Low-illumination physical examination image noise reduction enhancement method and device, equipment and storage medium

Abstract

The invention relates to the technical field of artificial intelligence, is suitable for financial and medical scenes, and particularly relates to a low-illumination physical examination image noise reduction enhancement method, device and equipment and a storage medium, the low-illumination physical examination image noise reduction enhancement method comprises the following steps: obtaining a to-be-processed low-illumination physical examination image; based on a Retinex decomposition model, acquiring a preliminarily estimated illumination component of the low-illumination physical examination image, acquiring an accurately estimated illumination image according to the preliminarily estimated illumination component, and enhancing the accurately estimated illumination image through gamma transformation to obtain an enhanced image; performing noise reduction on the enhanced image based on a BM3D algorithm to obtain a noise-reduced image; and combining the enhanced image and the noise reduction image to obtain a noise reduction enhanced image. The problem that in the prior art, when enhancement processing is carried out on the low-illumination physical examination image through an image enhancement algorithm, synchronous processing is not carried out on noise and amplified noise in the enhanced physical examination image is solved, and the accuracy and efficiency of physical examination image recognition are greatly improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology and is applicable to financial and medical scenarios, and in particular to a method, device, equipment and storage medium for noise reduction and enhancement of low-light physical examination images. Background Art

[0002] In some specific application scenarios in the financial and medical industries, such as insurance underwriting and telemedicine, users are required to upload health information such as medical images. However, traditional methods often rely on manual evaluation or analysis of uploaded images, which is not only time-consuming and labor-intensive, but also prone to human error. With the development of technology, insurance companies and medical institutions have begun to adopt intelligent medical image processing methods to automatically process user-uploaded medical images, providing efficient and accurate data support for insurance underwriting and medical decision-making, thereby improving efficiency and accuracy. However, the quality of user-uploaded medical images varies greatly, greatly affecting recognition results. Among them, low illumination causing the medical images to be too dark is a typical defect. When existing technologies use image enhancement algorithms to enhance the brightness and contrast of low-illuminance medical images, they do not simultaneously process the noise in the low-illuminance medical images and the amplified noise in the enhanced medical images, resulting in low medical image recognition accuracy and poor efficiency. Summary of the Invention

[0003] The present invention provides a method, device, equipment and storage medium for noise reduction and enhancement of low-illumination physical examination images, so as to solve the technical problems of low accuracy and poor efficiency of physical examination image recognition in existing methods.

[0004] In a first aspect, a method for noise reduction and enhancement of a low-light physical examination image is provided, the method comprising:

[0005] Acquiring a low-light physical examination image to be processed;

[0006] Obtaining a preliminary estimated illumination component of the low-illumination physical examination image based on a Retinex decomposition model, obtaining a precisely estimated illumination image based on the preliminary estimated illumination component, and enhancing the precisely estimated illumination image through a gamma transform to obtain an enhanced image;

[0007] An image acquisition module, used for acquiring low-light physical examination images to be processed;

[0008] an image enhancement module, configured to obtain a preliminary estimated illumination component of the low-light physical examination image based on a Retinex decomposition model, obtain a precisely estimated illumination image based on the preliminary estimated illumination component, and enhance the precisely estimated illumination image by gamma transform to obtain an enhanced image;

[0009] An image denoising module, configured to denoise the enhanced image based on a BM3D algorithm to obtain a denoised image;

[0010] An image merging module is used to merge the enhanced image with the denoised image to obtain a denoised enhanced image.

[0011] In a third aspect, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the above-mentioned low-illumination physical examination image noise reduction and enhancement method are implemented.

[0012] In a fourth aspect, a computer-readable storage medium is provided, which stores a computer program. When the computer program is executed by a processor, the steps of the above-mentioned low-illumination physical examination image noise reduction and enhancement method are implemented.

[0013] In the scheme implemented by the aforementioned low-light medical examination image noise reduction and enhancement method, apparatus, device, and storage medium, the low-light medical examination image to be processed is first enhanced based on the Retinex decomposition model and gamma shift to obtain an enhanced image. The enhanced image is then subjected to noise reduction based on the BM3D algorithm to obtain a reduced-noise image. Finally, the enhanced image and the reduced-noise image are merged to obtain a reduced-noise enhanced image. This solves the problem in the prior art of performing low-light medical examination image enhancement using an image enhancement algorithm, which fails to synchronously process the noise and the amplified noise in the enhanced medical examination image, thereby greatly improving the accuracy and efficiency of medical examination image recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments of the present invention. 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 creative work.

