High-definition digital imaging method for otolaryngology endoscopy

Abstract

The present invention relates to the field of endoscopic imaging technology, and in particular to a high-definition digital imaging method suitable for ENT endoscopes. The present invention analyzes the pixel value distribution between different super-pixel real-time blocks to obtain the dark area structure activity coefficient of each super-pixel real-time block; obtains the overall pixel correlation based on the pixel similarity between each super-pixel real-time block and the different super-pixel real-time blocks within the corresponding neighborhood in all corresponding areas of the endoscopic image; obtains the shadow interference coefficient of each super-pixel real-time block in combination with the fluctuation degree distribution of the pixel values ​​between each super-pixel real-time block and the different super-pixel real-time blocks within the corresponding neighborhood; and then obtains the detail fuzzy coefficient of each super-pixel real-time block, corrects the initial grid, obtains the corrected grid, and obtains the enhanced endoscopic image. The present invention improves the imaging quality by obtaining a suitable grid during the endoscopic image enhancement process.

Description

Technical Field

[0001] The present invention relates to the technical field of endoscopic imaging, and in particular to a high-definition digital imaging method suitable for ENT endoscopes. Background Art

[0002] With the development of medical imaging technology, the emergence of high-definition digital endoscopy technology has significantly improved image quality and facilitated the observation of lesions. However, due to the complex internal structure of tissues and mutual obstruction, and the relatively single light source of the endoscope, the image is affected by different lighting conditions, resulting in dark images and limited imaging clarity and details. Therefore, it is necessary to enhance the endoscopic image.

[0003] In the prior art, an adaptive histogram equalization (CLAHE) algorithm is applied to endoscopic images based on an empirically set grid, thereby enhancing contrast and improving image clarity. However, since the influence of light source light is not taken into account, different regional structures have different detail characteristics. The empirically set grid cannot adapt to the contrast enhancement requirements of dynamically changing areas. If the grid is small, it may lead to excessive local contrast enhancement, resulting in obvious blocking effects. If the grid is large, it may lead to insufficient local contrast enhancement, resulting in loss of image details and poor imaging effect. Summary of the Invention

[0004] In order to solve the technical problem that the empirically set grid size cannot adapt to the contrast enhancement requirements of dynamically changing areas, resulting in poor imaging effects, the present invention aims to provide a high-definition digital imaging method suitable for ENT endoscopes. The technical solutions adopted are as follows:

[0005] The present invention proposes a high-definition digital imaging method suitable for ENT endoscopes, the method comprising:

[0006] Acquire endoscopic images of the observation site in time sequence;

[0007] Obtain multiple superpixel real-time blocks of a real-time endoscopic image; obtain a dark area probability parameter for each superpixel real-time block based on differences in pixel value distributions between different superpixel real-time blocks; obtain a dark area structure activity coefficient for each superpixel real-time block based on pixel value distributions of corresponding areas between different adjacent endoscopic images of each superpixel real-time block and the dark area probability parameter;

[0008] Based on the pixel similarity between each superpixel real-time block and different superpixel real-time blocks within the corresponding neighborhood in all corresponding areas of the endoscopic image, the overall pixel correlation of each superpixel real-time block is obtained; based on the fluctuation degree distribution of pixel values ​​between each superpixel real-time block and different superpixel real-time blocks within the corresponding neighborhood, and the overall pixel correlation, the shadow interference coefficient of each superpixel real-time block is obtained;

[0009] According to the dark area structure activity coefficient and shadow interference coefficient of each superpixel real-time block, the detail fuzzy coefficient of each superpixel real-time block is obtained; according to the detail fuzzy coefficient of each superpixel real-time block, the initial grid is corrected to obtain a corrected grid;

[0010] The CLAHE algorithm is performed on each superpixel in the endoscopic image in real time according to the corrected grid to obtain the enhanced endoscopic image.

[0011] Furthermore, the method for obtaining the dark area probability parameter includes:

[0012] Obtain the mean of all pixel values ​​in each superpixel real-time block as the block pixel level; obtain the dark area probability coefficient based on the difference distribution of block pixel levels between different superpixel real-time blocks;

[0013] According to the difference in block pixel levels between each superpixel real-time block and different superpixel real-time blocks in the corresponding neighborhood, as well as the dark area possibility coefficients of different superpixel real-time blocks in the neighborhood, the dark area probability parameter of each superpixel real-time block is obtained. The difference in block pixel levels and the dark area possibility coefficients are both positively correlated with the dark area probability parameter.

[0014] Furthermore, the method for obtaining the possible coefficients of the dark area includes:

[0015] Obtain the pixel-level differences between different superpixel real-time blocks and normalize them as the block difference index;

[0016] If the block difference index between the super-pixel real-time blocks is less than the preset difference threshold, the corresponding super-pixel real-time blocks are regarded as the same type of blocks;

[0017] The ratio of the number of similar blocks corresponding to each superpixel real-time block to the number of all superpixel real-time blocks is obtained, and negative correlation mapping is performed as the possible dark area coefficient of each superpixel real-time block.

