Real-time enhancement method of endoscopic images for emergency treatment

Abstract

This invention relates to the field of image enhancement technology, specifically to a real-time enhancement method for endoscopic images used in emergency medicine. The method involves acquiring images using an endoscope, obtaining an initial adaptive convolution kernel based on the difference between the gradient values ​​of feature points and the average gradient values ​​of their four neighboring regions, and then negatively feeding back the kernel size by evaluating the results to obtain the optimal adaptive convolution kernel for each pixel in the endoscopic image. The optimal adaptive convolution kernel for non-feature points is obtained by utilizing differences in distance and grayscale values ​​between non-feature points and feature points. Finally, the optimal adaptive convolution kernel for the obtained endoscopic image is used to sharpen each pixel in the image, resulting in the best-sharpened image. This invention improves the efficiency of endoscopic image enhancement and helps reduce the computational load of the endoscopic image enhancement process.

Description

Technical Field

[0001] The present invention relates to the technical field of image enhancement, and in particular to a real-time enhancement method for endoscopic images used in emergency. Background Art

[0002] Endoscope is a common tool in the medical industry. With the continuous progress and development of medical teaching, the requirements for details in images involved in medical treatment are gradually increasing. However, due to the limitations of equipment accuracy, patient cooperation and other factors, the images taken by the endoscope may appear blurred or unrecognizable. At this time, it is generally necessary to sharpen the image to enhance the features of the image accordingly, so that the teaching materials of medical teaching are clearer and it is convenient to explain medical knowledge during the medical teaching process.

[0003] There are many methods for sharpening images, and Laplace sharpening is a commonly used method. In the Laplace sharpening process, the selection of the convolution kernel size often has different effects on the image sharpening results. In the Laplace sharpening process of the prior art, the convolution kernel is generally adjusted manually to achieve a better sharpening effect. This adjustment method relies on the experience of personnel and is not conducive to batch sharpening of images. Summary of the invention

[0004] In order to solve the technical problem that when Laplace sharpening is used in the prior art to process emergency endoscopic images, the sharpening result is usually optimized by manually adjusting the convolution kernel, which is not conducive to batch processing of images, the purpose of the present invention is to provide a real-time enhancement method for emergency endoscopic images, the method comprising:

[0005] Step S1: using an endoscope to collect images and pre-process the images;

[0006] Step S2: Select pixel points in the image that are greater than a first gradient value threshold as feature points, and obtain an initial adaptive convolution kernel based on the difference between the gradient value of the feature point and the average gradient value of the feature point and its four neighborhoods;

[0007] Step S3: Perform Laplace sharpening on the feature point, compare the grayscale value change difference between the pixel point and its four neighborhoods before and after sharpening, and construct a sharpening evaluation of the feature point;

[0008] Step S4: obtaining the best adaptive convolution kernel of the feature point according to the sharpening evaluation of the feature point;

[0009] Step S5: Obtain the best adaptive convolution kernel of the non-feature point according to the distance between the non-feature point and the feature point in the image, the gray value difference, and the difference in the gradient values ​​of the surrounding four neighborhoods;

[0010] Step S6: Perform Laplace sharpening on the image according to the optimal adaptive convolution kernel of the feature points and non-feature points to obtain a sharpened image.

[0011] Furthermore, in step S1, the image is captured by using an endoscope, and the image is preprocessed specifically as follows: the throat image of the patient is captured by using an endoscope, and the image is preprocessed including denoising and grayscale processing.

[0012] Furthermore, the step S2 is specifically as follows:

[0013] Calculate the gradient value of each pixel on the image and the average gradient value of all pixels;

[0014] Pixels with values ​​greater than the average gradient are recorded as feature points on the image, and other pixels are recorded as non-feature points on the image;

[0015] According to the difference between the gradient value of the feature point on the image and the average gradient value of its four neighborhoods, the center value of the initial adaptive convolution kernel corresponding to each feature pixel on the image is obtained;

[0016] According to the center value of the initial adaptive convolution kernel corresponding to each feature pixel point on the image, the initial adaptive convolution kernel corresponding to each feature point on the image is obtained.