[0015] Figure 1 2 is a schematic diagram of an application environment of a low-illumination physical examination image noise reduction and enhancement method provided by an embodiment of the present invention.

[0016] Figure 2 The figure is a flow chart of an implementation method of a low-illumination physical examination image noise reduction and enhancement method provided by an embodiment of the present invention.

[0017] Figure 3 yes Figure 2 Schematic diagram of a specific implementation of step S20.

[0018] Figure 4 yes Figure 2FIG. 1 is a flow chart of another specific implementation of step S20.

[0019] Figure 5 yes Figure 2 FIG. 5 is a flow chart of another specific implementation of step S20.

[0020] Figure 6 yes Figure 2 A flowchart of a specific implementation of step S30 is shown in FIG.

[0021] Figure 7 yes Figure 6 A flowchart of a specific implementation of step S32 is shown in FIG.

[0022] Figure 8 yes Figure 2 A flowchart of a specific implementation of step S40 is shown in FIG.

[0023] Figure 9 This is a schematic diagram of a low-light medical examination image noise reduction and enhancement device according to an embodiment of the present invention.

[0024] Figure 10 It is a structural diagram of a computer device in one embodiment of the present invention. DETAILED DESCRIPTION

[0025] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0026] The low-light physical examination image noise reduction and enhancement method provided by the embodiment of the present invention can be applied in the following fields: Figure 1 In the application environment, Figure 1 A schematic diagram of an application environment of a low-light medical examination image noise reduction and enhancement method provided by an embodiment of the present invention is shown as follows: Figure 1 As shown, in this application environment, a terminal 10 and a server 20 are included. The terminal 10 is a terminal device used to upload low-light physical examination images, and the server 20 is a processing end for performing noise reduction and enhancement on low-light physical examination images. The terminal 10 and the server 20 jointly execute the low-light physical examination image noise reduction and enhancement method. It should be noted that the terminal 10 and the server 20 can be smart phones, tablet computers, laptop computers, desktop computers, etc., but are not limited to these. The terminal 10 and the server 20 can be connected via Bluetooth, USB (Universal Serial Bus) or other communication connection methods, and the present invention is not limited thereto.

[0027] Figure 2 A flowchart of an implementation method of a low-light physical examination image noise reduction and enhancement method provided by an embodiment of the present invention is shown as follows: Figure 2 As shown, the low-light physical examination image noise reduction and enhancement method may specifically include the following steps:

[0028] S10: Acquire a low-light physical examination image to be processed.

[0029] In this step, the low-light health examination images to be processed are the health information uploaded by users during the underwriting process, including medical examination forms, special examination reports, medical records, and diagnosis certificates. This information is the core basis for insurance companies to assess the health risks of policyholders, so the clarity of this information greatly affects the accuracy of the underwriting process. However, when users upload these images, the environment in which they are shooting is often not optimal, so the image illumination is often too low.

[0030] S20. Obtain a preliminary estimated illumination component of the low-illumination physical examination image based on a Retinex decomposition model, obtain a precisely estimated illumination image according to the preliminary estimated illumination component, and enhance the precisely estimated illumination image by gamma transform to obtain an enhanced image.

[0031] In this step, the Retinex decomposition model assumes that the image consists of an illumination component (Illumination) and a reflectance component (Reflectance), that is, I = L × R, where I is the original image, which is a low-light physical examination image in this embodiment, L is the illumination component, and R is the reflectance component. Based on the Retinex decomposition model, a preliminary estimate of the illumination component of the low-light physical examination image is obtained, that is, the initial illumination component is extracted from the low-light physical examination image. As the basis for subsequent optimization, the method commonly used here is the RGB channel maximum method, which takes the maximum value of the R, G, and B channel values ​​of each pixel to construct a preliminary illumination component. Assuming that the illumination is not blocked in at least one color channel, its maximum value approximately represents the illumination intensity, but the result may contain noise and block artifacts and needs further optimization. Based on the preliminary estimated illumination component, an accurate estimate of the illumination image is obtained to optimize the initial illumination component so that it conforms to the low-frequency and smooth characteristics of the illumination. The illumination image is enhanced by gamma transformation, the brightness distribution of the illumination component is adjusted, and dark details are enhanced or highlight overexposure is suppressed.

