AI image enhancement method for display screen

Abstract

The invention relates to the technical field of medical image processing, in particular to an AI image enhancement method for a display screen. The method comprises the following steps: acquiring an endoscopic image data set, performing key frame processing on the endoscopic image data set, and determining a key image data set; acquiring a multi-light-source parameter set of the data collection terminal, performing endogenous enhancement processing on the key image data set based on the multi-light-source parameter set of the data collection terminal, and determining an endogenous image data set; obtaining a display terminal environment light source parameter set, performing exogenous enhancement processing on the internal source image data set based on the display terminal environment light source parameter set, and determining a composite enhanced image data set; and performing composite mark enhancement processing on the endoscopic image data set according to the composite enhanced image data set, and determining and outputting an enhanced image data set. The endoscopic image can still keep a good display effect under the dual influence of the environment where the data collection terminal is located and the environment where the display terminal is located, and image details are guaranteed.

Description

Technical Field

[0001] The present application relates to the field of medical image processing technology, and in particular to an AI image enhancement method for a display screen. Background Art

[0002] With the advancement of medical imaging technology, endoscopes have been widely used in various in vivo examinations and treatments of the digestive tract, respiratory tract, etc. When performing endoscopic examinations, doctors need to accurately judge the condition of the lesions. Clear medical images can improve the accuracy of diagnosis and avoid misdiagnosis and missed diagnosis.

[0003] However, the existing endoscopic image processing technology is easily affected by the dual influence of the environment in which the endoscope data collection terminal is located and the environment in which the image display terminal is located, which may cause the loss of details of the endoscopic image presented on the display terminal, thereby affecting the doctor's ability to judge the endoscopic image. Summary of the invention

[0004] The present application provides an AI image enhancement method for a display screen to solve the above-mentioned technical problems.

[0005] In a first aspect, the present application provides an AI image enhancement method for a display screen, the method comprising:

[0006] Acquire an endoscopic image data set, perform key frame processing on the endoscopic image data set, and determine a key image data set;

[0007] Acquire a multi-light source parameter set of a data collection terminal, and perform endogenous enhancement processing on the key image data set based on the multi-light source parameter set of the data collection terminal to determine an endogenous image data set;

[0008] Acquire a display terminal environment light source parameter set, and based on the display terminal environment light source parameter set, perform exogenous enhancement processing on the endogenous image data set to determine a composite enhanced image data set;

[0009] According to the composite enhanced image data set, composite marking enhancement processing is performed on the endoscopic image data set to determine and output an enhanced image data set.

[0010] Through this solution, key frame data in an endoscopic image data set is acquired and processed to obtain a key image data set, so as to reduce data redundancy and improve the efficiency of subsequent image enhancement processing. According to the multi-light source parameter set of the data collection terminal, the key image data set is endogenously enhanced to obtain an endogenous image data set. On this basis, combined with the display terminal environmental light source parameter set, the endogenous image data set is exogenously enhanced to obtain a composite enhanced image data set, which effectively reduces the interference of endogenous and exogenous light on the endoscopic image, so that the endoscopic image can still maintain a good display effect under the dual influence of the environment of the data collection terminal and the environment of the display terminal, and the image details are guaranteed to assist doctors in medical diagnosis. According to the composite enhanced image data set, the endoscopic image data set is further compositely marked for enhancement, and the corresponding enhanced image data set is determined and provided to the doctor, while maintaining the image enhancement processing effect, the data continuity is not destroyed.

[0011] Optionally, performing key frame processing on the endoscopic image data set to determine a key image data set includes:

[0012] Performing frame extraction processing on the endoscopic image data set to determine an endoscopic image data set;

[0013] According to a preset preprocessing rule, the endoscopic image data set is subjected to denoising and normalization processing to determine a normalized image data set;

[0014] Based on the normalized image data set, extract the lesion area and contour of each endoscopic image to determine the lesion estimated area and lesion contour features corresponding to each endoscopic image;

[0015] Based on the estimated lesion area and the lesion contour features, key frame processing is performed on the endoscopic image data set to determine a number of key frame images, and the key image data set is constructed based on these.

[0016] Through this scheme, frame extraction processing is performed on the endoscopic image data set to determine the endoscopic image data set, and according to the preset preprocessing rules, denoising and normalization processing is performed on the endoscopic image data set to determine the normalized image data set, so as to reduce noise interference and unify the data dimension, and improve the accuracy and efficiency of subsequent key frame processing. On this basis, the lesion estimated area and lesion contour features corresponding to each endoscopic image are extracted, and key frame processing is performed on the endoscopic image data set accordingly, so as to construct a key image data set, so that the key frame image conforms to the dynamic change characteristics of the endoscope, and significantly reduces data redundancy without losing image details.

[0017] Optionally, based on the estimated lesion area and the lesion contour features, key frame processing is performed on the endoscopic image dataset to determine a number of key frame images, and the key image dataset is constructed based on them, which is specifically the following formula:

[0018] K={I t ||A t -A t-1 |>ΔA or H(C t ,C t-1 )>D th for all t};

[0019] Wherein, K is the key image dataset, I t is the endoscopic image corresponding to the current frame time t, A t is the estimated area of ​​the lesion corresponding to the endoscopic image at the current frame time t, A t-1 is the estimated area of ​​the lesion corresponding to the endoscopic image at the previous frame time t-1, ΔA is the preset area change threshold, C t is the lesion contour feature corresponding to the endoscopic image at the current frame time t, C t-1 is the lesion contour feature corresponding to the endoscopic image at the previous frame time t-1, H() is the Hausdorff distance function, D th is the preset feature change threshold.

[0020] Through this scheme, mathematical analysis methods are used to perform key frame processing on the endoscopic image dataset according to the estimated lesion area and lesion contour characteristics, and several key frame images are extracted. Based on this, a key image dataset is constructed, so that the key frame image determination process is highly matched with the changes in the estimated lesion area and lesion contour characteristics under the dynamically changing endoscopic perspective, thereby improving the accuracy of the key frame images and the comprehensiveness and robustness of the key frame extraction process.

[0021] Optionally, the multi-light source parameter set of the data collection terminal includes multi-light source intensity information and multi-light source angle information, and the performing endogenous enhancement processing on the key image data set based on the multi-light source parameter set of the data collection terminal to determine the endogenous image data set includes:

[0022] Extracting the light source intensity of different light source illumination areas in each key image according to the multiple light source intensity information;

[0023] Analyze the key image data set to extract the light intensity value corresponding to each pixel in each key image;

[0024] Based on the light source illumination area to which the pixel point belongs, and according to the light source intensity and the light intensity value, light intensity normalization processing is performed on different light source illumination areas in each key image to determine a light intensity normalization processing result;

[0025] Determine the incident angle and reflection angle of different light sources in each key image within the corresponding light source illumination area according to the multi-light source angle information;

[0026] Based on the light source intensity, according to the incident angle and the reflection angle, determine the reflected light intensity values ​​of different light source illumination areas in each key image, and thereby determine the average reflected light intensity;

[0027] According to the reflected light intensity mean, the light source intensity and the light intensity normalization processing result, the light intensity in each pixel in the key image is adjusted to determine the adjusted light intensity value;

[0028] According to the adjusted light intensity value, each key image is subjected to endogenous enhancement processing, and an endogenous image data set is constructed according to all processed key images.

