Eye burn recovery degree evaluation method and system based on image recognition

Abstract

The invention provides an eye burn recovery degree evaluation method and system based on image recognition, and the method comprises the steps: carrying out the preprocessing of a to-be-evaluated eye image, so as to obtain a processed eye image; performing color correction processing on the processed eye image to obtain a color correction image; performing texture enhancement and brightness adjustment processing on the color correction image to obtain a first adjustment image and a second adjustment image, and fusing the first adjustment image and the second adjustment image to obtain a fused image; performing tensor iteration de-noising processing on the fused image to obtain a de-noised eye image; and inputting the de-noised eye image into a pre-trained burn recovery degree evaluation model for evaluation so as to output a target burn recovery degree evaluation result, the image brightness is effectively improved, image color distortion is avoided, the problem of edge blurring is solved, detail information of the image is fully highlighted, and the image quality is improved. And the accuracy of subsequent model evaluation is improved.

Description

A method and system for assessing the degree of recovery from eye burns based on image recognition. Technical Field

[0001] This invention belongs to the technical field of image processing, and specifically relates to a method and system for assessing the degree of recovery from eye burns based on image recognition. Background Technology

[0002] Ocular burns are a common ophthalmic emergency, with a lengthy and complex recovery process involving the healing and repair of multiple structures, including the cornea, conjunctiva, and eyelids. Clinicians typically rely on regular follow-ups to observe changes in the affected eye's appearance, such as the degree of congestion, neovascularization, scar formation, and corneal transparency, to subjectively assess the stage and outcome of the burn's recovery. This method, dependent on visual observation and clinical experience, suffers from problems such as inconsistent assessment standards, strong subjectivity, difficulty in quantification, and potential discrepancies in judgment among different physicians. With the development of telemedicine and precision medicine, objective, repeatable, and standardized assessment methods have become urgently needed.

[0003] In recent years, significant progress has been made in using computer vision and image recognition technologies to assist in medical diagnosis. In the field of ophthalmology, studies have attempted to automatically screen and classify diseases by analyzing ocular images (such as slit-lamp photographs and anterior segment optical coherence tomography images). Applying such technologies to the assessment of ocular burn recovery can theoretically enable the automatic extraction, quantitative analysis, and trend tracking of recovery characteristics, thereby improving the objectivity and efficiency of the assessment.

[0004] However, directly applying image recognition technology to the assessment of eye burns faces a series of unique technical challenges. The core bottleneck lies in the quality of the input image, which severely restricts the accuracy of subsequent feature extraction and model analysis: Severe color distortion (overall reddish tint): After an eye burn, the local tissue is in a state of inflammation and congestion, with dilated blood vessels, resulting in a significant reddish hue in the affected area. When shooting under conventional light sources (such as built-in flash or ambient light), the camera's white balance algorithm may not be able to accurately correct this, easily causing color imbalance in the entire image, resulting in an overall or regional reddish tint. This color cast not only obscures subtle color differences at different recovery stages (such as the difference between bright red and dark red), but also severely interferes with color feature-based segmentation and classification algorithms, leading to misjudgments of key recovery indicators such as the degree of congestion and the maturity of neovascularization.

[0005] Blurred image details and unclear textures: Burn sites may be accompanied by tissue edema, secretions, early corneal opacity, or later scar formation, all of which inherently reduce the clarity of local images. Furthermore, limitations in shooting conditions (such as patient cooperation, eyelid opening and closing, handheld device shake, and uneven ambient lighting) further exacerbate image blurring, noise, and loss of texture details. Clear texture information is crucial for identifying the healing status of the corneal epithelium, the fine structure of scars, and vascular morphology. Unclear textures directly lead to difficulties in feature extraction, making it impossible to effectively distinguish the subtle differences between edema, healing tissue, and normal tissue.

[0006] Insufficient overall image brightness and contrast: The aforementioned color cast and blurring issues, coupled with suboptimal lighting conditions, often result in images with low overall brightness, narrow dynamic range, and poor contrast. The boundaries between important anatomical structures and pathological features and the background, as well as between tissues at different recovery stages, are blurred. Low-contrast images significantly reduce the effectiveness of traditional image segmentation algorithms (such as thresholding and edge detection), making it difficult to accurately segment burn areas, let alone further analyze their internal features.

