the detail rendering capability of the endoscopic image and the image artifact interference caused by the fixed-pattern noise, taking into account the two core dimensions of image sharpness and noise defects, providing a reliable comprehensive evaluation basis for subsequent analysis and processing; and a target imaging level determination module, used to determine the target imaging level among different imaging levels based on the quality evaluation data, achieving accurate determination of the target imaging level and improving the efficiency of target imaging level determination.

[0118] Based on the above embodiments, optionally, the classification model includes a global spatial feature extraction module, a height-direction spatial feature extraction module, a pooling module, and a column-level classification module; the global spatial feature extraction module extracts global spatial features from the endoscopic image to obtain global spatial features; the height-direction spatial feature extraction module extracts features from the global spatial features, downsamples along the height direction corresponding to the global spatial features while maintaining the width direction resolution to obtain height-direction spatial features; the pooling module performs pooling processing on the height-direction spatial features to obtain target features; and the column-level classification module performs classification processing on the target features to obtain classification results.

[0119] Optionally, the image quality information includes at least one of sharpness information, local contrast information, glare information, and high-frequency noise information; the sharpness evaluation data determination module 320 is further configured to: perform weighted processing on the sharpness information, local contrast information, glare information, and high-frequency noise information to obtain sharpness evaluation data.

[0120] Optionally, the device further includes a sharpness evaluation data correction module, used to: acquire edge naturalness information, oversharpness penalty information and noise excess penalty information corresponding to the endoscopic image; and correct the sharpness evaluation data corresponding to the endoscopic image based on the edge naturalness information, oversharpness penalty information and noise excess penalty information to obtain corrected sharpness evaluation data.

[0121] Optionally, the quality assessment data determination module 350 is also used to: determine the quality assessment data of the endoscopic image based on the corrected sharpness assessment data and the column fixed pattern noise assessment data.

[0122] Optionally, the column fixed pattern noise information includes the encoding information corresponding to each vertical pixel column in the endoscopic image, and the encoding information characterizes whether column fixed pattern noise exists in the vertical pixel column; the column fixed pattern noise evaluation data determination module 340 is further used to: determine the column fixed pattern noise ratio, the maximum continuous column fixed pattern noise width, the number of column fixed pattern noise clusters, the odd-even column bias, and the column fixed pattern noise risk index corresponding to the endoscopic image based on the encoding information; and perform weighted processing on the column fixed pattern noise ratio, the maximum continuous column fixed pattern noise width, the number of column fixed pattern noise clusters, the odd-even column bias, and the column fixed pattern noise risk index to obtain the column fixed pattern noise evaluation data.

[0123] Optionally, the column fixed pattern noise evaluation data determination module 340 is also used to: perform weighted processing on the column fixed pattern noise ratio, the maximum continuous column fixed pattern noise width, the number of column fixed pattern noise clusters and the odd-even column bias to obtain the column fixed pattern noise risk index.

[0124] Optionally, the quality evaluation data determination module 350 is also used to: perform weighted processing on the sharpness evaluation data and the column fixed pattern noise evaluation data to obtain quality evaluation data.

[0125] Optionally, the device further includes an image sequence reacquisition module for: reacquiring a new image sequence based on the target endoscope when the image sequence is invalid.

[0126] Optionally, the clarity evaluation data determination module 320 is further configured to: perform channel detection on each endoscope image to obtain a first detection result corresponding to each endoscope image, wherein the first detection result characterizes whether the channel of the endoscope image meets the preset channel conditions; And / or, Exposure detection is performed on each endoscope image to obtain the second detection result corresponding to the endoscope image. The second detection result indicates whether the endoscope image is overexposed or underexposed. The validity of the image sequence is determined based on the first detection result and / or the second detection result to obtain the validity detection result. The image sequence is considered valid when the validity test result is that the channels of the endoscope image meet the channel conditions, and / or that none of the endoscope images are overexposed or underexposed; The image sequence is invalid when the validity test result is that the channel of the endoscope image does not meet the channel condition, and / or at least one endoscope image is overexposed and underexposed.

[0127] Optionally, the device further includes a column fixed pattern noise evaluation data filtering module, used to: after determining the column fixed pattern noise evaluation data of the endoscopic image based on the classification result, obtain the column fixed pattern noise evaluation data threshold; filter the column fixed pattern noise evaluation data based on the column fixed pattern noise evaluation data threshold to obtain the filtered column fixed pattern noise evaluation data.