[0017] Furthermore, the step S3 is specifically as follows:

[0018] According to the initial adaptive convolution kernel corresponding to each feature point on the image, Laplace sharpening is performed on the feature points;

[0019] The sharpening effect of the feature point on the image is evaluated based on the grayscale value of the feature point and its neighboring pixels before sharpening, the grayscale value of the feature point and its neighboring pixels after sharpening, and the grayscale value difference between the feature point and each pixel in the four neighborhoods after sharpening.

[0020] Furthermore, the difference between the grayscale value of a feature point and its neighboring pixels before sharpening and the grayscale value of the feature point and its neighboring pixels after sharpening, and the difference in grayscale value between the feature point and each pixel in the four neighborhoods after sharpening are positively correlated with the evaluation of the sharpening effect of the feature point on the image.

[0021] Furthermore, the step S4 is specifically as follows:

[0022] According to the evaluation results of the sharpening effect of each feature point on the image, the numerical range of the evaluation results belonging to the preset optimal result is set, and the center value of the convolution kernel that is not in the numerical range of the optimal result is sharpened or suppressed.

[0023] Furthermore, the sharpening enhancement or sharpening suppression of the center value of the convolution kernel that is not in the numerical range of the optimal result is specifically:

[0024] For evaluation results that are lower than the lower limit of the numerical range of the best result, increase the center value of the convolution kernel;

[0025] For evaluation results that are higher than the upper limit of the numerical range of the best result, reduce the center value of the convolution kernel.

[0026] Furthermore, according to increasing or decreasing the center value of the convolution kernel, Laplace sharpening is performed again on the feature points on the image until the sharpening evaluation result of the corresponding feature point is within the numerical range of the preset optimal result.

[0027] Furthermore, the step S5 is specifically as follows:

[0028] Obtain the degree of difference between a non-feature point and its nearest feature point on the image according to the distance between the non-feature point on the image and its nearest feature point, the gray value difference between the non-feature point on the image and its nearest feature point, and the average gradient difference between the non-feature point on the image and its four neighboring pixels and its nearest feature point and its four neighboring pixels;

[0029] Obtaining the center value of the best adaptive convolution kernel of a non-feature point on the image according to the center value of the best adaptive convolution kernel of a feature point on the image that is closest to the non-feature point and the degree of difference between the non-feature point on the image and the feature point that is closest to the non-feature point on the image;

[0030] According to the center value of the best adaptive convolution kernel of the non-feature point on the image, an adaptive convolution kernel of the non-feature point on the image is obtained.

[0031] Furthermore, the distance between a non-feature point on the image and its nearest feature point is negatively correlated with the degree of difference between the non-feature point and its nearest feature point, and the grayscale value difference between the non-feature point on the image and its nearest feature point, and the average gradient difference between the non-feature point and the pixels of its four neighborhoods on the image, the nearest feature point and the pixels of its four neighborhoods are positively correlated with the degree of difference between the non-feature point and its nearest feature point.

[0032] The present invention has the following beneficial effects:

[0033] 1. Improved the efficiency of endoscopic image enhancement for emergency use. By selecting pixels in the image that are greater than the first gradient value threshold as feature points, the initial adaptive convolution kernel is obtained based on the difference between the gradient value of the feature point and the average gradient value of its four neighborhoods; Laplace sharpening is performed on the feature point, and the difference in grayscale value changes between the pixel point and its four neighborhoods before and after sharpening is compared to construct a sharpening evaluation of the feature point; based on the sharpening evaluation of the feature point, the optimal adaptive convolution kernel of the feature point is obtained; based on the distance between the non-feature point and the feature point in the image, the difference in grayscale value, and the difference in gradient values ​​of the surrounding four neighborhoods, the optimal adaptive convolution kernel of the non-feature point is obtained; based on the optimal adaptive convolution kernel of the feature point and the non-feature point, Laplace sharpening is performed on the image to obtain a sharpening map. In this way, the sharpening effect of the convolution kernel is digitized using the evaluation method, avoiding artificial adjustment of the convolution kernel to optimize the sharpening effect, and improving the efficiency of endoscopic image enhancement for emergency use.