[0029] Through this scheme, the light source intensity of different light source illumination areas in each key image is extracted, and the brightness value corresponding to each pixel point in each key image in the key image data set is extracted as its corresponding light intensity value. By extracting the incident angle and the reflection angle, the multi-light source illumination effect of the endoscope can be modeled more accurately, reflecting the reflection and shadow effects of the endoscope light at different angles. By analyzing the reflected light intensity value of each light source illumination area, the real reflection characteristics of human tissue under the illumination conditions of different light source illumination areas are reflected. The average reflected light intensity obtained on this basis is used as the basis for subsequent reduction of local brightness fluctuations. Then, according to the reflected light intensity average, light source intensity and light intensity normalization processing results, the adjusted light intensity value required under multi-light source interference is reduced for each pixel point in the key image, and the brightness value of the pixel point in each key image is adjusted accordingly to achieve endogenous enhancement processing of each key image.

[0030] Optionally, based on the light source intensity, the reflected light intensity value of the different light source illumination areas in each key image is determined according to the incident angle and the reflection angle, specifically as follows:

[0031] R j =L j ·(k d ·cos(θ j )+k s ·(max(cos(φ j ),0)) G );

[0032] Among them, Rj is the reflected light intensity value corresponding to the j-th light source illumination area in the current key image, L j is the light source intensity corresponding to the j-th light source illumination area in the current key image, k d is the preset tissue diffuse reflectance, θ j is the incident angle corresponding to the j-th light source illumination area in the current key image, k s is the preset tissue mirror reflection coefficient, φ j is the reflection angle corresponding to the j-th light source illumination area in the current key image, and G is the preset gloss factor.

[0033] Through this scheme, mathematical analysis methods are used to scientifically quantify the reflected light intensity values ​​of different light source illumination areas in each key image based on the light source intensity, according to the incident angle and reflection angle, so that the reflected light intensity values ​​are consistent with the diffuse reflection characteristics and mirror reflection characteristics of human tissue to light in the current key image, thereby improving the accuracy and comprehensiveness of the reflected light intensity values.

[0034] Optionally, the light intensity in each pixel in each key image is adjusted according to the reflected light intensity mean, the light source intensity and the light intensity normalization processing result to determine the adjusted light intensity value, which is specifically the following formula:

[0035]

[0036] Among them, I t ′ (x, y) is the adjusted light intensity value of the pixel at the coordinate point (x, y) in the current key image, I t (x, y) is the light intensity value of the pixel at the coordinate point (x, y) in the current key image, L j is the light source intensity corresponding to the j-th light source illumination area in the current key image, is the light intensity normalization processing result, n is the total number of light source irradiation areas in the current key image, and R is the mean value of the reflected light intensity corresponding to the current key image.

[0037] Through this scheme, mathematical analysis methods are used to scientifically quantify the adjusted light intensity value required for the pixel point according to the mean value of the reflected light intensity, the light source intensity and the light intensity normalization processing results. The balanced adjustment based on the light source intensity difference provided by the mathematical formula is used to take the obtained adjusted light intensity value as the basis for adjusting the pixel brightness. This can make the light intensity of each pixel point in the endoscopic image under the influence of different light sources more consistent and reasonable, thereby reducing the interference of multiple light sources and enhancing the endoscopic image display effect.

[0038] Optionally, the display terminal ambient light parameter set includes basic ambient light intensity, ambient light fluctuation amplitude, ambient light fluctuation frequency and ambient light phase, and the exogenous enhancement processing is performed on the endogenous image data set based on the display terminal ambient light parameter set to determine the composite enhanced image data set, including:

[0039] Determining an ambient light characteristic value according to the basic ambient light intensity, the ambient light fluctuation amplitude, the ambient light fluctuation frequency and the ambient light illumination phase;

[0040] Based on the ambient light characteristic value, correct the color channel corresponding to each pixel in each image in the internal source image data set to determine the corresponding corrected pixel;

[0041] According to all the corrected pixel points corresponding to each image in the internal source image data set, a composite enhanced image corresponding to each image is generated, and the composite enhanced image data set is constructed based on this.

[0042] Through this scheme, the ambient light characteristic value that can reflect the comprehensive influence of the ambient light is determined according to the basic ambient light intensity, the ambient light fluctuation amplitude, the ambient light fluctuation frequency and the ambient light phase, and the color channel corresponding to each pixel in each image in the endogenous image data set is corrected to determine the corresponding corrected pixel to generate a composite enhanced image, and the composite enhanced image data set is constructed to reduce the light interference in the environment where the display is located, and further improve the display effect of the enhanced endoscopic image.

[0043] Optionally, based on the ambient light characteristic value, the color channel corresponding to each pixel in each image in the internal source image data set is corrected to determine the corresponding corrected pixel, which is specifically the following formula:

[0044] E′ t (x,y)=∑ c∈{R,G,B} (E′ t (x,y,c)+λ·M);

[0045] Among them, E′ t (x, y) is the corrected pixel point at the coordinate point (x, y) in the image corresponding to the current frame time t, c is the image color channel, c∈{R,G,B} indicates the use of RGB color channel, α is the preset channel influence coefficient, E′ t (x, y, c) is the image color channel of the pixel at the coordinate point (x, y) in the image corresponding to the current frame time t, λ is the preset environmental feature influence coefficient, and M is the ambient light feature value.

[0046] Through this scheme, mathematical analysis methods are used to describe the influence of ambient light on the color channels of pixels through mathematical language according to the ambient light characteristic values, so as to correct the color channels corresponding to each pixel in each image in the endogenous image data set, determine the corresponding corrected pixel, reduce the interference of ambient light on the display effect of the endoscopic image on the display terminal, and enhance the display effect of the endoscopic image on the display terminal.

[0047] Optionally, the ambient light characteristic value is determined according to the basic ambient light intensity, the ambient light fluctuation amplitude, the ambient light fluctuation frequency and the ambient light illumination phase, specifically as the following formula:

[0048] M = L0 + ΔL·sin(ωT + φ);

[0049] Among them, M is the ambient light characteristic value, L0 is the basic ambient light intensity, ΔL is the ambient light fluctuation amplitude, ω is the ambient light fluctuation frequency, T is the preset ambient light monitoring time, and φ is the ambient light phase.

[0050] Through this scheme, mathematical analysis is used to describe the periodic fluctuation of ambient light above the basic ambient light intensity through mathematical language according to the basic ambient light intensity, ambient light fluctuation amplitude, ambient light fluctuation frequency and ambient light phase, and the ambient light characteristic value reflecting the comprehensive influence of ambient light is accurately quantified to improve the accuracy and scientificity of the ambient light characteristic value.