[0018] Furthermore, the method for obtaining the dark area structure activity coefficient includes:

[0019] A corner detection method is used to obtain the corner points of each superpixel real-time block in the corresponding area of ​​each endoscopic image; the difference in the number of corresponding corner points of each superpixel real-time block between different adjacent endoscopic images is obtained as a first difference feature;

[0020] Obtaining the mean square error of the set composed of all pixel values ​​in the corresponding area between different adjacent endoscopic images of each superpixel real-time block as the second difference feature;

[0021] According to the first difference feature, the second difference feature and the dark area probability parameter corresponding to each superpixel real-time block, the dark area structure activity coefficient of each superpixel real-time block is obtained, and the first difference feature, the second difference feature and the dark area probability parameter are all positively correlated with the dark area structure activity coefficient.

[0022] Furthermore, the method for obtaining the overall pixel correlation includes:

[0023] For each superpixel real-time block and each superpixel real-time block within the corresponding neighborhood, the correlation coefficient of the set of block pixel levels in the corresponding area of ​​all endoscopic images between the superpixel real-time blocks is obtained as the pixel local correlation;

[0024] The average of the local correlations of the corresponding pixels between each superpixel real-time block and all superpixel real-time blocks in the corresponding neighborhood is obtained as the overall pixel correlation of each superpixel real-time block.

[0025] Furthermore, the method for obtaining the shadow interference coefficient includes:

[0026] Obtain the fluctuation degree of all pixel values ​​in each superpixel real-time block as the block pixel fluctuation feature;

[0027] Obtain the mean of the block pixel fluctuation characteristics of all superpixel real-time blocks within the neighborhood as the neighborhood pixel fluctuation feature;

[0028] The difference between the block pixel fluctuation characteristics and the neighborhood pixel fluctuation characteristics of each superpixel real-time block is obtained, and a negative correlation mapping is performed as the distribution stability; the corresponding distribution stability is weighted according to the overall pixel correlation of each superpixel real-time block to obtain the shadow interference coefficient of each superpixel real-time block.

[0029] Furthermore, the method for obtaining the detail fuzzy coefficient includes:

[0030] The dark area structure activity coefficient and shadow interference coefficient of each superpixel real-time block are fused to obtain the detail fuzzy coefficient of each superpixel real-time block.

[0031] Furthermore, the method for obtaining the modified grid includes:

[0032] Get the initial size of the initial grid;

[0033] Negative correlation mapping is performed on the detail fuzzy coefficient of each superpixel real-time block, and the product of the negative correlation mapping result and the initial size is obtained as the correction size to obtain the correction grid.

[0034] Furthermore, the preset difference threshold is 0.3.

[0035] Furthermore, the correlation coefficient is a Pearson correlation coefficient.

[0036] The present invention has the following beneficial effects:

[0037] The present invention obtains multiple super-pixel real-time blocks of real-time endoscopic images, which helps to more conveniently process and analyze local features of the image and reduce computational complexity; according to the difference in pixel value distribution between different super-pixel real-time blocks, the dark area probability parameter of each super-pixel real-time block is obtained, and by evaluating the distribution difference of pixel values ​​within the super-pixel real-time blocks, the dark area in the image can be identified; according to the pixel value distribution of the corresponding area between different adjacent endoscopic images of each super-pixel real-time block and the dark area probability parameter, the dark area structure activity coefficient of each super-pixel real-time block is obtained, and the dynamic change of the dark area is evaluated; according to the pixel similarity between each super-pixel real-time block and different super-pixel real-time blocks within the corresponding neighborhood range in all corresponding areas of the endoscopic images, the overall pixel correlation of each super-pixel real-time block is obtained, and the super-pixel real-time block is evaluated. The similarity between a block and other blocks within its neighborhood reflects the degree to which the superpixel real-time block is affected by other blocks; based on the distribution of pixel value fluctuations between each superpixel real-time block and different superpixel real-time blocks within the corresponding neighborhood, as well as the overall pixel correlation, the shadow interference coefficient of each superpixel real-time block is obtained; based on the dark area structure activity coefficient and shadow interference coefficient of each superpixel real-time block, the detail fuzziness coefficient of each superpixel real-time block is obtained, which reflects the degree of fuzziness of different areas in the image and helps to identify detail areas that need to be enhanced; based on the detail fuzziness coefficient of each superpixel real-time block, the initial grid is corrected to obtain a corrected grid so that it better adapts to detail changes in the image; based on the corrected grid, the CLAHE algorithm is performed on each superpixel real-time block in the endoscopic image to obtain an enhanced endoscopic image. The present invention improves imaging quality by obtaining a suitable grid for endoscopic image enhancement processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0039] Figure 1 A flowchart of a high-definition digital imaging method applicable to an ENT endoscope provided in one embodiment of the present invention;

[0040] Figure 2 A flow chart of a method for obtaining a dark area structure activity coefficient provided by one embodiment of the present invention;

[0041] Figure 3 A flow chart of a method for obtaining a shadow interference coefficient provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0042] To further illustrate the technical means and effectiveness of the present invention in achieving its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effectiveness of a high-definition digital imaging method for ENT endoscopes proposed in accordance with the present invention. In the following description, references to different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0043] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0044] The following describes in detail a specific solution of a high-definition digital imaging method applicable to an ENT endoscope provided by the present invention with reference to the accompanying drawings.