[0034] Second, it is helpful to reduce the amount of calculation of non-feature points in the image sharpening process, and is helpful to save computing power in the process of endoscopic image enhancement. The present invention records the degree of difference between non-feature points and feature points, as well as the best adaptive convolution kernel of non-feature points. If the degree of difference between non-feature points and feature points calculated subsequently exists in the record table, the best adaptive convolution kernel of this non-feature point is directly obtained, reducing the amount of calculation. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0036] Figure 1 A flow chart of a method for real-time enhancement of endoscopic images for emergency use provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0037] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the specific implementation method, structure, characteristics and effects of the real-time enhancement method of endoscopic images for emergency proposed by the present invention are described in detail below in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. Unless otherwise defined, all technical and scientific terms used in this document have the same meaning as commonly understood by technicians in the technical field of the present invention.

[0038] The scenario targeted by the present invention is: in the medical field, endoscopes are used in a variety of medical testing processes. In the process of medical emergency, a large part of the cases are respiratory infection-related diseases. In the process of throat testing, endoscopic images often have problems such as unclear picture details due to the influence of factors such as equipment, and medical judgments often require more detailed pictures for analysis to obtain more accurate judgment results. Therefore, Laplace sharpening technology is used to obtain the best effect after image sharpening through adaptive convolution kernels.

[0039] The specific scheme of the real-time enhancement method of endoscopic images for emergency treatment provided by the present invention is described in detail below with reference to the accompanying drawings.

[0040] See also Figure 1 , which shows a flow chart of a real-time enhancement method for emergency endoscopic images provided by an embodiment of the present invention, the method comprising the following steps:

[0041] Step S1: Use an endoscope to collect images and pre-process the images.

[0042] This embodiment takes the throat image as an example. First, an endoscope is used to collect an image of the patient's throat area, and denoising and grayscale processing are performed based on the specific information shown in the image of the patient's throat area, so that the features shown in the throat image are more obvious.

[0043] Step S2: Select pixel points in the image that are greater than a first gradient value threshold as feature points, and obtain an initial adaptive convolution kernel based on the difference between the gradient value of the feature point and the average gradient value of the feature point and its four neighborhoods.

[0044] Since the vascular lines in the throat area in the image show a difference from the surrounding area, the vascular lines obviously show a darker color than the surrounding area. The gradient of the pixel points belonging to the vascular lines is significantly greater than the gradient value of the pixel points belonging to the surrounding tissue. In the image sharpening algorithm, the Laplace sharpening effect is better than other sharpening algorithms, but the Laplace sharpening effect is related to the selection of the Laplace convolution kernel size, that is, the best sharpened image can be obtained by continuously adjusting the size of the Laplace convolution kernel. Therefore, this implementation adopts the Laplace sharpening method to sharpen the image collected by the endoscope. In this embodiment, the other steps in the algorithm are kept unchanged, and the image sharpening effects obtained by convolution using Laplace convolution kernels of different sizes are also different. Laplace sharpening often makes the changes in edge pixels more prominent. According to the different degree of expression of the difference in grayscale value changes between the pixel points and the surrounding areas, the effect required to achieve pixel sharpening is also different. In this embodiment, feature points are selected by the difference in gradient values ​​of the pixel points in the image. The best adaptive Laplace convolution kernel for the feature points is obtained.

[0045] In this embodiment, step S2 is specifically as follows: calculating the gradient value of each pixel on the image and the average gradient value of all pixels; recording the pixel points greater than the average gradient value as feature points on the image, and recording the other pixel points as non-feature points on the image; obtaining the center value of the initial adaptive convolution kernel corresponding to each feature pixel on the image based on the difference between the gradient value of the feature point on the image and the average gradient value of its four neighborhoods; obtaining the initial adaptive convolution kernel corresponding to each feature point on the image based on the center value of the initial adaptive convolution kernel corresponding to each feature pixel on the image.

[0046] More specifically, first calculate the gradient value of each pixel on the endoscopic image and construct a data sequence:

[0047]

[0048] In the formula, Indicates the first The gradient value of each pixel is . Represents the total number of pixels in the endoscope image.

[0049] Since blood vessels in endoscopic images are thinner, the gradient values ​​of pixels on blood vessels will be larger than those on other tissues in the throat. In addition, the number of pixels on blood vessels is relatively small, so we use the average gradient value of pixels on endoscopic images. Classification is performed, that is, in this implementation, the first gradient value threshold can be set to the average gradient value , the gradient value The pixel points are recorded as feature points on the endoscopic image, and the other pixel points are recorded as non-feature points on the endoscopic image.