[0051] Optionally, performing compound marking enhancement processing on the endoscopic image dataset according to the compound enhanced image dataset to determine and output an enhanced image dataset includes:

[0052] Based on the composite enhanced image data set, analyzing the endoscopic image data set, determining a number of similar frame images corresponding to each composite enhanced image in the endoscopic image data set;

[0053] According to the corrected pixel points of each composite enhanced image of the composite enhanced image data set, composite enhancement processing is performed on a plurality of similar frame images of each composite enhanced image to determine an enhanced similar frame image set;

[0054] The enhanced image data set is constructed and outputted according to the composite enhanced image data set and the enhanced similar frame image set.

[0055] Through this scheme, a number of similar frame images corresponding to each composite enhanced image in the composite enhanced image data set in the endoscopic image data set are extracted, and the similar frame images are subjected to composite enhancement processing according to the corrected pixel points corresponding to the composite enhanced images to obtain an enhanced similar frame image set. On this basis, an enhanced image data set is generated and output to maintain the continuity of the enhanced features in the corresponding endoscopic images, thereby providing a positive auxiliary effect for doctors' medical diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0057] Figure 1 A schematic diagram of an application scenario provided for an embodiment of the present application;

[0058] Figure 2 A flowchart of an AI image enhancement method for a display screen provided in one embodiment of the present application. DETAILED DESCRIPTION

[0059] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0060] In addition, the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article, unless otherwise specified, generally means that the associated objects before and after are in an "or" relationship.

[0061] The embodiments of the present application are further described in detail below in conjunction with the drawings in the specification.

[0062] Under the dual influence of the environment in which the endoscope data collection terminal is located and the environment in which the image display terminal is located, the existing endoscopic image processing technology is prone to cause the loss of details of the endoscopic image presented on the display terminal, thereby affecting the doctor's ability to judge the endoscopic image.

[0063] Based on this, the present application provides an AI image enhancement method for display screens. The key frame data in the endoscopic image data set is acquired and processed to obtain a key image data set to reduce data redundancy and improve the efficiency of subsequent image enhancement processing. According to the multi-light source parameter set of the data collection terminal, the key image data set is endogenously enhanced to obtain an endogenous image data set, and on this basis, combined with the display terminal environment light source parameter set, the endogenous image data set is exogenously enhanced to obtain a composite enhanced image data set, effectively reducing the interference of endogenous light and exogenous light on the endoscopic image, so that the endoscopic image can still maintain a good display effect under the dual influence of the environment in which the data collection terminal is located and the environment in which the display terminal is located, ensuring image details to assist doctors in medical diagnosis, and further composite marking enhancement processing is performed on the endoscopic image data set according to the composite enhanced image data set, and the corresponding enhanced image data set is determined and provided to the doctor, while maintaining the image enhancement processing effect, the data coherence is not destroyed.

[0064] Figure 1 This is a schematic diagram of an application scenario provided by this application. When examining the internal tissue structure of the human body through an endoscope, the method provided by this application is applied.

[0065] Specifically, the method of the present application is applied to any server, which communicates with a data collection terminal and a display terminal respectively. The server obtains and processes key frame data in the endoscopic image data set provided by the data collection terminal to obtain a key image data set, so as to reduce data redundancy and improve the efficiency of subsequent image enhancement processing. According to the data collection terminal multi-light source parameter set provided by the data collection terminal, the key image data set is endogenously enhanced to obtain an endogenous image data set, and on this basis, combined with the display terminal environment light source parameter set provided by the display terminal, the endogenous image data set is exogenously enhanced to obtain a composite enhanced image data set, which effectively reduces the interference of endogenous light and exogenous light on the endoscopic image, so that the endoscopic image can still maintain a good display effect under the dual influence of the environment in which the data collection terminal is located and the environment in which the display terminal is located, and the image details are guaranteed to assist doctors in medical diagnosis. According to the composite enhanced image data set, the endoscopic image data set is further compositely marked for enhancement, and the corresponding enhanced image data set is determined and provided to the doctor through the display terminal, while maintaining the image enhancement processing effect, the data continuity is not destroyed.

[0066] For specific implementation methods, please refer to the following embodiments.

[0067] Figure 2 This is a flow chart of an AI image enhancement method for a display screen provided in an embodiment of the present application. The method of this embodiment can be applied to the server in the above scenario. Figure 2As shown, the method includes:

[0068] S201: Acquire an endoscopic image data set, perform key frame processing on the endoscopic image data set, and determine a key image data set.

[0069] An endoscopic image dataset may be a collection of internal human body image data collected by an endoscope.

[0070] Key frame processing can be the process of extracting and preprocessing key frame images in an endoscopic image dataset that are highly relevant to the endoscopic diagnosis target.

[0071] The key image dataset may be a data set containing all key frame images.

[0072] Specifically, in the process of collecting images of the internal structure of the human body through an endoscope, the doctor inserts the endoscope into the patient's body through appropriate channels, such as the oral cavity, nasal cavity, anus, etc. The built-in camera of the endoscope captures the internal images of the human body in real time and outputs them through a display device for the doctor to observe. However, due to differences in observation targets for different patients, if each frame of the endoscopic image is intelligently enhanced, it is easy to cause data redundancy and a decrease in computing efficiency. For example, in the process of the endoscope passing through a predetermined channel to reach the observation target area, the image enhancement demand along the path is relatively low. Through image processing technology, key frames of the endoscopic image data set are extracted and preprocessed to obtain a key image data set, which can provide targeted data for the subsequent image enhancement process, effectively reduce data redundancy, and improve the efficiency of subsequent image enhancement to ensure the real-time nature of image enhancement, which is particularly important in the medical diagnosis process.

[0073] S202: Acquire a multi-light source parameter set of a data collection terminal, and perform endogenous enhancement processing on a key image data set based on the multi-light source parameter set of the data collection terminal to determine an endogenous image data set.

[0074] The data collection terminal may be an endoscope with multiple light sources. The multiple light source parameter set of the data collection terminal may be a set of lighting parameters of the multiple light sources set on the endoscope inside the human body. The multiple light source parameters of the data collection terminal may be provided by the data collection terminal.

[0075] Endogenous enhancement processing can be a process of processing image interference caused by multiple light sources of an endoscope located inside the human body to enhance the display effect of the endoscopic image.

[0076] The endogenous image data set may be a data set including all adjusted key frame images obtained after processing the image interference caused by multiple light sources of the endoscope inside the human body.

[0077] Specifically, in order to ensure the visibility of images collected by the endoscope inside the human body and the visibility of human tissues, multiple light sources are usually set on the endoscope to ensure the lighting effect and adapt to the needs of different observation targets. However, multiple light sources also bring more dynamic light interference to the endoscopic image. These dynamic light interferences may cover up important human tissue details and structures, which will have a negative impact on medical diagnosis. Therefore, after extracting the key image data set, mathematical analysis is used to automatically adjust each key image in the key image data set in batches according to the various lighting parameters in the multi-light source parameter set of the data collection terminal to obtain an endogenous image data set to enhance the display effect of the endoscopic image under the interference of multiple light sources of the endoscope, and at the same time provide scientific data support for subsequent further image enhancement processing.