[0032] S30 , performing noise reduction on the enhanced image based on the BM3D algorithm to obtain a noise-reduced image.

[0033] In this step, BM3D (3D Block Matching Filter) is a denoising method based on image self-similarity. Its core concepts include: block matching, which divides the image into multiple small blocks and globally searches for blocks with similar structures (similar blocks); joint filtering, which stacks similar blocks into a three-dimensional array, disperses noise to high-frequency areas through a three-dimensional transform (such as DCT or wavelet transform), and then suppresses noise through threshold shrinkage; and aggregate reconstruction, which weights and fuses the filtered blocks back into the original image, preserving detail while removing noise.

[0034] S40: Merge the enhanced image and the denoised image to obtain a denoised enhanced image.

[0035] In this step, the enhanced image and the denoised image are merged to obtain a denoised enhanced image. This can combine the contrast / details of the enhanced image with the clarity of the denoised image, while avoiding blurred details or residual noise caused by simple mixing, thereby improving the quality of the core image and the accuracy of subsequent image recognition.

[0036] Table 1 is a quantitative comparison of the image enhancement performance of different algorithms. The implementation scheme provided by the present invention has better overall performance than the commonly used image enhancement algorithms.

[0037] Table 1 Quantitative comparison of image enhancement performance of different algorithms

[0038]

[0039] The low-light medical examination image noise reduction and enhancement method of this embodiment first enhances the low-light medical examination image to be processed based on the Retinex decomposition model and gamma shift to obtain an enhanced image. Then, the enhanced image is subjected to noise reduction processing based on the BM3D algorithm to obtain a reduced-noise image. Finally, the enhanced image and the reduced-noise image are merged to obtain a reduced-noise enhanced image. This method solves the problem of prior art that, when enhancing low-light medical examination images using image enhancement algorithms, fails to synchronously process the noise and the amplified noise in the enhanced medical examination image, thereby greatly improving the accuracy and efficiency of medical examination image recognition.

[0040] In one specific embodiment, when user 1 is making a claim for health-related insurance (such as critical illness insurance, medical insurance, and life insurance), they need to upload health information such as physical examination images to verify the authenticity of their health declaration, assess whether the claim application complies with the terms and conditions, and prevent insurance fraud risks. However, when the user uploads the image, the uploaded physical examination image may have low illumination due to issues with the equipment, environment, and shooting angle. In this case, a low-illumination physical examination image A uploaded by user 1 is first obtained. Then, a preliminary estimate of the illumination components of low-illumination physical examination image A is obtained based on the Retinex decomposition model. A precise estimate of the illumination components is then obtained based on the preliminary estimate. The precise estimate of the illumination image is then enhanced using a gamma transform to obtain an enhanced image. The enhanced image is then denoised using the BM3D algorithm to obtain a denoised image. Finally, the enhanced image and the denoised image are merged to obtain a denoised enhanced image a of low-illumination physical examination image A. The denoised enhanced image a is then used to verify the authenticity of user 1's health declaration, assess whether the claim application complies with the terms and conditions, and assess whether there is insurance fraud risk.

[0041] In another specific embodiment, when user 2 is conducting an online telemedicine consultation, they are required to upload a physical examination image or a photo of the affected area for health assessment. However, due to issues with the equipment, environment, or shooting angle, the uploaded physical examination image may have low illumination. In this case, the low-illumination physical examination image B uploaded by user 2 is first obtained. Then, a preliminary estimate of the illumination components of low-illumination physical examination image B is obtained based on the Retinex decomposition model. Based on the preliminary estimate of the illumination components, a precise estimate of the illumination image is obtained. The precise estimate of the illumination image is then enhanced using a gamma transform to obtain an enhanced image. The enhanced image is then denoised using the BM3D algorithm to obtain a denoised image. Finally, the enhanced image and the denoised image are merged to obtain a denoised enhanced image b of low-illumination physical examination image B. User 2's health condition is then confirmed based on the denoised enhanced image b so that optimal medical advice can be given.