[0050] Then, select a 4-adjacent convolution kernel and fix the center value to -1, as shown below:

[0051]

[0052] According to the difference between the gradient value of the feature point in the endoscopic image and the average gradient value of its four neighborhoods, the center value of the initial adaptive convolution kernel corresponding to the i-th feature pixel point on the endoscopic image is obtained. In this embodiment, the calculation formula for the center value of the initial adaptive convolution kernel corresponding to the i-th feature pixel point on the constructed image is as follows:

[0053]

[0054] In the formula, Represents the gradient value of the i-th feature pixel, Represents the average gradient value of the i-th feature pixel and its four neighborhoods. Indicates the function of rounding down the calculation result; represents the normalization function; Represents the center value of the initial adaptive convolution kernel corresponding to the i-th feature pixel.

[0055] In the calculation formula of the center value of the initial adaptive convolution kernel corresponding to the i-th feature pixel point on the above image, the calculation is mainly based on the difference between the gradient value of the feature point in the endoscopic image and the gradient value in the surrounding neighborhood. For the feature point located on the blood vessel in the endoscopic image, the larger the gradient difference between it and the surrounding pixel points, the more special the point is in the endoscopic image, and the convolution kernel makes it more prominent and the details clearer.

[0056] At this point, the initial adaptive convolution kernel corresponding to the i-th feature pixel is obtained as follows:

[0057]

[0058] Furthermore, according to the obtained The adaptive convolution kernel of the feature pixel is used to Laplace sharpening is performed on the feature pixel and the pixels in its four neighborhoods.

[0059] Step S3: Perform Laplace sharpening on the feature points, compare the grayscale value changes of the pixel point and its four neighborhoods before and after sharpening, and construct a sharpening evaluation of the feature points.

[0060] Specifically, according to the initial adaptive convolution kernel corresponding to each feature point on the image, Laplace sharpening is performed on the feature point; according to the grayscale value of the feature point and its neighborhood pixels before sharpening, the grayscale value of the feature point and its neighborhood pixels after sharpening, and the difference in grayscale values ​​between the feature point and each pixel in the four neighborhoods after sharpening, the sharpening effect of the feature point on the image is evaluated.

[0061] Since the size of the average grayscale value in an image reflects the amount of image content, that is, the larger the average grayscale value of an image, the more details it contains. Before and after image sharpening, the grayscale value of the details at the edge of the image changes significantly. Based on the grayscale value difference between the pixel before and after sharpening, the grayscale value difference between the sharpened pixel and the pixels in the four neighborhoods can be used to obtain the first The evaluation method for the sharpening effect of the i-th feature pixel point is as follows:

[0062]

[0063] In the formula, Indicates the number of The gray value of a feature point and its neighboring pixels; After sharpening The gray value of a feature point and its neighboring pixels; After sharpening, The feature points and the The gray value difference of each pixel; represents the normalization function; Represents the evaluation result of the sharpening effect of the i-th feature pixel on the image.

[0064] In the evaluation formula for the sharpening effect of the i-th feature pixel point on the image constructed above, the grayscale value change of the feature point in the endoscopic image before and after sharpening and the grayscale value difference between the feature point and the pixels in the surrounding neighborhood after sharpening are mainly used. For the endoscopic image, the greater the difference in the grayscale value of the pixel point before and after sharpening, the greater the sharpening intensity. And the greater the grayscale difference between the feature point and the pixels in the surrounding neighborhood after sharpening, it also indicates that the sharpening intensity of the endoscopic image is greater. That is, the difference between the grayscale value of a feature point and its neighborhood pixels before sharpening and the grayscale value of the feature point and its neighborhood pixels after sharpening, and the grayscale value difference between the feature point and each pixel in the four neighborhoods after sharpening are respectively positively correlated with the evaluation of the sharpening effect of the feature point on the image.

[0065] Step S4: According to the sharpening evaluation of the feature points, the best adaptive convolution kernel of the feature points is obtained.

[0066] Specifically, according to the evaluation results of the sharpening effect of each feature point on the image, the numerical range of the evaluation results belonging to the preset optimal result is set, and the center value of the convolution kernel that is not in the numerical range of the optimal result is sharpened or suppressed.