[0078] S203: Acquire a display terminal environment light source parameter set, and perform exogenous enhancement processing on the endogenous image data set based on the display terminal environment light source parameter set to determine a composite enhanced image data set.

[0079] The display terminal may be an ultra-high definition display screen, and the display terminal environment light source parameter set may be a collection of light parameters in the environment where the display terminal is located. The display terminal environment light source parameter set may be obtained through a light sensor in the display terminal.

[0080] Exogenous enhancement processing may be a process of processing interference caused by ambient light outside the human body on the display effect of the endoscopic image, so as to enhance the display effect of the endoscopic image on the display terminal.

[0081] The composite enhanced image data set may be a data set including all adjusted endogenous images obtained after processing the interference caused by ambient light outside the human body on the display effect of the endoscopic image.

[0082] Specifically, in addition to the interference from multiple light sources of the endoscope device itself, the light in the environment where the display terminal is located will also directly affect the display effect of the endoscopic image in the real terminal, resulting in that even if the image data from the endoscope device is endogenously enhanced, the doctor is still difficult to obtain accurate image information from the display terminal due to the interference of exogenous ambient light. Through mathematical analysis, each endogenous image in the endogenous image data set is automatically adjusted in batches according to the various ambient light data of the display terminal ambient light source parameter set, and a composite enhanced image data set is obtained to enhance the display effect of the endogenous image that has completed endogenous enhancement in the display terminal under the interference of ambient light.

[0083] S204: Perform compound marking enhancement processing on the endoscopic image data set according to the compound enhanced image data set, and determine and output the enhanced image data set.

[0084] The composite marking enhancement processing may be a process of marking corresponding positions of a series of enhanced human tissue structures in the composite enhanced image data set in the endoscopic image data set.

[0085] The enhanced image data set may be an image data set obtained by performing continuous frame splicing processing on all images that have completed the composite mark enhancement processing.

[0086] Specifically, after obtaining a composite enhanced image data set, if the composite enhanced image data set is directly output, it is easy to cause data continuity to be destroyed, thereby causing the loss of details of dynamic changes in human tissue structure. According to the composite enhanced image data set, the corresponding enhanced target human tissue structure in all frame images of the endoscopic image data set is marked, and the image data is regenerated through further frame image stitching processing. While maintaining the image enhancement processing effect, the data continuity is not destroyed. After obtaining the corresponding enhanced image data set, the enhanced image data is output through the display terminal for doctors to make medical diagnosis reference.

[0087] Through this solution, key frame data in an endoscopic image data set is acquired and processed to obtain a key image data set, so as to reduce data redundancy and improve the efficiency of subsequent image enhancement processing. According to the multi-light source parameter set of the data collection terminal, the key image data set is endogenously enhanced to obtain an endogenous image data set. On this basis, combined with the display terminal environmental light source parameter set, the endogenous image data set is exogenously enhanced to obtain a composite enhanced image data set, which effectively reduces the interference of endogenous and exogenous light on the endoscopic image, so that the endoscopic image can still maintain a good display effect under the dual influence of the environment of the data collection terminal and the environment of the display terminal, and the image details are guaranteed to assist doctors in medical diagnosis. According to the composite enhanced image data set, the endoscopic image data set is further compositely marked for enhancement, and the corresponding enhanced image data set is determined and provided to the doctor, while maintaining the image enhancement processing effect, the data continuity is not destroyed.

[0088] In some embodiments, frame extraction processing is performed on the endoscopic image data set to determine the endoscopic image data set; according to preset preprocessing rules, denoising and normalization processing is performed on the endoscopic image data set to determine the normalized image data set; based on the normalized image data set, lesion area and contour extraction is performed on each endoscopic image to determine the lesion estimated area and lesion contour features corresponding to each endoscopic image; based on the lesion estimated area and lesion contour features, key frame processing is performed on the endoscopic image data set to determine a number of key frame images, and thereby construct a key image data set.

[0089] The frame extraction process may be a process of extracting each frame of the endoscopic image in the endoscopic image data set. The frame extraction process may be implemented by a computer vision processing library, such as OpenCV (Open Source Computer Vision Library).

[0090] The endoscopic image dataset may be a data set containing each frame of endoscopic image in the endoscopic image dataset.

[0091] The denoising and normalization processing may be a process of removing basic noise in the endoscopic image and performing normalization processing on different endoscopic images.

[0092] The preset preprocessing rule may be a preset processing rule for denoising and normalizing the endoscopic image, and the denoising rule in the preset preprocessing rule may adopt a Gaussian filtering algorithm.

[0093] The normalized image data set may be a data set of endoscopic images after all denoising and normalization are completed.

[0094] The estimated lesion area may be an estimated value of the area of ​​the lesion region corresponding to each endoscopic image.

[0095] The lesion contour feature may be a lesion area contour feature presented in each endoscopic image.

[0096] Specifically, during the process of examining the inside of the human body through an endoscope, since the position and angle of the endoscope are in dynamic change, the judgment of the key frame image in the endoscopic image data set cannot be based solely on whether a lesion area appears in the current endoscopic image. Instead, on this basis, the lesion contour and area in the next frame of the endoscopic image are compared with the current endoscopic image. If there is no significant change, it means that the position and angle of the endoscope have not changed significantly in the above two frames of images, and no more lesion information has been released in the endoscopic image. At this time, there is no need to determine the above two frames of endoscopic images as key frame images, only the current frame image needs to be determined as the key frame image. This can significantly reduce the amount of data calculation. Conversely, if the lesion area and contour change significantly in two consecutive frames of endoscopic images, both frames of endoscopic images need to be determined as key frame images to ensure that the lesion details are not lost.

[0097] For the above processing process, the endoscopic image dataset is subjected to frame extraction processing through the computer vision processing library to determine the endoscopic image dataset, and according to the preset preprocessing rules, all endoscopic image data in the endoscopic image dataset are subjected to automatic batch denoising and normalization processing to obtain a normalized image dataset. On this basis, the principal component analysis algorithm and contour extraction algorithm in the image processing technology are used to extract the estimated lesion area and lesion contour features corresponding to each endoscopic image respectively, which are used as the basis for key frame extraction. Through mathematical analysis, the key frame images of the endoscopic image dataset are batch extracted to construct a key image dataset.

[0098] Through this scheme, frame extraction processing is performed on the endoscopic image data set to determine the endoscopic image data set, and according to the preset preprocessing rules, denoising and normalization processing is performed on the endoscopic image data set to determine the normalized image data set, so as to reduce noise interference and unify the data dimension, and improve the accuracy and efficiency of subsequent key frame processing. On this basis, the lesion estimated area and lesion contour features corresponding to each endoscopic image are extracted, and key frame processing is performed on the endoscopic image data set accordingly, so as to construct a key image data set, so that the key frame image conforms to the dynamic change characteristics of the endoscope, and significantly reduces data redundancy without losing image details.