[0042] In other specific embodiments, the above-mentioned low-light physical examination image noise reduction and enhancement method is also applicable to medical care scenarios. For example, when user 3 is managing a chronic disease, he needs to monitor his condition regularly and upload physical examination images or examination pictures (such as urine routine, electrocardiogram, B-ultrasound images) to help the doctor adjust the treatment plan. When user 4 undergoes a postoperative review, he needs to regularly upload wound photos and imaging data (such as CT, MRI) to help the doctor evaluate the recovery and reduce the number of follow-up visits. In the above situations, it is easy for the images uploaded by users to have low illumination problems. At this time, the image data uploaded by users 3 and 4 are converted into noise-reduced and enhanced images through the above-mentioned low-light physical examination image noise reduction and enhancement method, which can significantly improve the efficiency and accuracy of doctors in adjusting treatment plans and evaluating recovery conditions.

[0043] Figure 3 yes Figure 2 A flow chart of a specific implementation method in step S20 is shown in FIG. Figure 3 As shown, in some embodiments, step S20 includes:

[0044] S21, obtain preliminary estimated illumination components by RGB color channel maximum value method,

[0045] L initial (x,y)=max(R(x,y),G(x,y),B(x,y))+∈, where L initial (x, y) is the preliminary estimated illumination component, R, G, and B are the red, green, and blue channel values ​​of the low-light physical examination image, respectively, and ∈ is a very small constant.

[0046] In this step, R(x,y), G(x,y), B(x,y) represent the pixel values ​​of the red, green, and blue channels of the original image at position (x,y) (usually ranging from 0 to 255). ∈ is a very small constant, such as 10 -3 , used to avoid division by zero in subsequent calculations. max() means taking the maximum value of the three color channels. Retinex theory believes that the brightness of each pixel in the image is mainly determined by the illumination intensity of the scene. In at least one color channel, the contribution of illumination is not absorbed or blocked by the object, so the maximum value of the channel can approximately represent the illumination intensity of the point, L initial (x, y) is the preliminary estimated illumination component of the low-light physical examination image, reflecting the brightness peak of the local area in the image.

[0047] Figure 4 yes Figure 2 Another specific implementation flow diagram of step S20 is as follows: Figure 4 As shown, in some embodiments, step S20 further includes:

[0048] S22. Construct an optimization objective function including a first-order derivative constraint according to the preliminary estimated illumination component, construct an augmented Lagrangian function according to the optimization objective function, and solve the augmented Lagrangian function to obtain the precisely estimated illumination image.

[0049] In this step, we first construct an optimization objective function containing first-order derivative constraints based on the preliminary estimated illumination components. in‖‖ F is the F norm, ‖‖1 is the l1 norm, is the data fidelity item, which is used to calculate the optimized illumination component L and the initial illumination image L initial The fidelity between α and α is a constant. It is a first-order derivative filter, which includes vertical filtering and horizontal filtering, and is used to maintain the smoothness of the overall structure and texture of the image.

[0050] Secondly, the augmented Lagrangian function is constructed according to the optimization objective function. This step requires the introduction of auxiliary variables G = ▽L to construct the augmented Lagrangian function. Among them, W is the weight matrix, G is the auxiliary variable, μ is the Lagrange coefficient, and ρ is the penalty coefficient. is a linear constraint that forces the gradient of the illumination component L is consistent with the auxiliary variable G, ensuring that the optimized illumination component satisfies both smoothness and the gradient sparsity assumption, decomposing the original problem into sub-problems that can be optimized alternately. It is a quadratic penalty term that accelerates the convergence of the algorithm by penalizing the difference between the gradient ▽L and the auxiliary variable G.

[0051] Finally, the augmented Lagrangian function is solved to obtain an accurate estimated illumination image L refined .

[0052] Figure 5 yes Figure 2 Another specific implementation flow diagram of step S20 is as follows: Figure 5 As shown, in some embodiments, step S20 further includes:

[0053] S231 : Normalize the pixel values ​​of the precisely estimated illumination image to obtain a first intermediate enhanced image.

[0054] In this step, the pixel values ​​of the accurately estimated illumination image are normalized to ensure the numerical stability of the exponential operation and avoid overflow.

[0055] S232: Set a gamma value according to the precisely estimated illumination image, and perform a gamma transform on the first intermediate enhanced image based on the gamma value to obtain a second intermediate enhanced image.