[0067] Since image sharpening is not the finer the better, when the image is sharpened to a certain extent, the image may lose details, increase noise, etc., which makes the image look unnatural. Insufficient image sharpening will cause the image to lack clarity and details, thereby affecting the visualization effect and quality of the image. Accordingly, in this embodiment, the size of the Laplace convolution kernel in the Laplace sharpening process of the pixel point is negatively fed back according to the sharpness evaluation of the obtained feature pixel point, and the size of the Laplace convolution kernel is continuously adjusted to make the image show the best sharpening effect.

[0068] In this embodiment, after obtaining the evaluation result of the sharpening effect of the i-th characteristic pixel point on the endoscopic image, the numerical range P of the evaluation result belonging to the best result is set to , further obtain a more suitable center value of the convolution kernel :

[0069]

[0070] In the formula, Represents the center value of the initial adaptive convolution kernel corresponding to the i-th feature pixel; Indicates the center value of the adaptive convolution kernel corresponding to the more suitable i-th feature pixel; Represents the evaluation result of the sharpening effect of the i-th feature pixel on the image.

[0071] In the formula, when the calculated If the value is small, it means that the sharpening effect of the endoscopic image is not very good. By increasing the center value of the convolution kernel, the characteristic pixels and their neighborhoods in the endoscopic image are re-sharpened. If the value is larger, it means that the sharpening effect of the endoscopic image is too detailed and needs to be suppressed to a certain extent. That is, the center value of the convolution kernel that is not in the numerical range of the best result is sharpened or suppressed. Specifically, if the evaluation result is lower than the lower limit of the numerical range of the best result, the center value of the convolution kernel is increased; if the evaluation result is higher than the upper limit of the numerical range of the best result, the center value of the convolution kernel is reduced.

[0072] So far, we have obtained The new convolution kernel size for feature pixels is:

[0073]

[0074] Furthermore, Laplace sharpening is performed again on the feature points on the image by increasing or decreasing the center value of the convolution kernel until the sharpening evaluation result of the corresponding feature point is within the numerical range of the preset optimal result. The new convolution kernel of the feature pixel is The characteristic pixel point returns to step S3 until the sharpening evaluation effect obtained falls within the range of P, thereby obtaining the first characteristic pixel point in the endoscope image. The optimal adaptive convolution kernel for feature pixels.

[0075] Step S5: Obtain the optimal adaptive convolution kernel of the non-feature point according to the distance between the non-feature point and the feature point in the image, the gray value difference, and the difference in the gradient values ​​of the surrounding four neighborhoods.

[0076] Specifically, the degree of difference between a non-feature point and its nearest feature point on the image is obtained based on the distance between it and the feature point closest to it, the difference in grayscale values ​​between the non-feature point and its nearest feature point on the image, and the average gradient difference between the non-feature point and the pixels of its four neighborhoods and the feature point closest to it and the pixels of its four neighborhoods; the center value of the best adaptive convolution kernel of the feature point closest to the non-feature point on the image is obtained based on the center value of the best adaptive convolution kernel of the non-feature point on the image and the degree of difference between the non-feature point on the image and its nearest feature point; the adaptive convolution kernel of the non-feature point on the image is obtained based on the center value of the best adaptive convolution kernel of the non-feature point on the image.

[0077] More specifically, first, the constructed The formula for calculating the difference between a non-feature pixel and its nearest feature point is as follows:

[0078]

[0079] In the formula, It indicates the first The distance between a non-feature pixel and its nearest feature point is obviously not 0; It indicates the first The gray value difference between a non-feature pixel and its nearest feature point; It indicates the first The average gradient difference between a non-feature pixel and its four neighboring pixels and its nearest feature point and its four neighboring pixels; Indicates The difference between a non-feature pixel and its nearest feature point.

