Image processing method, electronic system, and non-transitory computer readable medium

Abstract

An image processing method, an electronic system, and a non-transitory computer readable medium are provided. The image processing system is applied to an electronic system. The image processing system comprises the following steps: acquiring an initial image; performing degradation processing on the initial image according to a preset light source of a camera module to generate an input image; combining the image of the preset light source into the light area of the initial image to generate a ground live image; and processing the input image with an image processing network module to generate an output image, the processing of the input image with the image processing network module including training the image processing network module according to a comparison of the input image and the ground live image.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to U.S. patent application No. 63 / 334,487, filed on April 25, 2022, and the entire contents of the above patent application are incorporated by reference into this application. Technical Field

[0003] The present application relates to the field of data synthesis, and in particular, to an image processing method, an electronic system, and a non-temporary computer-readable medium. Background Art

[0004] In general, large amounts of training data are essential for data-driven algorithms such as those based on neural networks. Although it is theoretically possible to obtain training data pairs through experiments, the time and labor costs of obtaining hundreds or thousands of images are too expensive to be practically feasible. More specifically, for network-based algorithms that perform image restoration tasks, image pairs containing actual degraded images and ideal non-degraded images are required in order to train the neural network to learn the transition from degraded images to ideal non-degraded images.

[0005] However, the time required for physical measurements alone exceeds the normal cycle allowed by typical R&D time budgets. For example, typical data-driven neural network training requires the number of training data pairs to be at least on the order of 100,000. Assuming a data collection rate of 100 image pairs per hour, the time required will exceed 100 working days.

[0006] In addition, special hardware systems need to be designed and built to acquire degraded images and ideal non-degraded images. This data acquisition needs to minimize the differences in factors that are not related to the image degradation process. For example, time differences or perspective differences. In reality, undesirable differences between degraded images and ideal non-degraded images are inevitable and will affect the quality of neural network training. Summary of the invention

[0007] Technical issues

[0008] A novel data synthesis method is needed to generate realistic training data for training state-of-the-art neural network-based image restoration algorithms.

[0009] Solution to the problem

[0010] The image processing method of the present application is applied to an electronic system. The image processing method comprises the following steps: acquiring an initial image; performing degradation processing on the initial image according to a preset light source of a camera module to generate an input image; merging the image of the preset light source into the light area of ​​the initial image to generate a ground truth image; and processing the input image using an image processing network module to generate an output image, wherein the processing of the input image using the image processing network module comprises training the image processing network module according to a comparison between the output image and the ground truth image.

[0011] In one embodiment of the present application, the camera module is an under-screen camera.

[0012] In one embodiment of the present application, the image processing method also includes the following steps: determining whether the comparison of the image quality of the output image and the ground truth image meets a preset standard; and if the comparison of the image quality of the output image and the ground truth image meets the preset standard, completing the training of the image processing network module and associating the image processing network module with multiple weights and biases.

[0013] In one embodiment of the present application, the initial image includes a first initial image, the input image includes a first input image, and the ground truth image includes a first ground truth image. The image processing method further includes the following steps: if the comparison of the image quality of the first output image and the first ground truth image does not meet a preset standard: acquiring a second initial image; generating a second input image and a second ground truth image; and training the image processing network module according to the second input image and the second ground truth image.

[0014] In one embodiment of the present application, the initial image is selected from an image database, which includes a plurality of low-noise images whose image quality is higher than a threshold level, and the image database is configured to provide the plurality of low-noise images for training the image processing network module.

[0015] In an embodiment of the present application, the step of performing degradation processing on the initial image includes the following steps: applying a fog effect, a diffraction effect and one or more types of noise to the initial image.

[0016] In one embodiment of the present application, the one or more types of noise include shooting noise and read noise. The image processing method further includes the following steps: determining the shooting noise and the read noise according to camera characteristics. The step of performing degradation processing on the initial image also includes the following steps: adding the shooting noise and the read noise to the initial image.