[0056] In this step, a gamma value is set based on the accurately estimated illumination image. When the accurately estimated illumination image is in low illumination, a gamma value less than 1 is selected to increase the brightness of dark areas. When the accurately estimated illumination image is overexposed in highlights, a gamma value greater than 1 is selected to suppress bright areas and enhance contrast. After determining the gamma value, the first intermediate enhanced image is gamma-transformed based on the gamma value to obtain a second intermediate enhanced image.

[0057] S233: Denormalize the pixel values ​​of the second intermediate enhanced image to obtain an enhanced image.

[0058] In this step, the pixel values ​​of the second intermediate enhanced image are denormalized and mapped back to the same scale as the original image to make the result interpretable, thereby obtaining an enhanced image after gamma shift.

[0059] Figure 6 yes Figure 2 A specific implementation flow diagram of step S30 is shown in FIG. Figure 6 As shown, step S30 includes:

[0060] S31. Obtain a noise standard deviation in the enhanced image, and configure noise reduction parameters according to the standard deviation.

[0061] In this step, the local variance of each pixel neighborhood in the flat texture area of ​​the enhanced image is first obtained, and then the median of the local variance is squared to obtain the noise standard deviation σ. The noise reduction parameters include the spatial domain standard deviation σ space , color domain standard deviation σ color And the window size h. Spatial domain standard deviation σ space Usually the value is fixed at 1.5 to 3.0 (pixel unit). color Proportional to the noise standard deviation σ, σ color =k·σ, the proportional coefficient k is determined by the noise intensity. The proportional parameter for mild noise is [1.0~1.5], the proportional parameter for moderate noise is [1.5~2.5], and the proportional parameter for severe noise is (2.5~3.5). The window size h=2·[2σ space ]+1.

[0062] S32. Perform noise reduction on each color channel of the enhanced image according to the noise reduction parameters to obtain an intermediate noise reduction image of each color channel.

[0063] In this step, the enhanced image is split into three single-channel images of R, G, and B, and bilateral filtering is applied to each channel independently to obtain the intermediate denoised image of the three single-channel images.

[0064] S33: Merge all the intermediate denoised images to obtain a denoised image.

[0065] In this step, the intermediate denoised images of the three single-channel images are merged to obtain a denoised image.

[0066] Figure 7 yes Figure 6 A specific implementation flow diagram of step S32 is shown in FIG. Figure 7 As shown, in some embodiments, step S32 may include:

[0067] S321. If the color space of the enhanced image is an RGB color space, first convert the color space of the enhanced image from the RGB color space to the YUV color space, and then perform noise reduction on a single channel Y of the enhanced image to obtain the intermediate noise reduction image.

[0068] In this embodiment, the three color channels in the RGB color space are highly correlated, making direct processing inefficient and prone to color deviation. After converting the enhanced image's color space from RGB to YUV, noise reduction is performed on the single Y channel of the enhanced image to reduce computational complexity while maintaining relative color stability. The resulting intermediate denoised image then needs to be converted back to the RGB color space to reconstruct the standard color space image to ensure compatibility.

[0069] Figure 8 yes Figure 2 A specific implementation flow diagram of step S40 is shown in FIG. Figure 8 As shown, in some embodiments, step S40 includes:

[0070] S41, performing weighted fusion on the enhanced image and the denoised image to obtain a denoised enhanced image, R final =w·R enhanced +(1-w)R denoised , where R final To reduce noise and enhance the image, R enhanced To enhance the image, R denoised is the denoised image, and w is the weight.

[0071] In this step, the enhanced and denoised images are linearly combined, balancing the advantages of both to produce a denoised enhanced image. This balance is achieved between preserving detail in the enhanced image and maintaining smoothness in the denoised image. Its core strengths lie in flexibility and efficiency, and the effects can be further enhanced through adaptive weighting or multi-scale decomposition. In practical applications, fixed or dynamic weighting should be selected based on scenario requirements. Parameters should be optimized through a combination of objective metrics and subjective evaluation to ultimately produce a high-quality denoised enhanced image.

[0072] In some implementations, after step S41 , the process may further include performing an unsharp mask on the noise reduction and enhanced image to enhance edges.