[0080] In the above-mentioned construction In the calculation formula for the degree of difference between a non-feature pixel and its nearest feature point, the calculation is performed based on the distance between the non-feature pixel and the feature pixel on the endoscopic image, the grayscale value, and the gradient difference of the four neighborhoods. For the non-feature pixel on the endoscopic image, the closer the distance between it and the feature pixel on the endoscopic image, the greater the difference in grayscale value and gradient, indicating that the degree of difference between it and the feature pixel is greater. That is, the distance between a non-feature point on the image and its nearest feature point is negatively correlated with the degree of difference between the non-feature point and its nearest feature point, and the grayscale value difference between the non-feature point on the image and its nearest feature point, the average gradient difference between the non-feature point and its four neighborhood pixels, the nearest feature point, and its four neighborhood pixels is positively correlated with the degree of difference between the non-feature point and its nearest feature point.

[0081] Further, construct the first The center value of the optimal adaptive convolution kernel for non-feature pixels The calculation formula is as follows:

[0082]

[0083] In the formula, Indicates the distance on the endoscope image The center value of the best adaptive convolution kernel of the nearest feature pixel to the non-feature pixel; Indicates the first The degree of difference between a non-feature pixel and its nearest feature point; Indicates the function of rounding down the calculation result; represents the normalization function; Indicates the first The center value of the optimal adaptive convolution kernel for non-feature pixels.

[0084] According to the endoscopic image The center value of the optimal adaptive convolution kernel of non-feature pixels is obtained on the endoscopic image. The adaptive convolution kernel of non-feature pixels is as follows:

[0085]

[0086] According to the endoscopic image The adaptive convolution kernel of non-feature pixels is used to perform Laplace sharpening on the non-feature points. The same scheme as in step S3 is used to construct the sharpening evaluation of non-feature points until the first pixel on the endoscopic image is obtained. The optimal adaptive convolution kernel for non-feature pixels.

[0087] Further, the first The difference between a non-feature pixel and its nearest feature point is The best adaptive convolution kernels for non-feature pixels are recorded in Table 1:

[0088]

[0089] Table 1

[0090] It should be further explained that in judging the first In the process of finding the best adaptive convolution kernel for non-feature pixels, we obtain The difference between a non-feature pixel and its nearest feature point ,like In the above records, we can find the difference range, and directly obtain the best adaptive convolution kernel corresponding to this difference range as the first The optimal adaptive convolution kernel for non-feature pixels.

[0091] Step S6: Perform Laplace sharpening on the image according to the optimal adaptive convolution kernel of the feature points and non-feature points to obtain a sharpened image.

[0092] In this embodiment, the above process obtains the best adaptive convolution kernel of the characteristic pixel points and non-characteristic pixel points collected by the endoscope, that is, the best adaptive convolution kernel of each pixel point on the image is obtained. Laplace sharpening is performed on the image according to the best adaptive convolution kernel of the pixel points on the obtained endoscopic image to obtain a sharpened endoscopic image. The sharpened endoscopic image can be used as an image for doctors to evaluate the health status of patients.

[0093] The real-time enhancement method for endoscopic images for emergency use provided in this embodiment evaluates the sharpening effect of the adaptive convolution kernel on the feature pixel points on the endoscopic image, and negatively feeds back the convolution kernel size based on the evaluation result, so as to obtain the best adaptive convolution kernel for each pixel point on the image collected by the endoscope. The optimal adaptive convolution kernel for non-feature points is obtained by utilizing the difference in distance, grayscale value, etc. between non-feature points and feature points. Each pixel point on the image is sharpened according to the optimal adaptive convolution kernel on the obtained endoscopic image, so as to obtain the best sharpened image. The above scheme can improve the efficiency of endoscopic image enhancement used in emergency use, which is beneficial to reducing the amount of calculation of non-feature points during image sharpening processing, and is beneficial to saving computing power during endoscopic image enhancement.