[0007] Currently, while there are common methods such as histogram equalization, contrast stretching, and color correction in general image enhancement or some medical image preprocessing, they often fall short when dealing with scenarios like eye burn images, which have special pathological colors and complex degradation patterns: a single enhancement strategy may exacerbate noise or amplify color distortion; general white balance algorithms have difficulty distinguishing between pathological "red" and color cast caused by illumination; simple correction may lead to the loss of important pathological color information; and there is a lack of targeted joint optimization, failing to simultaneously and collaboratively solve the three interrelated problems of "color cast," "blur," and "low contrast."

[0008] Due to the lack of in-depth analysis and targeted processing of the degradation characteristics of ocular burn images, the image quality obtained by existing technical processes cannot meet the requirements of high-precision assessment models, and the robustness of image features is poor, ultimately resulting in low accuracy, reliability, and clinical acceptability of the recovery degree assessment results. Summary of the Invention

[0009] To address the aforementioned technical problems, this invention provides a method and system for assessing the degree of recovery from eye burns based on image recognition, thereby resolving the technical issues in the prior art.

[0010] In a first aspect, the present invention provides the following technical solution: a method for assessing the degree of recovery of eye burns based on image recognition, comprising: acquiring an image of the target eye to be assessed; preprocessing the image of the target eye to obtain a processed eye image; performing color correction processing on the processed eye image to obtain a color-corrected image; performing texture enhancement and brightness adjustment processing on the color-corrected image to obtain a first adjusted image and a second adjusted image respectively; fusing the first adjusted image and the second adjusted image to obtain a fused image; performing tensor iterative denoising processing on the fused image to obtain a denoised eye image; acquiring a pre-trained burn recovery degree assessment model; inputting the denoised eye image into the pre-trained burn recovery degree assessment model for evaluation, and outputting the burn recovery degree assessment result of the target.

[0011] Compared with existing technologies, the beneficial effects of this invention are as follows: First, this invention acquires an image of the target eye to be evaluated, preprocesses the image to obtain a processed eye image; then, it performs color correction processing on the processed eye image to obtain a color-corrected image; next, it performs texture enhancement and brightness adjustment processing on the color-corrected image to obtain a first adjusted image and a second adjusted image, respectively; the first adjusted image and the second adjusted image are then fused to obtain a fused image; next, it performs tensor iterative denoising processing on the fused image to obtain a denoised eye image; finally, it acquires a pre-trained burn recovery assessment model and applies the denoised eye image... The image is input into a pre-trained burn recovery assessment model for evaluation, and the output is the burn recovery assessment result of the target. This invention can effectively solve the problem of overall reddish color in the image by performing color correction, texture enhancement and brightness adjustment on the image. At the same time, it can also effectively improve the image brightness, avoid image color distortion and solve the problem of edge blurring. Then, tensor iterative denoising is performed on the image. Through the idea of ​​domain decomposition, the clear structure after the initial structure-guided filtering is used to further filter the removed noise, extract the signal loss in the noise, enhance the effect of structure-guided filtering, fully highlight the detailed information of the image and improve the accuracy of subsequent model evaluation.

[0012] Preferably, the step of performing color correction processing on the processed eye image to obtain a color-corrected image specifically includes: performing a first color correction processing on the red channel of the processed eye image to obtain a first corrected image. In the formula, The pixel value of the red channel of the first corrected image. These represent the pixel values ​​of the blue and red channels in the processed eye image, respectively. The blue and red channels of the first corrected image are used to calculate the mean values, respectively; the green channel of the first corrected image is then subjected to a second color correction process to obtain the second corrected image. In the formula, The pixel values ​​of the red channel in the second corrected image. These represent the pixel values ​​of the blue and green channels in the processed eye image, respectively. These are the mean values ​​of the blue and green channels of the processed eye image, respectively; for the second corrected image A third color correction process is performed to obtain a color-corrected image. : In the formula, It is a mapping function for nonlinear transformation.

[0013] Preferably, the step of performing texture enhancement and brightness adjustment processing on the color-corrected image to obtain a first adjusted image and a second adjusted image respectively includes: determining the initial differential order. The mask coefficient sequence is calculated based on the initial differential order. : ; , In the formula, The initial masking factor is... For the first A mask coefficient; selected from the mask coefficient sequence, excluding the initial mask coefficient. A mask coefficient, and based on the selected The first mask is constructed using the individual mask coefficients and the initial mask coefficient. With the second mask : , The first mask and the second mask are cross-producted to obtain a two-dimensional mask. The number of selected mask coefficients is adjusted and several two-dimensional masks of different scales are determined. The two-dimensional masks of different scales are convolved with the color correction image to obtain several convolution results. The several convolution results are fused at multiple scales to obtain a first adjusted image. The color correction image is converted from RGB space to HSV space to obtain a converted image. The converted image is brightness adjusted using a gamma correction method and converted back from HSV space to RGB space to obtain a second adjusted image.