[0099] In some embodiments, based on the estimated lesion area and lesion contour features, key frame processing is performed on the endoscopic image dataset to determine a number of key frame images, and a key image dataset is constructed based on them, which is specifically the following formula (1):

[0100] K={I t ||A t -A t-1 |>ΔA or H(C t ,C t-1 )>D th for all t}(1)

[0101] Among them, K is the key image dataset, I t is the endoscopic image corresponding to the current frame time t, A t is the estimated area of ​​the lesion in the endoscopic image at the current frame time t, A t-1 is the estimated area of ​​the lesion corresponding to the endoscopic image at the previous frame time t-1, ΔA is the preset area change threshold, C t is the corresponding lesion contour feature in the endoscopic image at the current frame time t, C t-1 is the lesion contour feature corresponding to the endoscopic image at the previous frame time t-1, H() is the Hausdorff distance function, D th is the preset feature change threshold.

[0102] The preset area change threshold may be an area threshold used to characterize whether the change in the lesion area is significant. The preset area change threshold may be obtained by analyzing and fitting the historical endoscopy dataset using a deep learning model.

[0103] The preset feature change threshold may be an area threshold used to characterize whether the change in the lesion contour is significant. The preset feature change threshold may be obtained by analyzing and fitting the historical endoscopy dataset using a deep learning model.

[0104] Specifically, by using |A in formula (1) t -A t-1 |>ΔA indicates that the area change of the lesion between the endoscopic images corresponding to the previous and next frames exceeds the preset area change threshold, which can be regarded as a signal for further release of lesion area information, and the current frame is selected as the key frame; through H(C t ,C t-1 )>D th It means that the change of the lesion contour between the endoscopic images corresponding to the previous and next frames exceeds the preset feature change threshold, which can be regarded as a signal for further release of the lesion contour information. The current frame is selected as the key frame. The Hausdorff distance is used as the judgment basis here, which can provide a more comprehensive perspective for the comparison of the lesion contour. It is suitable for the comparison process of the "lesion contour", which has a highly irregular feature. Even if the local contour of the lesion changes slightly, the Hausdorff distance can still reflect the significant change of the overall contour, so that the lesion contour comparison process has good robustness.

[0105] Through this scheme, mathematical analysis methods are used to perform key frame processing on the endoscopic image dataset according to the estimated lesion area and lesion contour characteristics, and several key frame images are extracted. Based on this, a key image dataset is constructed, so that the key frame image determination process is highly matched with the changes in the estimated lesion area and lesion contour characteristics under the dynamically changing endoscopic perspective, thereby improving the accuracy of the key frame images and the comprehensiveness and robustness of the key frame extraction process.

[0106] In some embodiments, based on the multi-light source intensity information, the light source intensity of different light source illumination areas in each key image is extracted; the key image data set is analyzed to extract the light intensity value corresponding to each pixel in each key image; based on the light source illumination area to which the pixel belongs, the light intensity normalization processing is performed on the different light source illumination areas in each key image according to the light source intensity and the light intensity value, and the light intensity normalization processing result is determined; based on the multi-light source angle information, the incident angle and reflection angle of the different light sources in each key image in the corresponding light source illumination area are determined; based on the light source intensity, the reflected light intensity value of the different light source illumination areas in each key image is determined according to the incident angle and the reflection angle, and the mean reflected light intensity is determined accordingly; based on the mean reflected light intensity, the light source intensity and the light intensity normalization processing result, the light intensity in the pixel point in each key image is adjusted to determine the adjusted light intensity value; based on the adjusted light intensity value, each key image is endogenously enhanced, and an endogenous image data set is constructed based on all processed key images.

[0107] The multi-light source parameter set of the data collection terminal includes multi-light source intensity information and multi-light source angle information.

[0108] The multi-light source intensity information may be the light source intensities corresponding to the multiple lighting components carried by the endoscope in different images.

[0109] The multi-light source angle information may be the light source irradiation angles corresponding to the multiple lighting components carried by the endoscope in different images.

[0110] The light source illumination area may be the illumination area that multiple illumination components carried by the endoscope are responsible for in different images.

[0111] The light source intensity may be the light source intensity of each light source irradiation area in each key image that is illuminated by the corresponding lighting component.

[0112] The light intensity value may be a pixel brightness value corresponding to each pixel point in each key image.

[0113] The incident angle may be an illumination angle of the light source of each illumination assembly within an illumination area of ​​the corresponding light source in each key image.

[0114] The reflection angle may be a light reflection angle in an area illuminated by a corresponding light source in each key image.

[0115] The reflected light intensity value may be the light reflection intensity of the area illuminated by the corresponding light source in each key image.

[0116] The adjusted light intensity value may be a value required to adjust the brightness of different pixel values ​​in each image in order to reduce the interference of multiple light sources in the current endoscope.

[0117] Specifically, in the process of inspecting the inside of the human body using an endoscopic device equipped with multiple light sources, the brightness and angle of the photos of the multiple lighting components on the endoscope need to be changed accordingly due to the changes in the position and angle of the endoscope, in order to ensure the basic quality of the endoscopic image. This results in differences in the light source intensity and illumination angle in the illumination areas responsible for the different lighting components on the endoscope in different key images, which in turn affects the presentation of light and dark details of human tissues in different areas of the key image. By recording the multi-light source intensity information of the changes of different lighting components of the endoscope during the image collection process, the light source intensity of the different light source illumination areas in each key image is extracted. At the same time, through the computer vision processing library, the brightness value corresponding to each pixel point in each key image in the key image data set is used as its corresponding light intensity value to clarify the brightness distribution of each illumination area in the image, and through mathematical analysis, the different light sources in the image are analyzed according to the light source intensity and light intensity value. The light intensity of the source illumination area is normalized to ensure that the light intensity of each light source illumination area is adjusted in the same dimension in the future. By extracting the incident angle and the reflection angle, the multi-light source illumination effect of the endoscope can be modeled more accurately, reflecting the reflection and shadow effects of the endoscope light at different angles, making the analysis of the image brightness more accurate. The reflected light intensity value of each light source illumination area is quantified by mathematical analysis to reflect the real reflection characteristics of human tissue under the illumination conditions of different light source illumination areas. The mean reflected light intensity is obtained based on this, which is used as the basis for reducing local brightness fluctuations in the future. Through mathematical analysis, according to the mean reflected light intensity, light source intensity and light intensity normalization processing results, the adjusted light intensity value required under multi-light source interference is reduced for each pixel in the key image, and the brightness value of the pixel in each key image is adjusted accordingly, so as to achieve endogenous enhancement processing of each key image, and construct an endogenous image data set based on the key image after processing.

[0118] Through this scheme, the light source intensity of different light source illumination areas in each key image is extracted, and the brightness value corresponding to each pixel point in each key image in the key image data set is extracted as its corresponding light intensity value. By extracting the incident angle and the reflection angle, the multi-light source illumination effect of the endoscope can be modeled more accurately, reflecting the reflection and shadow effects of the endoscope light at different angles. By analyzing the reflected light intensity value of each light source illumination area, the real reflection characteristics of human tissue under the illumination conditions of different light source illumination areas are reflected. The average reflected light intensity obtained on this basis is used as the basis for subsequent reduction of local brightness fluctuations. Then, according to the reflected light intensity average, light source intensity and light intensity normalization processing results, the adjusted light intensity value required under multi-light source interference is reduced for each pixel point in the key image, and the brightness value of the pixel point in each key image is adjusted accordingly to achieve endogenous enhancement processing of each key image.