[0045] See also Figure 1 , which shows a flow chart of a high-definition digital imaging method applicable to an ENT endoscope provided by an embodiment of the present invention, the specific method includes:

[0046] Step S1: Acquire endoscopic images of the observation area in time series.

[0047] In an embodiment of the present invention, since the internal structure of the observed part is complex and there are occlusions between them, the detailed features of the local area may be difficult to present clearly when the image light source is insufficient. Therefore, in order to improve the quality of image imaging, it is necessary to analyze the endoscopic image; first, a multi-camera endoscope is used, and a high-intensity cold light source such as an LED or fiber optic light source is set at the front end of the camera to provide a stable light source, and a CMOS high-definition camera at the front end of the endoscope is used to capture the endoscopic image of the observed part in real time, and automatically store it in the device cloud storage system according to the acquisition time; the endoscopic image of the observed part is acquired in time sequence; wherein, in an embodiment of the present invention, the observed part is the ear, nose, and throat.

[0048] It should be noted that in order to facilitate the subsequent image processing process, the acquired endoscopic images are preprocessed to enhance the image quality. It should be noted that the image preprocessing operation is a technical means well known to those skilled in the art and can be specifically set according to the specific implementation scenario. In one embodiment of the present invention, a denoising and cleaning algorithm is used for image processing to remove interference signals in the image, which can highlight the contours and details of the image and make the image clearer. The image is then linearly normalized to unify the contrast and brightness differences between different images. The specific means are technical means well known to those skilled in the art and will not be described in detail here.

[0049] Step S2: obtaining multiple superpixel real-time blocks of the real-time endoscopic image; obtaining the dark area probability parameter of each superpixel real-time block based on the difference in pixel value distribution between different superpixel real-time blocks; obtaining the dark area structure activity coefficient of each superpixel real-time block based on the pixel value distribution of the corresponding area between different adjacent endoscopic images of each superpixel real-time block and the dark area probability parameter.

[0050] Endoscopic images typically contain rich details and complex texture information, making it difficult to process pixels directly. To reduce computational complexity, images are segmented into multiple superpixel blocks, which helps simplify image analysis and understanding. Multiple superpixel blocks are obtained in real time for real-time endoscopic images.

[0051] It should be noted that, in one embodiment of the present invention, superpixel segmentation is used to obtain multiple superpixel real-time blocks for real-time endoscopic images. Superpixel segmentation can aggregate pixels in real-time endoscopic images into several larger pixel blocks based on similarity criteria such as color, texture, or shape, i.e., superpixel real-time blocks. This helps significantly reduce the dimensionality of image processing while retaining important image information. The specific superpixel segmentation is a technical means well known to those skilled in the art and will not be described in detail here.

[0052] Due to the differences in spatial position and light reflection characteristics of different tissue structures, the light reflection characteristics are random, resulting in some areas of the image reflecting less light, resulting in the formation of local dark areas in the image. Light reflection varies significantly between regions. By analyzing the differences in pixel value distribution between different superpixel real-time blocks, we can reflect the changes in light between different areas in the image. The greater the variation, the more likely it is to present a single feature, and the larger the dark area probability parameter. Based on the differences in pixel value distribution between different superpixel real-time blocks, the dark area probability parameter of each superpixel real-time block is obtained.

[0053] Preferably, in one embodiment of the present invention, the method for obtaining the dark area probability parameter includes:

[0054] Obtain the mean of all pixel values ​​in each superpixel real-time block as the block pixel level; obtain the dark area probability coefficient based on the difference distribution of block pixel levels between different superpixel real-time blocks;

[0055] Based on the difference in block pixel levels between each superpixel real-time block and different superpixel real-time blocks in the corresponding neighborhood, as well as the dark area possibility coefficients of different superpixel real-time blocks in the neighborhood, the dark area probability parameter of each superpixel real-time block is obtained. The difference in block pixel levels and the dark area possibility coefficients are positively correlated with the dark area probability parameter.

[0056] Among them, the greater the difference in the pixel level between super-pixel real-time blocks, the more inconsistent the pixel distribution between super-pixel real-time blocks, the more single features the super-pixel real-time blocks present, the greater the dark area possibility coefficient, and the greater the credibility of the existence of dark areas.