[0014] Preferably, the step of fusing the first adjusted image and the second adjusted image to obtain a fused image specifically involves: performing multi-scale fusion of the first adjusted image and the second adjusted image using the Laplacian pyramid to obtain a fused image.

[0015] Preferably, the step of performing tensor iterative denoising on the fused image to obtain a denoised eye image includes: determining the x-gradient matrix of the fused image in the x and y directions, respectively. With the gradient matrix of y The gradient tensor set of each point is calculated based on the x-gradient matrix and the y-gradient matrix. : In the formula, , , The gradient tensors are the first, second, and third gradient tensors, respectively. Gaussian filtering is applied sequentially to the set of gradient tensors to obtain the target gradient tensor. Eigenvalue decomposition is then performed on the target gradient tensor to obtain the feature set. With the set of feature vectors ; Calculate the image confidence set based on the feature set. and image correlation coefficient set : ; ; ; In the formula, The confidence scores for the first, second, and third images are respectively. These are the correlation coefficients of the first, second, and third images, respectively. These are the first, second, and third feature values ​​in the feature set, respectively; based on the image confidence set... and image correlation coefficient set Determine the image diffusion tensor : ; ; ; In the formula, These are the eigenvalues ​​of the first, second, and third tensors, respectively. The initial feature values ​​are used; the fused image is iteratively updated based on the image diffusion tensor until the iteration stopping condition is met to obtain the first updated fused image. And the first processing noise: In the formula, The first , The updated fused image after the next iteration The iteration step size, For gradient operators, For the first Image diffusion tensor after the next iteration These are the target gradient tensors at small and large scales, respectively; the first updated fused image is used as the basis for this. To guide the iterative update of the first processed noise, the iteration stops repeatedly until the condition is met, so as to obtain a second updated fused image and a second processed noise; the first updated fused image and the second updated fused image are combined to obtain a denoised eye image.

[0016] Preferably, the burn recovery assessment model is the DenseNet model.

[0017] Secondly, the present invention provides the following technical solution: an image recognition-based system for assessing the degree of recovery of eye burns, the system comprising: a preprocessing module for acquiring an image of the target eye to be assessed, and preprocessing the image of the target eye to obtain a processed eye image; a correction module for performing color correction processing on the processed eye image to obtain a color-corrected image; an adjustment module for performing texture enhancement and brightness adjustment processing on the color-corrected image to obtain a first adjusted image and a second adjusted image respectively, and fusing the first adjusted image and the second adjusted image to obtain a fused image; a denoising module for performing tensor iterative denoising processing on the fused image to obtain a denoised eye image; and an evaluation module for acquiring a pre-trained burn recovery degree assessment model, inputting the denoised eye image into the pre-trained burn recovery degree assessment model for evaluation, and outputting the burn recovery degree assessment result of the target.

[0018] Preferably, the burn recovery assessment model is the DenseNet model.

[0019] Thirdly, the present invention provides the following technical solution: a computer, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described image recognition-based method for assessing the degree of recovery of eye burns.

[0020] Fourthly, the present invention provides the following technical solution: a storage medium storing a computer program, wherein when the computer program is executed by a processor, it implements the above-described image recognition-based method for assessing the degree of recovery of eye burns. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 is a flowchart of the image recognition-based method for assessing the degree of recovery of eye burns provided in Embodiment 1 of the present invention; Figure 2 is a structural block diagram of the image recognition-based system for assessing the degree of recovery of eye burns provided in Embodiment 2 of the present invention; Figure 3 is a schematic diagram of the hardware structure of a computer provided in another embodiment of the present invention.

[0023] The embodiments of the present invention will be further described below with reference to the accompanying drawings. Detailed Implementation

[0024] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain embodiments of the present invention, and should not be construed as limiting the present invention.

[0025] In Embodiment 1 of the present invention, as shown in Figure 1, a method for assessing the degree of recovery of eye burns based on image recognition includes: S1, acquiring an image of the target's eye to be assessed, and preprocessing the image of the eye to be assessed to obtain a processed eye image; specifically, the image of the target's eye to be assessed can be acquired by a shooting device. In fact, the image of the eye to be assessed can be captured from multiple angles to fully acquire image features. The preprocessing process specifically includes steps such as image translation and rotation, image size cropping, and format conversion in the prior art, which will not be elaborated here.