[0128] Optionally, the device further includes a column fixed pattern noise evaluation data update module, used for: after determining the column fixed pattern noise evaluation data corresponding to the endoscope image based on the classification result, performing span detection on the column fixed pattern noise evaluation data corresponding to each endoscope image to obtain a span detection result, the span detection result indicating whether the span of the column fixed pattern noise evaluation data corresponding to each endoscope image is consistent; if the span detection result indicates that the span of the column fixed pattern noise evaluation data corresponding to each endoscope image is consistent, updating the column fixed pattern noise evaluation data corresponding to each endoscope image to obtain updated column fixed pattern noise evaluation data.

[0129] Optionally, the column fixed pattern noise evaluation data update module is also used to: when the span detection result shows that the span of the column fixed pattern noise evaluation data corresponding to each endoscope image is inconsistent, normalize the column fixed pattern noise evaluation data corresponding to each endoscope image based on the preset normalization rule to obtain the normalized column fixed pattern noise evaluation data corresponding to each endoscope image.

[0130] Optionally, the device further includes a target imaging level adjustment module for: sending the target imaging level to the target endoscope; and adjusting the imaging level of the target endoscope to the target imaging level.

[0131] Optionally, the image quality information includes Tenengrad gradient sharpness features, Brenner gray-level difference contrast features, Laplacian variance sharpness features, gradient magnitude percentile information, and sharpness information; the sharpness evaluation data determination module 320 is further used to: perform weighted processing on the Tenengrad gradient sharpness features, Brenner gray-level difference contrast features, Laplacian variance sharpness features, and gradient magnitude percentile information to obtain sharpness information.

[0132] Optionally, the image quality information includes first gray-level percentile information, second gray-level percentile information, and local contrast information; the sharpness evaluation data determination module 320 is also used to: determine local contrast information based on the first gray-level percentile information and the second gray-level percentile information.

[0133] Optionally, the image quality information includes the proportion of overexposed pixels, the third grayscale percentile information, and glare information; the sharpness evaluation data determination module 320 is further configured to: perform weighted processing on the proportion of overexposed pixels to obtain first glare evaluation data; calculate the difference between the third grayscale percentile information and the preset bright pixel threshold, and determine second glare evaluation data based on the difference between the three grayscale percentile information and the preset bright pixel threshold; and sum the first glare evaluation data and the second glare evaluation data to obtain glare information.

[0134] Optionally, the image quality information includes grayscale enhancement information, grayscale smoothing information, and high-frequency noise information at each pixel location; the sharpness evaluation data determination module 320 is further configured to: determine the grayscale residual information corresponding to each pixel location based on the grayscale enhancement information and the grayscale smoothing information at the corresponding pixel location; determine the standard deviation corresponding to the grayscale residual information based on the grayscale residual information at each pixel location; and determine the high-frequency noise information based on the standard deviation.

[0135] Optionally, the image quality information includes grayscale enhancement information, edge binary information, and edge density information at each pixel location; the sharpness evaluation data determination module 320 is also used to: determine the proportion of edge pixels based on the edge binary information; and determine the edge density information based on the proportion of edge pixels.

[0136] The image processing apparatus provided in this embodiment of the invention can execute an image processing method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of executing the method.

[0137] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. The electronic device 10 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (such as helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0138] like Figure 5As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 may also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0139] Multiple components in electronic device 10 are connected to input / output (I / O) interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of monitors, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0140] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as an image processing method.

[0141] In some embodiments, an image processing method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or mounted on electronic device 10 via read-only memory (ROM) 12 and / or communication unit 19. When the computer program is loaded into random access memory (RAM) 13 and executed by processor 11, one or more steps of an image processing method described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to perform an image processing method by any other suitable means (e.g., by means of firmware).

[0142] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transferring data and instructions to the storage system, the at least one input device, and the at least one output device.

[0143] A computer program for implementing an image processing method of the present invention can be written in any combination of one or more programming languages. Such computer programs can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer program causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The computer program can be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0144] This invention also provides a computer-readable storage medium storing computer instructions for causing a processor to execute an image processing method, the method comprising: The process involves acquiring an image sequence from the target endoscope, including images of the endoscope at different imaging settings. If the image sequence is valid, image quality information is obtained for each endoscope image, and the sharpness evaluation data is determined based on this information. Each endoscope image is then classified using a classification model to obtain classification results, including column-fixed pattern noise information. Based on the classification results, column-fixed pattern noise evaluation data is determined for each endoscope image. Finally, based on the sharpness evaluation data and the column-fixed pattern noise evaluation data, the target imaging setting is determined within each imaging setting.