[0057] In one embodiment of the present invention, the formula for the dark area probability parameter is expressed as:

[0058] ;

[0059] in, Indicates the Dark area probability parameter of superpixel real-time segmentation; Indicates the neighborhood The dark area probability coefficient of superpixel real-time block; Indicates the The block pixel level of superpixel real-time block; Indicates the neighborhood The block pixel level of superpixel real-time block; Indicates the The number of superpixel real-time blocks within the neighborhood of a superpixel real-time block; Indicates taking the absolute value; Represents the normalization function.

[0060] In the formula of the dark zone probability parameter, Indicates calculation of The superpixel is divided into blocks in real time and the corresponding neighborhood is divided into The difference in the pixel level between the real-time superpixel blocks is greater. The greater the difference, the less similar the pixel distribution between the blocks is, the more likely there is a dark area influence, the greater the dark area possibility coefficient, and the larger the dark area probability parameter.

[0061] It should be noted that, in one embodiment of the present invention, the neighborhood range is a range consisting of all adjacent superpixel real-time blocks centered on each superpixel real-time block, and the central superpixel real-time block is not included in the range; in other embodiments of the present invention, the neighborhood range can be set according to specific circumstances, and is not limited or elaborated here.

[0062] Preferably, considering that the smaller the difference in pixel level of the blocks is, the closer the pixel distribution of the superpixel real-time blocks is, the more likely they are to correspond to the same type of tissue areas; in one embodiment of the present invention, the method for obtaining the possible coefficient of the dark area includes:

[0063] Obtain the pixel-level differences between different superpixel real-time blocks and normalize them as the block difference index;

[0064] If the block difference index between the super-pixel real-time blocks is less than the preset difference threshold, the corresponding super-pixel real-time blocks are regarded as the same type of blocks;

[0065] The ratio of the number of similar blocks corresponding to each superpixel real-time block to the number of all superpixel real-time blocks is obtained, and negative correlation mapping is performed as the possible dark area coefficient of each superpixel real-time block.

[0066] It should be noted that, in one embodiment of the present invention, the preset difference threshold is 0.3; in other embodiments of the present invention, the preset difference threshold may be set according to specific circumstances, which is not limited or elaborated herein.

[0067] It should be noted that, in one embodiment of the present invention, since the ratio of the number of similar blocks corresponding to each superpixel real-time block to the number of all superpixel real-time blocks is 0-1, the difference between the positive integer 1 and the ratio can be calculated to represent the contrast value for negative correlation mapping. The larger the ratio, the smaller the difference. In other embodiments of the present invention, the reciprocal of the contrast value can also be taken for negative correlation mapping. In order for the denominator of the formula to be 0, an artificially set threshold, such as 0.01, needs to be added. The specific means are technical means well known to those skilled in the art and will not be elaborated here.

[0068] In one embodiment of the present invention, the formula for the dark area possible coefficient is expressed as:

[0069] ;

[0070] in, Indicates the The dark area probability coefficient of superpixel real-time block; Indicates the The number of similar blocks corresponding to each superpixel real-time block; Indicates the number of all superpixel real-time patches.

[0071] Taking into account the changes in the light reflection characteristics of the lesion area of ​​the observation site, dark areas will appear. When the light source angle changes, the pixel value distribution will also change accordingly. By analyzing the pixel value distribution of the corresponding areas between adjacent endoscopic images, the changes in the dark areas in the image can be captured. The dark area probability parameter is used to quantify the possibility of the appearance of dark areas in the image. The larger the dark area probability parameter, the more dark area structures there are. According to the pixel value distribution of the corresponding areas between different adjacent endoscopic images of each superpixel real-time block and the dark area probability parameter, the dark area structure activity coefficient of each superpixel real-time block is obtained.

[0072] Preferably, in one embodiment of the present invention, the method for obtaining the dark area structure activity coefficient can be found in Figure 2 , which shows a flow chart of a method for obtaining the dark area structure activity coefficient, including:

[0073] Step S201: Use a corner detection method to obtain the corner points of each superpixel real-time block in the corresponding area of ​​each endoscopic image; obtain the difference in the number of corresponding corner points of each superpixel real-time block between different adjacent endoscopic images as the first difference feature.

[0074] When the dark area of ​​the image is affected by a lesion, the lesion area usually has the characteristics of a depression or a mass, and the distribution of corner points is relatively complex, which causes the number of corner points to change significantly when the light source changes. The greater the difference in the number of corner points, the greater the first difference feature.

[0075] It should be noted that, in the embodiments of the present invention, existing corner detection algorithms may be used to obtain corner points within a region, such as the Shi-Tomasi corner detection algorithm. Specific methods are well known to those skilled in the art and will not be elaborated herein.

[0076] Step S202: Obtain the mean square error of the set composed of all pixel values ​​in the corresponding area between different adjacent endoscopic images of each superpixel real-time block as the second difference feature.

[0077] The mean square error is used to measure the difference in pixel value distribution in the corresponding area of ​​each superpixel real-time block between different adjacent endoscopic images, reflecting the changes in tissue structure. Since the pixel value distribution in the dark area of ​​the lesion changes significantly when the light source angle changes, the greater the difference in pixel value distribution, the greater the mean square error, and the greater the second difference feature.