[0026] S2. Perform color correction processing on the processed eye image to obtain a color-corrected image; wherein, step S2 includes: S21. Perform a first color correction processing on the red channel of the processed eye image to obtain a first corrected image: In the formula, The pixel value of the red channel of the first corrected image. These represent the pixel values ​​of the blue and red channels in the processed eye image, respectively. These are the average values ​​of the blue and red channels for processing eye images, respectively. Specifically, for processed eye images obtained from actual shooting, due to the burn area, shooting environment, and skin tone, the overall image will have a weak red tint. Therefore, by color correction of the red and green channels, the overall brightness and contrast of the image are further improved to achieve effective restoration of image colors and solve the image color cast problem.

[0027] S22. Perform a second color correction process on the green channel of the first corrected image to obtain a second corrected image: In the formula, The pixel values ​​of the red channel in the second corrected image. These represent the pixel values ​​of the blue and green channels in the processed eye image, respectively. S23, process the blue and green channels of the eye image, respectively; S24, process the second corrected image. A third color correction process is performed to obtain a color-corrected image. : In the formula, This is a mapping function for nonlinear transformation; specifically, after correction, the white balance needs to be corrected again. The purpose of white balance is to eliminate color deviation caused by low illumination. However, since the color deviation caused by scene lighting is nonlinear, the white balance can be effectively corrected by using a nonlinear color mapping function.

[0028] S3. Perform texture enhancement and brightness adjustment processing on the color-corrected image to obtain a first adjusted image and a second adjusted image respectively. Then, fuse the first adjusted image and the second adjusted image to obtain a fused image. Step S3 includes: S31. Determine the initial differential order. The mask coefficient sequence is calculated based on the initial differential order. : ; , In the formula, The initial masking factor is... For the first S32, Select mask coefficients from the mask coefficient sequence, excluding the initial mask coefficients. A mask coefficient, and based on the selected The first mask is constructed using the individual mask coefficients and the initial mask coefficient. With the second mask : , S33. Perform an outer product calculation on the first mask and the second mask to obtain a two-dimensional mask. Adjust the number of selected mask coefficients and determine several two-dimensional masks of different scales. Perform convolution processing on the two-dimensional masks of different scales with the color correction image to obtain several convolution results. Perform multi-scale fusion on the several convolution results to obtain a first adjusted image. Specifically, the two-dimensional mask in this application is generally 3×3, 5×5, 7×7, or 9×9 in size. Then, multi-scale fusion is performed to make the image data that was originally difficult to use due to low contrast and blurred details usable, and to extract more valuable information from the data.

[0029] Specifically, in general, first-order and second-order differential algorithms are used to enhance images, such as Gaussian filtering and Laplacian sharpening. While these algorithms have achieved certain results in enhancing the edge information of images, they perform poorly in processing the texture details of images. These algorithms often suffer from blurring and distortion when processing texture details, resulting in insufficient display of image details and easy introduction of noise. However, the fractional derivative method used in this application processes the image with an initial mask coefficient of 0.5. A larger initial coefficient results in a significant enhancement effect, emphasizing high-frequency information, while a smaller coefficient results in a milder enhancement effect, emphasizing low-frequency information. Therefore, this application sets the initial coefficient to 0.5, balancing both low- and high-frequency information. This method has non-linear amplitude-frequency characteristics, enabling it to adaptively adjust the enhancement intensity according to the signal frequency, which is more in line with the human visual system and the natural characteristics of images. It effectively overcomes the blurring and distortion problems that traditional first-order (such as the Sobel operator) and second-order (such as the Laplacian operator) derivative methods tend to encounter when processing complex textures. It can significantly improve the clarity and contrast of subtle textures, weak edges, and complex patterns in images, allowing previously inconspicuous details to be fully displayed.

[0030] S34. Convert the color-corrected image from RGB space to HSV space to obtain a converted image. Use gamma correction method to adjust the brightness of the converted image and convert it back from HSV space to RGB space to obtain a second adjusted image.

[0031] The step of fusing the first adjusted image and the second adjusted image to obtain a fused image specifically involves: performing multi-scale fusion of the first adjusted image and the second adjusted image using the Laplacian pyramid to obtain a fused image.