[0119] In some embodiments, based on the light source intensity, according to the incident angle and the reflection angle, the reflected light intensity value of the different light source illumination areas in each key image is determined, specifically as follows:

[0120] R j =L j ·(k d ·cos(θ j )+k s ·(max(cos(φ j ),0)) G )(2)

[0121] Among them, R j is the reflected light intensity value corresponding to the j-th light source illumination area in the current key image, L j is the light source intensity corresponding to the jth light source illumination area in the current key image, k d is the preset tissue diffuse reflectance, θ j is the incident angle corresponding to the jth light source illumination area in the current key image, k s is the preset tissue mirror reflection coefficient, φ j is the reflection angle corresponding to the area illuminated by the jth light source in the current key image, and G is the preset gloss factor.

[0122] The preset tissue diffuse reflectance coefficient may be a coefficient used to characterize the ability of different tissues inside the human body to evenly scatter and reflect light in all directions. The preset tissue diffuse reflectance coefficient may be obtained by analyzing and fitting historical endoscopic image data.

[0123] The preset tissue specular reflection coefficient may be a coefficient that characterizes the ability of different tissues inside the human body to reflect light in a specular reflection manner. The preset tissue specular reflection coefficient may be obtained by analyzing and fitting historical endoscopic image data.

[0124] The preset gloss factor may be a parameter used to characterize the smoothness of the surfaces of different tissues inside the human body. The preset gloss factor may be obtained by analyzing and fitting historical endoscopic image data.

[0125] Specifically, the surface characteristics of different types of tissues in the human body are very different, and they are not completely ideal diffuse reflection or specular reflection materials. The internal tissues of the human body usually show a combination of diffuse reflection and specular reflection characteristics to light. d ·cos(θ j ) describes the diffuse reflection characteristics of human tissue in the current key image, and cos(θ j ) represents the effect of the incident angle of the light source on the intensity of diffuse reflected light. The smaller the incident angle, the stronger the reflected light intensity. It describes the scattering of light by human tissue. At the same time, through ks ·(max(cos(φ j ),0)) G Describes the specular reflection characteristics of human tissue, reflects the reflection effect produced by the interaction between light and the smooth surface of human tissue, and quantifies the reflected light intensity value based on the corresponding light source intensity according to the above-mentioned diffuse reflection characteristics and specular reflection characteristics to improve the scientific nature of the reflected light intensity value.

[0126] Through this scheme, mathematical analysis methods are used to scientifically quantify the reflected light intensity values ​​of different light source illumination areas in each key image based on the light source intensity, according to the incident angle and reflection angle, so that the reflected light intensity values ​​are consistent with the diffuse reflection characteristics and mirror reflection characteristics of human tissue to light in the current key image, thereby improving the accuracy and comprehensiveness of the reflected light intensity values.

[0127] In some embodiments, the light intensity in each pixel in each key image is adjusted according to the reflected light intensity mean, the light source intensity and the light intensity normalization processing result, and the adjusted light intensity value is determined, which is specifically the following formula (3):

[0128]

[0129] Among them, I t ′ (x, y) is the adjusted light intensity value of the pixel at the coordinate point (x, y) in the current key image, I t (x, y) The light intensity value of the pixel at the coordinate point (x, y) in the current key image, L j is the light source intensity corresponding to the j-th light source illumination area in the current key image, is the result of light intensity normalization processing, n is the total number of light source irradiation areas in the current key image, is the mean value of the reflected light intensity corresponding to the current key image.

[0130] Specifically, through formula (3) Normalize the light intensity value of each coordinate point to the corresponding light source intensity to unify the dimension, and then Evaluate the deviation relationship between the intensity of each light source and the mean value of the reflected light intensity, using The square deviations between all light source intensities and the mean of reflected light intensities are summed and averaged to reflect the overall difference between the light source intensity and the reflected light intensity. Formula (5) is used to provide a balanced adjustment based on the light source intensity difference during the processing of the endoscopic image. The adjusted light intensity value is used as the basis for adjusting the pixel brightness. This can make the light intensity of each pixel in the endoscopic image more consistent and reasonable under the influence of different light sources, thereby reducing the interference of multiple light sources and enhancing the endoscopic image display effect.

[0131] Through this scheme, mathematical analysis methods are used to scientifically quantify the adjusted light intensity value required for the pixel point according to the mean value of the reflected light intensity, the light source intensity and the light intensity normalization processing results. The balanced adjustment based on the light source intensity difference provided by the mathematical formula is used to take the obtained adjusted light intensity value as the basis for adjusting the pixel brightness. This can make the light intensity of each pixel point in the endoscopic image under the influence of different light sources more consistent and reasonable, thereby reducing the interference of multiple light sources and enhancing the endoscopic image display effect.

[0132] In some embodiments, ambient light characteristic values ​​are determined based on basic ambient light intensity, ambient light fluctuation amplitude, ambient light fluctuation frequency and ambient light phase; based on the ambient light characteristic values, color channels corresponding to pixels in each image in the endogenous image data set are corrected to determine corresponding corrected pixels; based on all corrected pixels corresponding to each image in the endogenous image data set, a composite enhanced image corresponding to each image is generated, and a composite enhanced image data set is constructed thereby.

[0133] The display terminal ambient light parameter set includes basic ambient light intensity, ambient light fluctuation amplitude, ambient light fluctuation frequency and ambient light phase.

[0134] The basic ambient light intensity may be the overall light intensity in the environment where the display terminal is located.

[0135] The ambient light fluctuation amplitude may be an average fluctuation amplitude of the light intensity in the environment where the display terminal is located within a preset monitoring period.

[0136] The ambient light fluctuation frequency may be an average fluctuation frequency of the light intensity in the environment where the display terminal is located within a preset monitoring period.

[0137] The ambient light phase may be phase information reflecting the light fluctuation period in the environment where the display terminal is located.

[0138] The ambient light characteristic value may be a quantitative value used to characterize the comprehensive influence of light in the current environment where the display terminal is located.

[0139] The corrected pixel point may be a corresponding pixel value obtained after correcting the pixel point for the purpose of reducing the influence of ambient light.

[0140] The composite enhanced image may be an enhanced endoscopic image obtained by combining endogenous enhancement with exogenous enhancement.