[0078] Step S203: According to the first difference feature, the second difference feature and the dark area probability parameter corresponding to each superpixel real-time block, the dark area structure activity coefficient of each superpixel real-time block is obtained, and the first difference feature, the second difference feature and the dark area probability parameter are all positively correlated with the dark area structure activity coefficient.

[0079] Among them, the larger the first difference feature, the greater the difference in the number of corner points in the corresponding areas between adjacent endoscopic images; the larger the second difference feature, the greater the difference in pixel value distribution in the corresponding areas between adjacent endoscopic images, and the more likely it is that there is a dark area with lesions; the larger the dark area probability parameter, the greater the dark area structure activity coefficient.

[0080] In one embodiment of the present invention, the formula for the dark area structure activity coefficient is expressed as:

[0081] ;

[0082] in, Indicates the The dark area structure activity coefficient of superpixel real-time block; Indicates the Dark area probability parameter of superpixel real-time segmentation; Indicates the Superpixel real-time blocking The set of all pixel values ​​in the corresponding area of ​​an endoscopic image; Indicates the Superpixel real-time blocking The set of all pixel values ​​in the corresponding area of ​​an endoscopic image; Indicates the Superpixel real-time blocking and The difference in the number of corner points in the corresponding area between the endoscopic images is the first difference feature; Indicates the Superpixel real-time blocking and The mean square error of the set of all pixel values ​​in the corresponding area between the endoscopic images is the second difference feature; Indicates the number of all endoscopic images.

[0083] In the formula of the dark area structure activity coefficient, Indicates calculation of Superpixel real-time blocking and The product of the first difference feature and the second difference feature in the corresponding area between the endoscopic images is larger, the larger the first difference feature, the larger the second difference feature, the more likely the structure is to change, and the more likely the dark area of ​​the lesion is to exist. It represents the mean value used to quantify the relationship between all adjacent endoscopic images. The larger the mean value, the more significant the change, the greater the possibility of lesions, and the larger the dark area probability parameter, the more active the dark area structure.

[0084] Step S3: Based on the pixel similarity between each superpixel real-time block and different superpixel real-time blocks in the corresponding areas of all endoscopic images, the overall pixel correlation of each superpixel real-time block is obtained; based on the fluctuation degree distribution of pixel values ​​between each superpixel real-time block and different superpixel real-time blocks in the corresponding neighborhood, as well as the overall pixel correlation, the shadow interference coefficient of each superpixel real-time block is obtained.

[0085] In reality, local areas of an image are affected by the complex surrounding structure. The closer the pixel values ​​of adjacent areas, the more consistent the global features of the corresponding areas in all endoscopic images. Analyzing pixel similarity reflects the consistency of the global features between superpixel real-time blocks across all corresponding areas in the endoscopic image, reflecting the overall structure of the image. Based on the pixel similarity between each superpixel real-time block and different superpixel real-time blocks within its corresponding neighborhood across all corresponding areas in the endoscopic image, the overall pixel correlation of each superpixel real-time block is obtained.

[0086] Preferably, in one embodiment of the present invention, the method for obtaining the overall pixel correlation includes:

[0087] For each superpixel real-time block and each superpixel real-time block within the corresponding neighborhood, the correlation coefficient of the set of block pixel levels in the corresponding area of ​​all endoscopic images between the superpixel real-time blocks is obtained as the pixel local correlation;

[0088] The average of the local correlations of the corresponding pixels between each superpixel real-time block and all superpixel real-time blocks in the corresponding neighborhood is obtained as the overall pixel correlation of each superpixel real-time block.

[0089] In one embodiment of the present invention, the formula for the overall pixel correlation is expressed as:

[0090] ;

[0091] in, Indicates the The overall pixel correlation of superpixels in real-time blocks; Indicates the The number of different superpixel real-time blocks within the neighborhood of a superpixel real-time block; Indicates the The superpixel real-time blocking is a set of pixel levels in the corresponding area of ​​all endoscopic images; Indicates the neighborhood The superpixel real-time blocking is a set of pixel levels in the corresponding area of ​​all endoscopic images; Indicates the Superpixel real-time block and neighborhood The correlation coefficient of the set composed of the block pixel levels in the corresponding area of ​​all endoscopic images between the real-time superpixel blocks is the local correlation between the corresponding pixels.

[0092] In the overall pixel correlation, Indicates the The average value of the local pixel correlation between a superpixel real-time block and all superpixel real-time blocks in the neighborhood, that is, the overall pixel correlation. The more similar the pixel value distribution is, the more consistent the block pixel level is, the greater the local pixel correlation is, and the greater the overall pixel correlation is.

[0093] It should be noted that, in one embodiment of the present invention, the correlation coefficient is a Pearson correlation coefficient. The specific Pearson correlation coefficient is a technical means well known to those skilled in the art and will not be described in detail here.