[0094] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0095] The embodiments described above are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, a person skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A real-time enhancement method for endoscopic images for emergency treatment, characterized in that: The method comprises: Step S1: using an endoscope to collect images and pre-process the images; Step S2: Select pixel points in the image that are greater than a first gradient value threshold as feature points, and obtain an initial adaptive convolution kernel based on the difference between the gradient value of the feature point and the average gradient value of the feature point and its four neighborhoods; Step S3: Perform Laplace sharpening on the feature point, compare the grayscale value change difference between the pixel point and its four neighborhoods before and after sharpening, and construct a sharpening evaluation of the feature point; Step S4: obtaining the best adaptive convolution kernel of the feature point according to the sharpening evaluation of the feature point; Step S5: Obtain the best adaptive convolution kernel of the non-feature point according to the distance between the non-feature point and the feature point in the image, the gray value difference, and the difference in the gradient values ​​of the surrounding four neighborhoods; Step S6: performing Laplace sharpening on the image according to the best adaptive convolution kernel of the feature points and non-feature points to obtain a sharpened image; The step S2 is specifically as follows: Calculate the gradient value of each pixel on the image and the average gradient value of all pixels; Pixels with values ​​greater than the average gradient are recorded as feature points on the image, and other pixels are recorded as non-feature points on the image; According to the difference between the gradient value of the feature point on the image and the average gradient value of its four neighborhoods, the center value of the initial adaptive convolution kernel corresponding to each feature pixel on the image is obtained; According to the center value of the initial adaptive convolution kernel corresponding to each feature pixel point on the image, the initial adaptive convolution kernel corresponding to each feature point on the image is obtained; The step S4 is specifically as follows: According to the evaluation results of the sharpening effect of each feature point on the image, the numerical range of the evaluation results belonging to the preset optimal result is set, and the center value of the convolution kernel that is not in the numerical range of the optimal result is sharpened or suppressed; The sharpening enhancement or sharpening suppression of the center value of the convolution kernel that is not in the numerical range of the best result is specifically as follows: For evaluation results that are lower than the lower limit of the numerical range of the best result, increase the center value of the convolution kernel; For evaluation results that are higher than the upper limit of the numerical range of the best result, reduce the center value of the convolution kernel; The step S5 is specifically as follows: Obtain the degree of difference between a non-feature point and its nearest feature point on the image according to the distance between the non-feature point on the image and its nearest feature point, the gray value difference between the non-feature point on the image and its nearest feature point, and the average gradient difference between the non-feature point on the image and its four neighboring pixels and its nearest feature point and its four neighboring pixels; Obtaining the center value of the best adaptive convolution kernel of a non-feature point on the image according to the center value of the best adaptive convolution kernel of a feature point on the image that is closest to the non-feature point and the degree of difference between the non-feature point on the image and the feature point that is closest to the non-feature point on the image; According to the center value of the best adaptive convolution kernel of the non-feature point on the image, an adaptive convolution kernel of the non-feature point on the image is obtained.

2. The real-time enhancement method for emergency endoscopic images according to claim 1, characterized in that: In the step S1, an endoscope is used to collect images, and the image is preprocessed specifically as follows: an endoscope is used to collect a throat image of a patient, and the image is preprocessed including denoising and grayscale processing.

3. The real-time enhancement method for emergency endoscopic images according to claim 1, characterized in that: The step S3 is specifically as follows: According to the initial adaptive convolution kernel corresponding to each feature point on the image, Laplace sharpening is performed on the feature points; The sharpening effect of the feature point on the image is evaluated based on the grayscale value of the feature point and its neighboring pixels before sharpening, the grayscale value of the feature point and its neighboring pixels after sharpening, and the grayscale value difference between the feature point and each pixel in the four neighborhoods after sharpening.

4. The real-time enhancement method for emergency endoscopic images according to claim 3, characterized in that: The difference between the grayscale value of a feature point and its neighboring pixels before sharpening and the grayscale value of the feature point and its neighboring pixels after sharpening, and the difference in grayscale value between the feature point and each pixel in the four neighborhoods after sharpening are positively correlated with the evaluation of the sharpening effect of the feature point on the image.

5. The real-time enhancement method for emergency endoscopic images according to claim 1, characterized in that: By increasing or decreasing the center value of the convolution kernel, Laplace sharpening is performed again on the feature points on the image until the sharpening evaluation result of the corresponding feature point is within the numerical range of the preset optimal result.

6. The real-time enhancement method for emergency endoscopic images according to claim 1, characterized in that: The distance between a non-feature point on the image and its nearest feature point is negatively correlated with the degree of difference between the non-feature point and its nearest feature point, and the grayscale value difference between the non-feature point on the image and its nearest feature point, and the average gradient difference between the non-feature point and its four neighboring pixels, its nearest feature point and its four neighboring pixels are positively correlated with the degree of difference between the non-feature point and its nearest feature point.

Images