[0032] S4. Perform tensor iterative denoising processing on the fused image to obtain a denoised eye image; wherein, step S4 includes: S41. Determine the x-gradient matrix of the fused image in the x and y directions respectively. With the gradient matrix of y The gradient tensor set of each point is calculated based on the x-gradient matrix and the y-gradient matrix. : In the formula, , , These are the first, second, and third gradient tensors, respectively; S42, Gaussian filtering is applied sequentially to the set of gradient tensors to obtain the target gradient tensor, and eigenvalue decomposition is performed on the target gradient tensor to obtain the feature set. With the set of feature vectors S43. Calculate the image confidence set based on the feature set. and image correlation coefficient set : ; ; ; In the formula, The confidence scores for the first, second, and third images are respectively. These are the correlation coefficients of the first, second, and third images, respectively. These are the first, second, and third feature values ​​in the feature set, respectively; S44, based on the image confidence set... and image correlation coefficient set Determine the image diffusion tensor : ; ; ; In the formula, These are the eigenvalues ​​of the first, second, and third tensors, respectively. These are the initial eigenvalues; specifically, for three eigenvalues, if If all three images are approximately equal to 0, then the confidence scores of all three images are approximately 0. Approximately and all approximately 0, and Greater than If the confidence level of the first image is approximately 1, and the confidence levels of the others are approximately 0, then... Since the relationship is decreasing, the confidence levels of the second and third images are approximately 1, while the confidence level of the first image is approximately 0.

[0033] S45. Iteratively update the fused image based on the image diffusion tensor until the iteration stopping condition is met to obtain the first updated fused image. And the first processing noise: In the formula, The first , The updated fused image after the next iteration The iteration step size, For gradient operators, For the first Image diffusion tensor after the next iteration These are the target gradient tensors at small and large scales, respectively; S46, update the fused image using the first tensor. To guide the repeated iterative updates of the first processed noise until the iteration stop condition is met, a second updated fused image and a second processed noise are obtained; S47, the first updated fused image and the second updated fused image are combined to obtain a denoised eye image.

[0034] Specifically, for the above method, based on the diffusion tensor generated by the original structure-guided filtering, the eigenvalues ​​are modified. First, structure-guided filtering is used to obtain the main structural information of the image data, and the generated diffusion tensor is used for domain decomposition. This allows detailed information to be preserved more accurately without blurring or loss due to the use of structure-guided filtering alone. Second, the domain decomposition method is combined with the structure-guided filtering method. First, structure-guided filtering is used to decompose the data into a first updated fused image and a first processed noise. The structure-guided filtering method uses its own structure for filtering, while the domain decomposition method uses a clear structure to filter the noise to extract the signal loss of the noise. That is, the structure tensor of the first updated fused image is used to filter the first processed noise, decomposing the first processed noise into a second updated fused image and a second processed noise. Finally, the first updated fused image and the second updated fused image are combined to obtain the result.

[0035] S5. Obtain a pre-trained burn recovery assessment model, input the denoised eye image into the pre-trained burn recovery assessment model for evaluation, and output the burn recovery assessment result of the target.

[0036] Specifically, the burn recovery assessment model is a DenseNet model. By inputting the image into the pre-trained DenseNet model and then processing it, the model can output the burn recovery assessment result of the target.

[0037] The first embodiment of this invention provides a method for assessing the degree of recovery of eye burns based on image recognition. This invention first acquires an image of the target eye to be assessed, preprocesses the image to obtain a processed eye image, then performs color correction processing on the processed eye image to obtain a color-corrected image, then performs texture enhancement and brightness adjustment processing on the color-corrected image to obtain a first adjusted image and a second adjusted image, respectively, and then fuses the first adjusted image and the second adjusted image to obtain a fused image, then performs tensor iterative denoising processing on the fused image to obtain a denoised eye image, and finally obtains a pre-trained burn recovery degree assessment model. The denoised eye image is input into a pre-trained burn recovery assessment model for evaluation, outputting the burn recovery assessment result of the target. This invention effectively solves the problem of overall reddishness in the image by performing color correction, texture enhancement, and brightness adjustment on the image. It can also effectively improve image brightness, avoid image color distortion, and solve the problem of edge blurring. Then, tensor iterative denoising is performed on the image. Through the idea of ​​domain decomposition, the clear structure after the initial structure-guided filtering is used to further filter the removed noise, extract the signal loss in the noise, enhance the effect of structure-guided filtering, fully highlight the detailed information of the image, and improve the accuracy of subsequent model evaluation.