[0145] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0146] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0147] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0148] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0149] This invention also provides a computer program product, including a computer program that, when executed by a processor, implements an image processing method according to any embodiment of the invention.

[0150] In implementing the computer program product, computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0151] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0152] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. An image processing method, characterized in that, include: Acquire an image sequence captured by the target endoscope, the image sequence including endoscope images at different imaging settings; If the image sequence is a valid sequence, obtain the image quality information corresponding to each of the endoscope images, and determine the clarity evaluation data of the endoscope images based on the image quality information; Based on the classification model, each of the endoscope images is classified to obtain the classification result corresponding to each endoscope image. The classification result includes column fixed pattern noise information in the endoscope image. Based on the classification results, determine the column fixed pattern noise evaluation data corresponding to the endoscopic image; The quality evaluation data of the endoscope image is determined based on the sharpness evaluation data and the column fixed pattern noise evaluation data. The target imaging level is determined based on the quality evaluation data among the different imaging levels.

2. The method according to claim 1, characterized in that, The classification model includes a global spatial feature extraction module, a height-oriented spatial feature extraction module, a pooling module, and a column-level classification module; The global spatial feature extraction module performs global spatial feature extraction on the endoscopic image to obtain global spatial features; The height direction spatial feature extraction module extracts features from the global spatial features by downsampling along the height direction corresponding to the global spatial features while maintaining the width direction resolution, thereby obtaining the height direction spatial features; The pooling module performs pooling processing on the spatial features in the height direction to obtain the target features; The column-level classification module performs classification processing on the target features to obtain the classification result.

3. The method according to claim 1, characterized in that, The image quality information includes at least one of sharpness information, local contrast information, glare information, and high-frequency noise information; The step of determining the sharpness evaluation data of the endoscopic image based on the image quality information includes: The sharpness information, the local contrast information, the glare information, the high-frequency noise information, and the edge density information are weighted and processed to obtain the sharpness evaluation data.

4. The method according to claim 3, characterized in that, The method further includes: Obtain edge naturalness information, oversharpness penalty information, and noise excess penalty information corresponding to the endoscopic image; Based on the edge naturalness information, the oversharpness penalty information, and the noise excess penalty information, the sharpness evaluation data corresponding to the endoscopic image is corrected to obtain the corrected sharpness evaluation data.

5. The method according to claim 4, characterized in that, The method further includes: The quality evaluation data of the endoscopic image is determined based on the corrected sharpness evaluation data and the column fixed pattern noise evaluation data.

6. The method according to claim 1, characterized in that, The column fixed pattern noise information includes the encoding information corresponding to each vertical pixel column in the endoscope image, and the encoding information indicates whether there is column fixed pattern noise in the vertical pixel column; The step of determining the column fixed-pattern noise evaluation data corresponding to the endoscopic image based on the classification result includes: Based on the encoded information, determine the column fixed pattern noise ratio, the maximum continuous column fixed pattern noise width, the number of column fixed pattern noise clusters, the odd / even column bias, and the column fixed pattern noise risk index corresponding to the endoscopic image. The column fixed pattern noise ratio, the maximum continuous column fixed pattern noise width, the number of column fixed pattern noise clusters, the odd / even column bias, and the column fixed pattern noise risk index are weighted to obtain the column fixed pattern noise evaluation data.

7. The method according to claim 6, characterized in that, The process for determining the fixed-pattern noise risk index is as follows: The column fixed pattern noise ratio, the maximum continuous column fixed pattern noise width, the number of column fixed pattern noise clusters, and the odd / even column bias are weighted to obtain the column fixed pattern noise risk index.

8. The method according to claim 1, characterized in that, The process of determining the quality assessment data of the endoscopic image based on the sharpness assessment data and the column fixed pattern noise assessment data includes: The clarity evaluation data and the column fixed pattern noise evaluation data are weighted to obtain the quality evaluation data.

9. The method according to claim 1, characterized in that, The method further includes: If the image sequence is invalid, a new image sequence is acquired based on the target endoscope.

10. The method according to claim 1, characterized in that, The validity detection process for the image sequence is as follows: Channel detection is performed on each of the endoscope images to obtain a first detection result corresponding to each endoscope image. The first detection result indicates whether the channel of the endoscope image meets the preset channel conditions. And / or, Exposure detection is performed on each of the endoscope images to obtain a second detection result corresponding to the endoscope image. The second detection result indicates whether the endoscope image is overexposed or underexposed. Based on the first detection result and / or the second detection result, the image sequence is subjected to validity detection to obtain a validity detection result; The image sequence is considered valid when the validity detection result is that the channels of each endoscope image meet the channel conditions, and / or that none of the endoscope images are overexposed or underexposed. The image sequence is invalid when the validity detection result is that at least one channel of the endoscope image does not meet the channel condition, and / or at least one endoscope image is overexposed or underexposed.