[0094] The degree of pixel value fluctuation reflects the brightness or color changes of the superpixel real-time block in the image. Because shadow interference causes significant differences in pixel values ​​between the shadow area and the surrounding normal area, this leads to pixel value fluctuations. By combining the pixel value fluctuation distribution and the overall pixel correlation, we can more comprehensively assess the impact of shadow interference on the image. Therefore, based on the pixel value fluctuation distribution between each superpixel real-time block and different superpixel real-time blocks in the corresponding neighborhood, as well as the overall pixel correlation, we obtain the shadow interference coefficient for each superpixel real-time block.

[0095] Preferably, in one embodiment of the present invention, the method for obtaining the shadow interference coefficient is as follows: Figure 3 , which shows a flow chart of a method for obtaining a shadow interference coefficient, including:

[0096] Step S301: Obtain the fluctuation degree of all pixel values ​​in each superpixel real-time block as the block pixel fluctuation feature.

[0097] The degree of fluctuation of all pixel values ​​reflects the distribution characteristics of pixel values. The greater the degree of fluctuation, the more uneven the pixel value distribution and the less interference from shadows. The smaller the degree of fluctuation, the more uniform the pixel value distribution and the more likely it is that the tissue or instrument blocks the light source of the endoscope, causing the brightness of the area to decrease or disappear completely, and the greater the interference from shadows.

[0098] It should be noted that in one embodiment of the present invention, the degree of fluctuation is expressed by calculating the variance. The larger the variance, the greater the degree of fluctuation, and the smaller the variance, the smaller the degree of fluctuation. In other embodiments of the present invention, the standard deviation and range may also be used to express the degree of fluctuation. The specific means are well known to those skilled in the art and will not be described in detail here.

[0099] Step S302: Obtain the mean of the block pixel fluctuation characteristics of all superpixel real-time blocks within the neighborhood range as the neighborhood pixel fluctuation feature.

[0100] The overall level of the block pixel fluctuation characteristics of all superpixel real-time blocks in the neighborhood is quantified by taking the average value, which serves as the benchmark for subsequent comparisons.

[0101] Step S303: Obtain the difference between the block pixel fluctuation characteristics and the neighborhood pixel fluctuation characteristics of each superpixel real-time block, and perform negative correlation mapping as the distribution stability; weight the corresponding distribution stability according to the overall pixel correlation of each superpixel real-time block to obtain the shadow interference coefficient of each superpixel real-time block.

[0102] In one embodiment of the present invention, the shadow interference coefficient is expressed as follows:

[0103] ;

[0104] in, Indicates the Shadow interference coefficient of superpixel real-time block; Indicates the The fluctuation degree of all pixel values ​​in a superpixel real-time block, that is, the block pixel fluctuation characteristics; Indicates the The average value of the block pixel fluctuation characteristics of all superpixel real-time blocks in the neighborhood of a superpixel real-time block, that is, the neighborhood pixel fluctuation characteristics; Indicates the The overall pixel correlation of superpixels in real-time blocks; Represents a natural constant.

[0105] In the formula for the shadow interference coefficient, The exponential function with a natural constant as the base will be Perform negative correlation mapping, i.e., the degree of distribution stability; Indicates calculation of The difference between the block pixel fluctuation characteristics of a superpixel real-time block and the neighborhood pixel fluctuation characteristics, the larger the difference, the larger the block pixel fluctuation characteristics of the corresponding superpixel real-time block, the more uneven the pixel value distribution in the block, the smaller the distribution stability, and the less interference from shadows. Conversely, the larger the difference, the smaller the block pixel fluctuation characteristics of the corresponding superpixel real-time block, the more uniform the pixel value distribution in the block, the greater the distribution stability, and the greater the shadow interference; the distribution stability is weighted based on the overall pixel correlation. Therefore, the greater the overall pixel correlation, the more each superpixel real-time block is affected by the surrounding blocks containing complex structures, the greater the distribution stability, the greater the shadow interference, and the larger the shadow interference coefficient.

[0106] Step S4: Obtain the detail fuzzy coefficient of each superpixel real-time block according to the dark area structure activity coefficient and shadow interference coefficient of each superpixel real-time block; and correct the initial grid according to the detail fuzzy coefficient of each superpixel real-time block to obtain a corrected grid.

[0107] By comprehensively considering the dark area structure activity coefficient and the shadow interference coefficient, the degree of detail blurring of the superpixel real-time block is more comprehensively evaluated. The dark area structure activity coefficient assesses the likelihood that the superpixel real-time block belongs to the lesion dark area. The larger the dark area structure activity coefficient, the higher the likelihood of belonging to the lesion dark area. The shadow interference coefficient reflects the degree to which the superpixel real-time block is obscured by shadows. The larger the shadow interference coefficient, the greater the degree of shadow obscuration, resulting in more complex and numerous detail features, and the detail features becoming more blurred and unclear. The detail blurring coefficient of each superpixel real-time block is obtained based on the dark area structure activity coefficient and shadow interference coefficient of each superpixel real-time block.