[0038] As shown in Figure 2, Embodiment 2 of the present invention provides an image recognition-based system for assessing the degree of eye burn recovery. The system includes: a preprocessing module 1, used to acquire an image of the target eye to be assessed and preprocess the image to obtain a processed eye image; a correction module 2, used to perform color correction processing on the processed eye image to obtain a color-corrected image; an adjustment module 3, used to perform texture enhancement and brightness adjustment processing on the color-corrected image to obtain a first adjusted image and a second adjusted image respectively, and to fuse the first adjusted image and the second adjusted image to obtain a fused image; a denoising module 4, used to perform tensor iterative denoising processing on the fused image to obtain a denoised eye image; and an evaluation module 5, used to acquire a pre-trained burn recovery degree assessment model, input the denoised eye image into the pre-trained burn recovery degree assessment model for evaluation, and output the burn recovery degree assessment result of the target.

[0039] The correction module 2 is used to: perform a first color correction process on the red channel of the processed eye image to obtain a first corrected image. In the formula, The pixel value of the red channel of the first corrected image. These represent the pixel values ​​of the blue and red channels in the processed eye image, respectively. The blue and red channels of the first corrected image are used to calculate the mean values, respectively; the green channel of the first corrected image is then subjected to a second color correction process to obtain the second corrected image. In the formula, The pixel values ​​of the red channel in the second corrected image. These represent the pixel values ​​of the blue and green channels in the processed eye image, respectively. These are the mean values ​​of the blue and green channels of the processed eye image, respectively; for the second corrected image A third color correction process is performed to obtain a color-corrected image. : In the formula, It is a mapping function for nonlinear transformation.

[0040] Among them, adjustment module 3 is used to: determine the initial differential order. The mask coefficient sequence is calculated based on the initial differential order. : ; , In the formula, The initial masking factor is... For the first A mask coefficient; selected from the mask coefficient sequence, excluding the initial mask coefficient. A mask coefficient, and based on the selected The first mask is constructed using the individual mask coefficients and the initial mask coefficient. With the second mask : , The first mask and the second mask are cross-producted to obtain a two-dimensional mask. The number of selected mask coefficients is adjusted and several two-dimensional masks of different scales are determined. The two-dimensional masks of different scales are convolved with the color correction image to obtain several convolution results. The several convolution results are fused at multiple scales to obtain a first adjusted image. The color correction image is converted from RGB space to HSV space to obtain a converted image. The converted image is brightness adjusted using a gamma correction method and converted back from HSV space to RGB space to obtain a second adjusted image.

[0041] The adjustment module 3 is further configured to: perform multi-scale fusion of the first adjusted image and the second adjusted image using the Laplacian pyramid to obtain a fused image.

[0042] The denoising module 4 is used to: determine the x-gradient matrix of the fused image in the x and y directions, respectively. With the gradient matrix of y The gradient tensor set of each point is calculated based on the x-gradient matrix and the y-gradient matrix. : In the formula, , , The gradient tensors are the first, second, and third gradient tensors, respectively. Gaussian filtering is applied sequentially to the set of gradient tensors to obtain the target gradient tensor. Eigenvalue decomposition is then performed on the target gradient tensor to obtain the feature set. With the set of feature vectors ; Calculate the image confidence set based on the feature set. and image correlation coefficient set : ; ; ; In the formula, The confidence scores for the first, second, and third images are respectively. These are the correlation coefficients of the first, second, and third images, respectively. These are the first, second, and third feature values ​​in the feature set, respectively; based on the image confidence set... and image correlation coefficient set Determine the image diffusion tensor : ; ; ; In the formula, These are the eigenvalues ​​of the first, second, and third tensors, respectively. The initial feature values ​​are used; the fused image is iteratively updated based on the image diffusion tensor until the iteration stopping condition is met to obtain the first updated fused image. And the first processing noise: In the formula, The first , The updated fused image after the next iteration The iteration step size, For gradient operators, For the first Image diffusion tensor after the next iteration These are the target gradient tensors at small and large scales, respectively; the first updated fused image is used as the basis for this. To guide the iterative update of the first processed noise, the iteration stops repeatedly until the condition is met, so as to obtain a second updated fused image and a second processed noise; the first updated fused image and the second updated fused image are combined to obtain a denoised eye image.

[0043] Specifically, the burn recovery assessment model is the DenseNet model.

[0044] In other embodiments of the present invention, the present invention provides the following technical solution: a computer, including a memory 102, a processor 101, and a computer program stored in the memory 102 and executable on the processor 101, wherein the processor 101 executes the computer program to implement the image recognition-based method for assessing the degree of recovery of eye burns as described above.

[0045] Specifically, the processor 101 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of the present invention.