11. The method according to claim 1, characterized in that, After determining the column-fixed pattern noise evaluation data of the endoscopic image based on the classification results, the method further includes: Obtain the threshold for noise evaluation data in a fixed column pattern; The column-fixed pattern noise evaluation data is filtered based on the threshold of the column-fixed pattern noise evaluation data to obtain the filtered column-fixed pattern noise evaluation data.

12. The method according to claim 1, characterized in that, After determining the column-fixed pattern noise evaluation data corresponding to the endoscopic image based on the classification results, the method further includes: Span detection is performed on the column fixed pattern noise evaluation data corresponding to each of the endoscopic images to obtain span detection results. The span detection results characterize whether the span of the column fixed pattern noise evaluation data corresponding to each of the endoscopic images is consistent. If the span detection result shows that the span of the column fixed pattern noise evaluation data corresponding to each of the endoscope images is consistent, the column fixed pattern noise evaluation data corresponding to each of the endoscope images is updated to obtain the updated column fixed pattern noise evaluation data.

13. The method according to claim 12, characterized in that, The method further includes: When the span detection results show that the span of the column fixed pattern noise evaluation data corresponding to each endoscope image is inconsistent, the column fixed pattern noise evaluation data corresponding to each endoscope image is normalized based on the preset normalization rule to obtain the normalized column fixed pattern noise evaluation data corresponding to each endoscope image.

14. The method according to claim 1, characterized in that, The method further includes: Send the target imaging setting to the target endoscope; The target endoscope will adjust the imaging setting to the target imaging setting.

15. The method according to claim 1, characterized in that, The image quality information includes Tenengrad gradient sharpness features, Brenner gray-level difference contrast features, Laplacian variance sharpness features, gradient magnitude percentile information, and sharpness information. The process for determining the sharpness information is as follows: The sharpness information is obtained by weighting the Tenengrad gradient sharpness feature, the Brenner gray-level difference contrast feature, the Laplacian variance sharpness feature, and the gradient magnitude percentile information.

16. The method according to claim 1, characterized in that, The image quality information includes first gray percentile information, second gray percentile information, and local contrast information; The process for determining the local contrast information is as follows: The local contrast information is determined based on the first grayscale percentile information and the second grayscale percentile information.

17. The method according to claim 1, characterized in that, The image quality information includes the proportion of overexposed pixels, the third grayscale percentile information, and glare information; The process for determining the glare information is as follows: The overexposed pixel ratio is weighted to obtain the first glare evaluation data; Calculate the difference between the third grayscale percentile information and the preset bright pixel threshold, and determine the second glare evaluation data based on the difference between the three grayscale percentile information and the preset bright pixel threshold; The glare information is obtained by summing the first glare evaluation data and the second glare evaluation data.

18. The method according to claim 1, characterized in that, The image quality information includes grayscale enhancement information, grayscale smoothing information, and high-frequency noise information at each pixel location; The process for determining the high-frequency noise information is as follows: The grayscale residual information corresponding to each pixel position is determined based on the grayscale enhancement information and the grayscale smoothing information of each pixel position. The standard deviation corresponding to the grayscale residual information is determined based on the grayscale residual information at each pixel position; The high-frequency noise information is determined based on the standard deviation.

19. An image processing apparatus, characterized in that, include: An image sequence acquisition module is used to acquire an image sequence captured by the target endoscope, the image sequence including endoscope images at different imaging settings; The sharpness evaluation data determination module is used to obtain image quality information corresponding to each of the endoscope images when the image sequence is a valid sequence, and to determine the sharpness evaluation data of the endoscope images based on the image quality information. The classification result determination module is used to classify each of the endoscope images based on the classification model to obtain the classification result corresponding to each of the endoscope images. The classification result includes column fixed pattern noise information in the endoscope image. A column-fixed pattern noise evaluation data determination module is used to determine the column-fixed pattern noise evaluation data corresponding to the endoscopic image based on the classification result. A quality assessment data determination module is used to determine the quality assessment data of the endoscopic image based on the sharpness assessment data and the column fixed pattern noise assessment data; The target imaging level determination module is used to determine the target imaging level among different imaging levels based on the quality evaluation data.

20. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the image processing method according to any one of claims 1-18.