[0141] Specifically, in the process of evaluating the influence of ambient light on the display effect of endoscopic images of display terminals, the main starting point is the basic ambient light intensity, ambient light fluctuation amplitude, ambient light fluctuation frequency and ambient light phase in the ambient light source parameter set of the display terminal. Among them, when the basic ambient light intensity is too high, the endoscopic image will lose details due to highlight interference, and the highlight area in the endoscopic image will appear to be overexposed, with bright spots appearing, thereby covering up important static details; a larger ambient light fluctuation amplitude will cause the dynamic process of the endoscopic image to be blurred, thereby losing important dynamic details; a higher ambient light fluctuation frequency will cause the visual effect of the endoscopic image to flicker, making it difficult to show the continuity of the endoscopic image; unstable ambient light photography If the display is not properly positioned, it will cause uneven distribution of shadows and highlights on the display surface, which will have a negative effect on the doctor's observation. Through mathematical analysis, according to the basic ambient light intensity, ambient light fluctuation amplitude, ambient light fluctuation frequency and ambient light phase, a comprehensive quantitative analysis is performed to obtain the ambient light characteristic value that can reflect the comprehensive impact of ambient light. On this basis, through mathematical analysis, the color channel corresponding to each pixel in each image in the endogenous image data set is further corrected to obtain the corrected pixel value to determine the corresponding corrected pixel point. Through image reconstruction technology, according to all the corrected pixels corresponding to each image, a composite enhanced image corresponding to each image is generated, and a composite enhanced image data set is constructed based on this.

[0142] Through this scheme, the ambient light characteristic value that can reflect the comprehensive influence of the ambient light is determined according to the basic ambient light intensity, the ambient light fluctuation amplitude, the ambient light fluctuation frequency and the ambient light phase, and the color channel corresponding to each pixel in each image in the endogenous image data set is corrected to determine the corresponding corrected pixel to generate a composite enhanced image, and the composite enhanced image data set is constructed to reduce the light interference in the environment where the display is located, and further improve the display effect of the enhanced endoscopic image.

[0143] In some embodiments, based on the ambient light characteristic value, the color channel corresponding to each pixel in each image in the internal source image data set is corrected to determine the corresponding corrected pixel, which is specifically the following formula (4):

[0144] E′ t (x,y)=∑ c∈{R,G,B} (E t ′ (x,y,c)+λ·M) (4)

[0145] Among them, E′ t (x, y) is the corrected pixel point at the coordinate point (x, y) in the image corresponding to the current frame time t, c is the image color channel, c∈{R,G,B} indicates the use of RGB color channel, α is the preset channel influence coefficient, E′ t(x, y, c) is the image color channel of the pixel at the coordinate point (x, y) in the image corresponding to the current frame time t, λ presets the environmental feature influence coefficient, and M is the ambient light feature value.

[0146] The preset environmental characteristic influence coefficient may be a coefficient used to reflect the influence of ambient light on a color channel, and the preset environmental characteristic influence coefficient may be obtained by fitting display screen experimental data.

[0147] Specifically, all pixel color channels in each image are traversed by formula (4), and the influence of ambient light is mapped to the offset adjustment of the pixel color channel according to the ambient light characteristic value, and the corresponding corrected pixel is obtained to reduce the interference of ambient light on the display effect of the endoscopic image on the display terminal.

[0148] Through this scheme, mathematical analysis methods are used to describe the influence of ambient light on the color channels of pixels through mathematical language according to the ambient light characteristic values, so as to correct the color channels corresponding to each pixel in each image in the endogenous image data set, determine the corresponding corrected pixel, reduce the interference of ambient light on the display effect of the endoscopic image on the display terminal, and enhance the display effect of the endoscopic image on the display terminal.

[0149] In some embodiments, the ambient light characteristic value is determined according to the basic ambient light intensity, the ambient light fluctuation amplitude, the ambient light fluctuation frequency and the ambient light illumination phase, specifically as the following formula (5):

[0150] M=L0+ΔL·sin(ωT+φ)(5)

[0151] Among them, M is the ambient light characteristic value, L0 is the basic ambient light intensity, ΔL is the ambient light fluctuation amplitude, ω is the ambient light fluctuation frequency, T is the preset ambient light monitoring time, and φ is the ambient light phase.

[0152] The preset ambient light monitoring time may be a preset time for monitoring the ambient light through a light sensor built into the display terminal, and the preset ambient light monitoring time may be set according to specific needs.

[0153] Specifically, since the ambient light usually changes periodically, the periodic change characteristics of the ambient light are described by the sinusoidal function ΔL·sin(ωT+φ) in formula (5). On this basis, combined with the basic ambient light intensity, the ambient light characteristic value is quantified. Formula (5) can be regarded as a stable part (basic ambient light intensity) combined with a variable part (fluctuation given by the sinusoidal function) to describe the dynamic change characteristics of the ambient light.

[0154] Through this scheme, mathematical analysis is used to describe the periodic fluctuation of ambient light above the basic ambient light intensity through mathematical language according to the basic ambient light intensity, ambient light fluctuation amplitude, ambient light fluctuation frequency and ambient light phase, and the ambient light characteristic value reflecting the comprehensive influence of ambient light is accurately quantified to improve the accuracy and scientificity of the ambient light characteristic value.

[0155] In some embodiments, based on the composite enhanced image data set, the endoscopic image data set is analyzed to determine a number of similar frame images corresponding to each composite enhanced image in the endoscopic image data set; based on the corrected pixel points of each composite enhanced image in the composite enhanced image data set, composite enhancement processing is performed on a number of similar frame images of each composite enhanced image to determine an enhanced similar frame image set; based on the composite enhanced image data set and the enhanced similar frame image set, an enhanced image data set is constructed and output.

[0156] The similar frame image may be a frame image in the endoscopic image data set having features highly similar to those of the composite enhanced image.

[0157] The enhanced similar frame image set may be an image data set obtained by enhancing the corresponding similar frame images according to the enhanced features in the composite enhanced image.

[0158] Specifically, the principal component analysis algorithm is used to extract the feature vectors corresponding to each image in the composite enhanced image data set and the endoscopic image data set, and the feature vectors of each composite enhanced image are compared with the feature vectors of each endoscopic image through a feature similarity algorithm, such as a cosine similarity algorithm, to extract a number of similar frame images corresponding to each composite enhanced image in the endoscopic image data set, and according to the corrected pixel points of each composite enhanced image in the composite enhanced image data set, the same pixel correction processing is performed on the similar frame images corresponding to the composite enhanced image to achieve composite enhancement processing on the similar frame images, and an enhanced similar frame image set is obtained, and an enhanced image data set is generated according to the frame sequences corresponding to all images in the composite enhanced image data set and the enhanced similar frame image set, and the enhanced image data set is output through a display terminal.

[0159] Through this scheme, a number of similar frame images corresponding to each composite enhanced image in the composite enhanced image data set in the endoscopic image data set are extracted, and the similar frame images are subjected to composite enhancement processing according to the corrected pixel points corresponding to the composite enhanced images to obtain an enhanced similar frame image set. On this basis, an enhanced image data set is generated and output to maintain the continuity of the enhanced features in the corresponding endoscopic images, thereby providing a positive auxiliary effect for doctors' medical diagnosis.

Claims

1. An AI image enhancement method for a display screen, characterized in that: include: Acquire an endoscopic image data set, perform key frame processing on the endoscopic image data set, and determine a key image data set; Acquire a multi-light source parameter set of a data collection terminal, and perform endogenous enhancement processing on the key image data set based on the multi-light source parameter set of the data collection terminal to determine an endogenous image data set; Acquire a display terminal environment light source parameter set, and based on the display terminal environment light source parameter set, perform exogenous enhancement processing on the endogenous image data set to determine a composite enhanced image data set; According to the composite enhanced image data set, composite marking enhancement processing is performed on the endoscopic image data set to determine and output an enhanced image data set.