[0073] The low-light physical examination image noise reduction and enhancement method provided by the embodiment of the present invention can be constructed based on artificial intelligence, and the relevant data can be acquired and processed based on artificial intelligence technology to achieve unattended low-light physical examination image noise reduction and enhancement. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. Basic artificial intelligence technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. Artificial intelligence software technology mainly includes several major directions such as computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0074] In one embodiment, a low-light physical examination image noise reduction and enhancement device is provided, which corresponds to the low-light physical examination image noise reduction and enhancement method in the above embodiment. Figure 9 As shown, the low-light physical examination image noise reduction and enhancement device includes:

[0075] The image acquisition module 100 is used to acquire low-illumination physical examination images to be processed.

[0076] The image enhancement module 200 is used to obtain a preliminary estimated illumination component of the low-light physical examination image based on a Retinex decomposition model, obtain a precisely estimated illumination image based on the preliminary estimated illumination component, and enhance the precisely estimated illumination image through gamma transformation to obtain an enhanced image.

[0077] The image denoising module 300 is configured to perform denoising on the enhanced image based on a BM3D algorithm to obtain a denoised image.

[0078] The image merging module 400 is configured to merge the enhanced image with the denoised image to obtain a denoised enhanced image.

[0079] Specifically, the image enhancement module 200 includes a first estimation unit.

[0080] The first estimation unit is used to obtain a preliminary estimated illumination component, L, by using the RGB color channel maximum value method. initial (x,y)=max(R(x,y),G(x,y),B(x,y))+∈, where L initial (x, y) is the preliminary estimated illumination component, R, G, and B are the red, green, and blue channel values ​​of the low-light physical examination image, respectively, and ∈ is a very small constant.

[0081] Specifically, the image enhancement module 200 further includes a second estimation unit.

[0082] The second estimation unit is used to construct an optimization objective function containing a first-order derivative constraint according to the preliminary estimated illumination component, construct an augmented Lagrangian function according to the optimization objective function, and solve the augmented Lagrangian function to obtain the accurately estimated illumination image.

[0083] Specifically, the image enhancement module 200 further includes a first enhancement unit, a second enhancement unit, and a third enhancement unit.

[0084] The first enhancement unit is configured to normalize the pixel values ​​of the precisely estimated illumination image to obtain a first intermediate enhanced image.

[0085] The second enhancement unit is configured to set a gamma value according to the precisely estimated illumination image, and perform a gamma transform on the first intermediate enhanced image based on the gamma value to obtain a second intermediate enhanced image.

[0086] The third enhancement unit is configured to denormalize the pixel values ​​of the second intermediate enhanced image to obtain an enhanced image.

[0087] Specifically, the image denoising module 300 further includes a denoising parameter acquiring unit, a channel denoising unit and a denoising merging unit.

[0088] The noise reduction parameter acquisition unit is used to obtain the noise standard deviation in the enhanced image and configure the noise reduction parameters according to the standard deviation.

[0089] A channel denoising unit is configured to perform denoising on each color channel of the enhanced image according to the denoising parameters to obtain an intermediate denoised image of each color channel.

[0090] A denoising merging unit, configured to merge all the intermediate denoised images to obtain a denoised image

[0091] Specifically, the channel denoising unit is also used to convert the color space of the enhanced image from RGB color space to YUV color space if the color space of the enhanced image is RGB color space, and then denoise the single channel Y of the enhanced image to obtain the intermediate denoised image.

[0092] Specifically, the image merging module 400 is further configured to perform weighted fusion of the enhanced image and the denoised image to obtain a denoised enhanced image. final =w·R enhanced +(1-w)R denoised , where R final To reduce noise and enhance the image, R enhanced To enhance the image, R denoisedis the denoised image, and w is the weight.

[0093] For the specific definition of the low-light physical examination image noise reduction and enhancement device, please refer to the definition of the low-light physical examination image noise reduction and enhancement method above, which will not be repeated here. The various modules in the above-mentioned low-light physical examination image noise reduction and enhancement device can be implemented in whole or in part through software, hardware and their combination. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0094] Figure 10 FIG. 1 is a schematic structural diagram of an embodiment of a computer device according to the present application, as shown in FIG. Figure 10 As shown, the present application also provides a computer device, including:

[0095] The memory and the processor store computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor executes any step of the above-mentioned low-illumination physical examination image noise reduction and enhancement method.