[0046] The memory 102 may include a large-capacity memory for data or instructions. For example, and not limitingly, the memory 102 may include a hard disk drive (HDD), a floppy disk drive, a solid-state drive (SSD), flash memory, an optical disk drive, a magneto-optical disk drive, magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, the memory 102 may include removable or non-removable (or fixed) media. Where appropriate, the memory 102 may be internal or external to a data processing device. In a particular embodiment, the memory 102 is non-volatile memory. In a particular embodiment, the memory 102 includes read-only memory (ROM) and random access memory (RAM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable read-only memory (PROM), an erasable read-only memory (EPROM), an electrically erasable read-only memory (EEPROM), an electrically alterable read-only memory (EAROM), or flash memory, or a combination of two or more of these. Where appropriate, the RAM can be Static Random-Access Memory (SRAM) or Dynamic Random-Access Memory (DRAM). DRAM can be Fast Page Mode Dynamic Random Access Memory (FPMDRAM), Extended Data Out Dynamic Random Access Memory (EDODRAM), Synchronous Dynamic Random-Access Memory (SDRAM), etc.

[0047] The memory 102 can be used to store or cache various data files that need to be processed and / or used for communication, as well as possible computer program instructions executed by the processor 101.

[0048] The processor 101 reads and executes the computer program instructions stored in the memory 102 to implement the above-mentioned image recognition-based method for assessing the degree of recovery of eye burns.

[0049] In some embodiments, the computer may further include a communication interface 103 and a bus 100. As shown in FIG3, the processor 101, memory 102, and communication interface 103 are connected through the bus 100 and communicate with each other.

[0050] The communication interface 103 is used to enable communication between the various modules, devices, units, and / or equipment in the embodiments of the present invention. The communication interface 103 can also enable data communication with other components such as external devices, image / data acquisition devices, databases, external storage, and image / data processing workstations.

[0051] Bus 100 includes hardware, software, or both, that couples components of a computer device together. Bus 100 includes, but is not limited to, at least one of the following: data bus, address bus, control bus, expansion bus, and local bus. For example, and not as a limitation, bus 100 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a memory bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses, or a combination of two or more of these. Where appropriate, bus 100 may include one or more buses. Although specific buses are described and illustrated in the embodiments of the present invention, the present invention is contemplated by any suitable bus or interconnect.

[0052] The computer can use the image recognition-based eye burn recovery assessment system to execute the image recognition-based eye burn recovery assessment method of the present invention, thereby realizing the image recognition-based eye burn recovery assessment.

[0053] In some further embodiments of the present invention, in conjunction with the above-described image recognition-based method for assessing the degree of recovery of eye burns, the present invention provides the following technical solution: a storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the above-described image recognition-based method for assessing the degree of recovery of eye burns.

[0054] Those skilled in the art will understand that the logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can mean any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0055] More specific examples of readable media (a non-exhaustive list) include: electrical connections (electronic devices) with one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0056] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0057] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0058] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A method for assessing the degree of recovery from eye burns based on image recognition, characterized in that, include: Obtain the target's eye image to be evaluated, and preprocess the eye image to obtain a processed eye image; The processed eye image is then subjected to color correction processing to obtain a color-corrected image; The color-corrected image is subjected to texture enhancement and brightness adjustment processing to obtain a first adjusted image and a second adjusted image, respectively. The first adjusted image and the second adjusted image are then fused to obtain a fused image. Tensor iteration denoising processing is performed on the fused image to obtain a denoised eye image. A pre-trained burn recovery degree assessment model is obtained, and the denoised eye image is input into the pre-trained burn recovery degree assessment model for evaluation to output the burn recovery degree assessment result of the target.

2. The method for assessing the degree of recovery from eye burns based on image recognition according to claim 1, characterized in that, The step of performing color correction processing on the processed eye image to obtain a color-corrected image specifically includes: performing a first color correction processing on the red channel of the processed eye image to obtain a first corrected image. In the formula, The pixel value of the red channel of the first corrected image. These represent the pixel values ​​of the blue and red channels in the processed eye image, respectively. The blue and red channels of the first corrected image are used to calculate the mean values, respectively; the green channel of the first corrected image is then subjected to a second color correction process to obtain the second corrected image. In the formula, The pixel values ​​of the red channel in the second corrected image. These represent the pixel values ​​of the blue and green channels in the processed eye image, respectively. These are the mean values ​​of the blue and green channels of the processed eye image, respectively; for the second corrected image A third color correction process is performed to obtain a color-corrected image. ： In the formula, It is a mapping function for nonlinear transformation.