[0108] Preferably, in one embodiment of the present invention, the method for obtaining the detail fuzziness coefficient includes:

[0109] The dark area structure activity coefficient and shadow interference coefficient of each superpixel real-time block are fused to obtain the detail fuzzy coefficient of each superpixel real-time block.

[0110] It should be noted that, in some embodiments of the present invention, the dark area structure activity coefficient and the shadow interference coefficient can be fused by addition or multiplication. The specific means are technical means well known to those skilled in the art and will not be elaborated here.

[0111] In one embodiment of the present invention, the formula for the detail blur coefficient is expressed as: ;in, Indicates the Detail blur coefficient of superpixel real-time block; Indicates the Shadow interference coefficient of superpixel real-time block; Indicates the The dark area structure activity coefficient of superpixels real-time block.

[0112] If the window size is set to a smaller size, although local detail information can be better captured, global information may be ignored; if the window size is set to a larger size, although global information can be better considered, more noise and computational complexity may be introduced; by obtaining the detail fuzzy coefficient to reflect the distribution state of the detail features within the block, it is helpful to adjust the size of the initial grid; the initial grid is corrected according to the detail fuzzy coefficient of each superpixel real-time block to obtain a corrected grid.

[0113] Preferably, in one embodiment of the present invention, the method for obtaining the modified grid includes:

[0114] Get the initial size of the initial grid;

[0115] The detail fuzzy coefficient of each superpixel real-time block is negatively correlated and mapped, and the product of the negative correlation mapping result and the initial size is obtained as the correction size to obtain the correction grid.

[0116] In one embodiment of the present invention, the formula for correcting the size is expressed as:

[0117] ;

[0118] in, Indicates the The real-time block of superpixels corresponds to the revised size of the initial grid; Indicates the Detail blur coefficient of superpixel real-time block; Indicates the Superpixels are divided into blocks corresponding to the initial grid in real time; Represents a natural constant.

[0119] In the formula for correcting the size, It means that the exponential function with the natural constant as the base is used to convert For negative correlation mapping, the larger the detail blur coefficient, the lower the contrast of the corresponding area, the more targeted local enhancement is needed, the smaller the initial size is adjusted, and the smaller the correction size is.

[0120] Step S5: Perform the CLAHE algorithm on each superpixel in the endoscopic image in real time according to the corrected grid to obtain an enhanced endoscopic image.

[0121] After obtaining the corrected grid of each superpixel real-time block, the superpixel real-time block is adaptively enhanced based on the CLAHE algorithm to enhance the detail information in the block, so that the originally hidden or blurred image details are fully displayed, and the information content of the image is increased; the specific CLAHE algorithm is a technical means well known to those skilled in the art and will not be described in detail here.

[0122] It should be noted that the super-pixel real-time blocks are enhanced in sequence from the upper left corner to the lower right corner of the endoscopic image. After the enhanced endoscopic image is obtained, it can also be smoothed by Gaussian filtering to eliminate the fast effect within the super-pixel real-time blocks, ensuring that the final endoscopic image is smoother and clearer, which helps to understand the tissue structure within the observation area and improve the efficiency of subsequent diagnostic analysis.

[0123] In summary, the present invention analyzes the pixel value distribution between different superpixel real-time blocks in a real-time endoscopic image to obtain the dark area structure activity coefficient of each superpixel real-time block; obtains the overall pixel correlation of each superpixel real-time block based on the pixel similarity between each superpixel real-time block and different superpixel real-time blocks within the corresponding neighborhood range in all endoscopic image corresponding areas; obtains the shadow interference coefficient of each superpixel real-time block by combining the fluctuation degree distribution of pixel values ​​between each superpixel real-time block and different superpixel real-time blocks within the corresponding neighborhood range; and then obtains the detail fuzzy coefficient of each superpixel real-time block, corrects the initial grid, obtains the corrected grid, and obtains the enhanced endoscopic image. The present invention improves the imaging quality by obtaining a suitable grid for endoscopic image enhancement processing.

[0124] It should be noted that the order in which the embodiments of the present invention are described above is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0125] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

Claims

1. A high-definition digital imaging method suitable for ENT endoscopes, characterized in that: The method comprises: Acquire endoscopic images of the observation site in time sequence; Obtain multiple superpixel real-time blocks of a real-time endoscopic image; obtain a dark area probability parameter for each superpixel real-time block based on differences in pixel value distributions between different superpixel real-time blocks; obtain a dark area structure activity coefficient for each superpixel real-time block based on pixel value distributions of corresponding areas between different adjacent endoscopic images of each superpixel real-time block and the dark area probability parameter; Based on the pixel similarity between each superpixel real-time block and different superpixel real-time blocks within the corresponding neighborhood in all corresponding areas of the endoscopic image, the overall pixel correlation of each superpixel real-time block is obtained; based on the fluctuation degree distribution of pixel values ​​between each superpixel real-time block and different superpixel real-time blocks within the corresponding neighborhood, and the overall pixel correlation, the shadow interference coefficient of each superpixel real-time block is obtained; According to the dark area structure activity coefficient and shadow interference coefficient of each superpixel real-time block, the detail fuzzy coefficient of each superpixel real-time block is obtained; according to the detail fuzzy coefficient of each superpixel real-time block, the initial grid is corrected to obtain a corrected grid; The CLAHE algorithm is performed on each superpixel in the endoscopic image in real time according to the corrected grid to obtain the enhanced endoscopic image.