[0096] The present application also provides a computer-readable storage medium. When the computer-readable instructions are executed by one or more processors, the one or more processors execute any step in the low-illumination physical examination image noise reduction and enhancement method. It can be understood that the readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0097] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the scope of protection of the present invention.

Claims

1. A method for noise reduction and enhancement of low-light physical examination images, characterized in that: include: Acquiring a low-light physical examination image to be processed; Obtaining a preliminary estimated illumination component of the low-illumination physical examination image based on a Retinex decomposition model, obtaining a precisely estimated illumination image based on the preliminary estimated illumination component, and enhancing the precisely estimated illumination image through a gamma transform to obtain an enhanced image; Denoising the enhanced image based on the BM3D algorithm to obtain a denoised image; The enhanced image and the denoised image are combined to obtain a denoised enhanced image.

2. The method for noise reduction and enhancement of low-light physical examination images according to claim 1, wherein: The obtaining of a preliminary estimated illumination component of the low-illumination physical examination image based on a Retinex decomposition model includes: The initial estimated illumination component is obtained by the RGB color channel maximum method, L initial (x,y)=max(R(x,y),G(x,y),B(x,y))+∈, where L initial (x, y) is the preliminary estimated illumination component, R, G, and B are the red, green, and blue channel values ​​of the low-light physical examination image, respectively, and ∈ is a very small constant.

3. The method for noise reduction and enhancement of low-light physical examination images according to claim 1, wherein: The obtaining of an accurately estimated illumination image according to the preliminarily estimated illumination components comprises: An optimization objective function including a first-order derivative constraint is constructed according to the preliminary estimated illumination component, an augmented Lagrangian function is constructed according to the optimization objective function, and the augmented Lagrangian function is solved to obtain the accurately estimated illumination image.

4. The method for noise reduction and enhancement of low-light physical examination images according to claim 1, wherein: The step of enhancing the precisely estimated illumination image by gamma transform to obtain an enhanced image includes: Normalizing pixel values ​​of the precisely estimated illumination image to obtain a first intermediate enhanced image; setting a gamma value according to the precisely estimated illumination image, and performing a gamma transform on the first intermediate enhanced image based on the gamma value to obtain a second intermediate enhanced image; The pixel values ​​of the second intermediate enhanced image are denormalized to obtain an enhanced image.

5. The method for noise reduction and enhancement of low-light physical examination images according to claim 1, wherein: The denoising of the enhanced image based on the BM3D algorithm to obtain a denoised image includes: Obtaining a noise standard deviation in the enhanced image, and configuring noise reduction parameters according to the standard deviation; Performing noise reduction on each color channel of the enhanced image according to the noise reduction parameters to obtain an intermediate noise reduction image of each color channel; All the intermediate denoised images are merged to obtain a denoised image.

6. The method for noise reduction and enhancement of low-light physical examination images according to claim 5, characterized in that: The step of performing noise reduction on each color channel of the enhanced image according to the noise reduction parameters to obtain an intermediate noise reduction image of each color channel includes: If the color space of the enhanced image is an RGB color space, the color space of the enhanced image is first converted from the RGB color space to the YUV color space, and then the single channel Y of the enhanced image is denoised to obtain the intermediate denoised image.

7. The method for noise reduction and enhancement of low-light physical examination images according to claim 1, wherein: The merging of the enhanced image and the denoised image to obtain the denoised enhanced image includes: The enhanced image and the denoised image are weightedly fused to obtain the denoised enhanced image, R final =w·R enhanced +(1-w)R denoised , where R final To reduce noise and enhance the image, R enhanced To enhance the image, R denoised is the denoised image, and w is the weight.

8. A low-light physical examination image noise reduction and enhancement device, characterized in that: include: An image acquisition module, used for acquiring low-light physical examination images to be processed; an image enhancement module, configured to obtain a preliminary estimated illumination component of the low-light physical examination image based on a Retinex decomposition model, obtain a precisely estimated illumination image based on the preliminary estimated illumination component, and enhance the precisely estimated illumination image by gamma transform to obtain an enhanced image; An image denoising module, configured to denoise the enhanced image based on a BM3D algorithm to obtain a denoised image; An image merging module is used to merge the enhanced image with the denoised image to obtain a denoised enhanced image.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the low-illumination physical examination image noise reduction and enhancement method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method for denoising and enhancing low-illumination physical examination images as claimed in any one of claims 1 to 7 are implemented.

Images