3. The method for assessing the degree of recovery from eye burns based on image recognition according to claim 1, characterized in that, The step of performing texture enhancement and brightness adjustment processing on the color-corrected image to obtain a first adjusted image and a second adjusted image respectively includes: determining the initial differential order. The mask coefficient sequence is calculated based on the initial differential order. ： ； ， In the formula, The initial masking factor is... For the first A mask coefficient; selected from the mask coefficient sequence, excluding the initial mask coefficient. A mask coefficient, and based on the selected The first mask is constructed using the individual mask coefficients and the initial mask coefficient. With the second mask ： ， The first mask and the second mask are cross-producted to obtain a two-dimensional mask. The number of selected mask coefficients is adjusted and several two-dimensional masks of different scales are determined. The two-dimensional masks of different scales are convolved with the color correction image to obtain several convolution results. The several convolution results are fused at multiple scales to obtain a first adjusted image. The color correction image is converted from RGB space to HSV space to obtain a converted image. The converted image is brightness adjusted using a gamma correction method and converted back from HSV space to RGB space to obtain a second adjusted image.

4. The method for assessing the degree of recovery from eye burns based on image recognition according to claim 1, characterized in that, The step of fusing the first adjusted image and the second adjusted image to obtain a fused image specifically involves: performing multi-scale fusion of the first adjusted image and the second adjusted image using the Laplacian pyramid to obtain a fused image.

5. The method for assessing the degree of recovery from ocular burns based on image recognition according to claim 1, characterized in that, The step of performing tensor iterative denoising on the fused image to obtain a denoised eye image includes: determining the x-gradient matrix of the fused image in the x and y directions, respectively. With the gradient matrix of y The gradient tensor set of each point is calculated based on the x-gradient matrix and the y-gradient matrix. ： In the formula, 、 、 The gradient tensors are the first, second, and third gradient tensors, respectively. Gaussian filtering is applied sequentially to the set of gradient tensors to obtain the target gradient tensor. Eigenvalue decomposition is then performed on the target gradient tensor to obtain the feature set. With the set of feature vectors ; Calculate the image confidence set based on the feature set. and image correlation coefficient set ： ； ； ； In the formula, The confidence scores for the first, second, and third images are respectively. These are the correlation coefficients of the first, second, and third images, respectively. These are the first, second, and third feature values ​​in the feature set, respectively; based on the image confidence set... and image correlation coefficient set Determine the image diffusion tensor ： ； ； ； In the formula, These are the eigenvalues ​​of the first, second, and third tensors, respectively. The initial feature values ​​are used; the fused image is iteratively updated based on the image diffusion tensor until the iteration stopping condition is met to obtain the first updated fused image. And the first processing noise: In the formula, The first 、 The updated fused image after the next iteration The iteration step size, For gradient operators, For the first Image diffusion tensor after the next iteration These are the target gradient tensors at small and large scales, respectively; the first updated fused image is used as the basis for this. To guide the iterative update of the first processed noise, the iteration stops repeatedly until the condition is met, so as to obtain a second updated fused image and a second processed noise; the first updated fused image and the second updated fused image are combined to obtain a denoised eye image.

6. The method for assessing the degree of recovery from ocular burns based on image recognition according to claim 1, characterized in that, The burn recovery assessment model is specifically the DenseNet model.

7. A system for assessing the degree of recovery from eye burns based on image recognition, characterized in that, The system includes: a preprocessing module for acquiring an image of the target's eye to be evaluated, and preprocessing the image to obtain a processed eye image; a correction module for performing color correction processing on the processed eye image to obtain a color-corrected image; an adjustment module for performing texture enhancement and brightness adjustment processing on the color-corrected image to obtain a first adjusted image and a second adjusted image, and fusing the first adjusted image and the second adjusted image to obtain a fused image; a denoising module for performing tensor iterative denoising processing on the fused image to obtain a denoised eye image; and an evaluation module for acquiring a pre-trained burn recovery degree evaluation model, inputting the denoised eye image into the pre-trained burn recovery degree evaluation model for evaluation, and outputting the burn recovery degree evaluation result of the target.

8. The image recognition-based eye burn recovery assessment system according to claim 7, characterized in that, The burn recovery assessment model is specifically the DenseNet model.

9. A computer comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the image recognition-based method for assessing the degree of recovery from eye burns as described in any one of claims 1 to 6.

10. A storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the image recognition-based method for assessing the degree of recovery from eye burns as described in any one of claims 1 to 6.

Images