2. A high-definition digital imaging method suitable for ENT endoscope according to claim 1, characterized in that: The method for obtaining the dark area probability parameter includes: Obtain the mean of all pixel values ​​in each superpixel real-time block as the block pixel level; obtain the dark area probability coefficient based on the difference distribution of block pixel levels between different superpixel real-time blocks; According to the difference in block pixel levels between each superpixel real-time block and different superpixel real-time blocks in the corresponding neighborhood, as well as the dark area possibility coefficients of different superpixel real-time blocks in the neighborhood, the dark area probability parameter of each superpixel real-time block is obtained. The difference in block pixel levels and the dark area possibility coefficients are both positively correlated with the dark area probability parameter.

3. A high-definition digital imaging method suitable for ENT endoscope according to claim 2, characterized in that: The method for obtaining the possible coefficients of the dark area includes: Obtain the pixel-level differences between different superpixel real-time blocks and normalize them as the block difference index; If the block difference index between the super-pixel real-time blocks is less than the preset difference threshold, the corresponding super-pixel real-time blocks are regarded as the same type of blocks; The ratio of the number of similar blocks corresponding to each superpixel real-time block to the number of all superpixel real-time blocks is obtained, and negative correlation mapping is performed as the possible dark area coefficient of each superpixel real-time block.

4. The high-definition digital imaging method for ENT endoscope according to claim 1, characterized in that: The method for obtaining the dark area structure activity coefficient includes: A corner detection method is used to obtain the corner points of each superpixel real-time block in the corresponding area of ​​each endoscopic image; the difference in the number of corresponding corner points of each superpixel real-time block between different adjacent endoscopic images is obtained as a first difference feature; Obtaining the mean square error of the set composed of all pixel values ​​in the corresponding area between different adjacent endoscopic images of each superpixel real-time block as the second difference feature; According to the first difference feature, the second difference feature and the dark area probability parameter corresponding to each superpixel real-time block, the dark area structure activity coefficient of each superpixel real-time block is obtained, and the first difference feature, the second difference feature and the dark area probability parameter are all positively correlated with the dark area structure activity coefficient.

5. The high-definition digital imaging method for ENT endoscope according to claim 2, characterized in that: The method for obtaining the overall pixel correlation includes: For each superpixel real-time block and each superpixel real-time block within the corresponding neighborhood, the correlation coefficient of the set of block pixel levels in the corresponding area of ​​all endoscopic images between the superpixel real-time blocks is obtained as the pixel local correlation; The average of the local pixel correlations between each superpixel real-time block and all superpixel real-time blocks in the corresponding neighborhood is obtained as the overall pixel correlation of each superpixel real-time block.

6. The high-definition digital imaging method for ENT endoscope according to claim 1, characterized in that: The method for obtaining the shadow interference coefficient includes: Obtain the fluctuation degree of all pixel values ​​in each superpixel real-time block as the block pixel fluctuation feature; Obtain the mean of the block pixel fluctuation characteristics of all superpixel real-time blocks within the neighborhood as the neighborhood pixel fluctuation feature; The difference between the block pixel fluctuation characteristics and the neighborhood pixel fluctuation characteristics of each superpixel real-time block is obtained, and a negative correlation mapping is performed as the distribution stability; the corresponding distribution stability is weighted according to the overall pixel correlation of each superpixel real-time block to obtain the shadow interference coefficient of each superpixel real-time block.

7. The high-definition digital imaging method for ENT endoscope according to claim 1, characterized in that: The method for obtaining the detail fuzzy coefficient includes: The dark area structure activity coefficient and shadow interference coefficient of each superpixel real-time block are fused to obtain the detail fuzzy coefficient of each superpixel real-time block.

8. The high-definition digital imaging method for ENT endoscope according to claim 1, characterized in that: The method for obtaining the modified grid includes: Get the initial size of the initial grid; Negative correlation mapping is performed on the detail fuzzy coefficient of each superpixel real-time block, and the product of the negative correlation mapping result and the initial size is obtained as the correction size to obtain the correction grid.

9. The high-definition digital imaging method for ENT endoscope according to claim 3, characterized in that: The preset difference threshold is 0.

3.

10. The high-definition digital imaging method for ENT endoscope according to claim 5, characterized in that: The correlation coefficient is the Pearson correlation coefficient.

Images