2. The method according to claim 1, characterized in that The step of performing key frame processing on the endoscopic image data set to determine a key image data set includes: Performing frame extraction processing on the endoscopic image data set to determine an endoscopic image data set; According to a preset preprocessing rule, the endoscopic image data set is subjected to denoising and normalization processing to determine a normalized image data set; Based on the normalized image data set, extract the lesion area and contour of each endoscopic image to determine the lesion estimated area and lesion contour features corresponding to each endoscopic image; Based on the estimated lesion area and the lesion contour features, key frame processing is performed on the endoscopic image data set to determine a number of key frame images, and the key image data set is constructed based on these.

3. The method according to claim 2, characterized in that Based on the estimated lesion area and the lesion contour features, the endoscopic image data set is subjected to key frame processing to determine a number of key frame images, and the key image data set is constructed based on the key frame images, which is specifically the following formula: K={I t ||A t -A t-1 |>ΔA or H(C t ,C t-1 )>D th for all t}; Wherein, K is the key image dataset, I t is the endoscopic image corresponding to the current frame time t, A t is the estimated area of ​​the lesion corresponding to the endoscopic image at the current frame time t, A t-1 is the estimated area of ​​the lesion corresponding to the endoscopic image at the previous frame time t-1, ΔA is the preset area change threshold, C t is the lesion contour feature corresponding to the endoscopic image at the current frame time t, C t-1 is the lesion contour feature corresponding to the endoscopic image at the previous frame time t-1, H() is the Hausdorff distance function, D th is the preset feature change threshold.

4. The method according to claim 3, characterized in that The multi-light source parameter set of the data collection terminal includes multi-light source intensity information and multi-light source angle information. The performing of endogenous enhancement processing on the key image data set based on the multi-light source parameter set of the data collection terminal to determine the endogenous image data set includes: Extracting the light source intensity of different light source illumination areas in each key image according to the multiple light source intensity information; Analyze the key image data set to extract the light intensity value corresponding to each pixel in each key image; Based on the light source illumination area to which the pixel point belongs, and according to the light source intensity and the light intensity value, light intensity normalization processing is performed on different light source illumination areas in each key image to determine a light intensity normalization processing result; Determine the incident angle and reflection angle of different light sources in each key image within the corresponding light source illumination area according to the multi-light source angle information; Based on the light source intensity, according to the incident angle and the reflection angle, determine the reflected light intensity values ​​of different light source illumination areas in each key image, and thereby determine the average reflected light intensity; According to the reflected light intensity mean, the light source intensity and the light intensity normalization processing result, the light intensity in each pixel in the key image is adjusted to determine the adjusted light intensity value; According to the adjusted light intensity value, each key image is subjected to endogenous enhancement processing, and an endogenous image data set is constructed according to all processed key images.

5. The method according to claim 4, characterized in that Based on the light source intensity, the reflected light intensity values ​​of different light source illumination areas in each key image are determined according to the incident angle and the reflection angle, specifically as follows: R j =L j ·(k d ·cos(θ j )+k s ·(max(cos(φ j ),0)) G ); Among them, R j is the reflected light intensity value corresponding to the j-th light source illumination area in the current key image, L j is the light source intensity corresponding to the j-th light source illumination area in the current key image, k d is the preset tissue diffuse reflectance, θ j is the incident angle corresponding to the j-th light source illumination area in the current key image, k s is the preset tissue mirror reflection coefficient, φ j is the reflection angle corresponding to the j-th light source illumination area in the current key image, and G is the preset gloss factor.

6. The method according to claim 5, characterized in that According to the reflected light intensity mean, the light source intensity and the light intensity normalization processing result, the light intensity in each pixel in the key image is adjusted to determine the adjusted light intensity value, which is specifically the following formula: Among them, I t ′ (x, y) is the adjusted light intensity value of the pixel at the coordinate point (x, y) in the current key image, I t (x, y) is the light intensity value of the pixel at the coordinate point (x, y) in the current key image, L j is the light source intensity corresponding to the j-th light source illumination area in the current key image, is the result of the light intensity normalization processing, n is the total number of light source irradiation areas in the current key image, is the mean value of the reflected light intensity corresponding to the current key image.

7. The method according to claim 5, characterized in that The display terminal ambient light parameter set includes basic ambient light intensity, ambient light fluctuation amplitude, ambient light fluctuation frequency and ambient light illumination phase. The exogenous enhancement processing is performed on the endogenous image data set based on the display terminal ambient light parameter set to determine a composite enhanced image data set, including: Determining an ambient light characteristic value according to the basic ambient light intensity, the ambient light fluctuation amplitude, the ambient light fluctuation frequency and the ambient light illumination phase; Based on the ambient light characteristic value, correct the color channel corresponding to each pixel in each image in the internal source image data set to determine the corresponding corrected pixel; According to all the corrected pixel points corresponding to each image in the internal source image data set, a composite enhanced image corresponding to each image is generated, and the composite enhanced image data set is constructed based on this.

8. The method according to claim 7, characterized in that Based on the ambient light characteristic value, the color channel corresponding to each pixel in each image in the internal source image data set is corrected to determine the corresponding corrected pixel, which is specifically the following formula: E′ t (x,y)=∑ c∈{R,G,B} (E′ t (x,y,c)+λ·M); Among them, E′ t (x, y) is the corrected pixel point at the coordinate point (x, y) in the image corresponding to the current frame time t, c is the image color channel, c∈{R,G,B} indicates the use of RGB color channel, α is the preset channel influence coefficient, E′ t (x, y, c) is the image color channel of the pixel at the coordinate point (x, y) in the image corresponding to the current frame time t, λ is the preset environmental feature influence coefficient, and M is the ambient light feature value.

9. The method according to claim 8, characterized in that The determining of the ambient light characteristic value according to the basic ambient light intensity, the ambient light fluctuation amplitude, the ambient light fluctuation frequency and the ambient light illumination phase is specifically the following formula: M = L0 + ΔL·sin(ωT + φ); Among them, M is the ambient light characteristic value, L0 is the basic ambient light intensity, ΔL is the ambient light fluctuation amplitude, ω is the ambient light fluctuation frequency, T is the preset ambient light monitoring time, and φ is the ambient light phase.

10. The method according to claim 9, characterized in that The step of performing composite marking enhancement processing on the endoscopic image dataset according to the composite enhanced image dataset, and determining and outputting an enhanced image dataset comprises: Based on the composite enhanced image data set, analyzing the endoscopic image data set, determining a number of similar frame images corresponding to each composite enhanced image in the endoscopic image data set; According to the corrected pixel points of each composite enhanced image of the composite enhanced image data set, composite enhancement processing is performed on a plurality of similar frame images of each composite enhanced image to determine an enhanced similar frame image set; The enhanced image data set is constructed and outputted according to the composite enhanced image data set and the enhanced similar frame image set.

Images