[0017] In one embodiment of the present application, the ground truth image does not have the fogging effect or the one or more types of noise, and the light source in the ground truth image is convolved with the preset light source of the camera module.

[0018] In one embodiment of the present application, the step of acquiring the initial image includes the following steps: acquiring a digital image in a JPEG image format; and converting the digital image from the JPEG image format to a RAW image format to generate an initial image in a RAW image format.

[0019] In one embodiment of the present application, the preset light source includes an array of point light sources. Each point light source includes a fiber light source covered by a pinhole with a radius less than 100 times the wavelength of light, and each of the point light sources is configured to provide incident light with a range of incident angles. The image processing method further includes the following steps: representing each point light source using a point source function to generate the input image and the ground truth image.

[0020] In one embodiment of the present application, each of the point source functions has a high dynamic range with a brightness greater than a threshold ratio, and each of the point source functions applies a series of exposure times for each of a plurality of incident angles, and a weighted average is applied in a corresponding input image or a corresponding ground truth image.

[0021] In one embodiment of the present application, the initial image includes a plurality of saturated pixels and a plurality of unsaturated pixels, and the unsaturated pixels are different from the saturated pixels. The step of performing degradation processing on the initial image includes the following steps: performing degradation processing on the initial image includes changing the preset light sources of the plurality of unsaturated pixels and the plurality of saturated pixels; and merging the image of the preset light source into the light region of the initial image includes applying an enhanced light source to the plurality of saturated pixels.

[0022] In an embodiment of the present application, the enhanced light source is represented by a plurality of point source functions with different shapes and intensity distributions.

[0023] In one embodiment of the present application, the step of changing the preset light source also includes the following steps: using multiple spatially averaged point source functions to represent the preset light source to degenerate the multiple unsaturated pixels; and using multiple point source functions with different shapes and intensity distributions to represent the preset light source to degenerate the multiple saturated pixels.

[0024] The electronic system of the present application includes one or more processors; and a memory, which is coupled to the one or more processors and is configured to store multiple instructions. When the instructions are executed by the one or more processors, the one or more processors perform the above-mentioned image processing method.

[0025] The non-transitory computer-readable medium of the present application is configured to store a plurality of instructions. When the instructions are executed by one or more processors, the one or more processors perform the above-mentioned image processing method.

[0026] Effects of the Invention

[0027] In summary, the image processing method, electronic system and non-transitory computer-readable medium according to the present application can automatically generate a large amount of training data and effectively train the image processing network module.

[0028] In order to make the above contents easier to understand, several embodiments with accompanying drawings will be described in detail below. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 is a schematic diagram of an electronic system according to an embodiment of the present application.

[0030] Figure 2 is a flowchart of an image processing method according to an embodiment of the present application.

[0031] Figure 3 is a flowchart of an image processing method according to another embodiment of the present application.

[0032] Figure 4A is a schematic diagram of an input image according to an embodiment of the present application.

[0033] Figure 4B is a schematic diagram of an input image according to an embodiment of the present application.

[0034] Figure 5A is a schematic diagram of measuring a point spread function according to an embodiment of the present application.

[0035] Figure 5B is a schematic diagram of a set of reconstructed point spread functions according to an embodiment of the present application.

[0036] Figure 6 is a flow chart of a data simulation pipeline according to an embodiment of the present application.

[0037] Figure 7 The flowchart of generating a simulated noise-free degraded image according to an embodiment of the present application is shown in FIG.

[0038] Figure 8 is a noise synthesis flow chart according to an embodiment of the present application. DETAILED DESCRIPTION

[0039] Figure 1 is a schematic diagram of an electronic system according to an embodiment of the present application. Figure 1In one embodiment of the present application, the electronic system 100 includes a processor 110 and a memory 120, and the memory 120 can store an image processing network module 121 and related instructions. The processor 110 is (electronically) coupled to the memory 120, and can execute the image processing network module 121 and related instructions to implement the image processing method of the present application. In one embodiment of the present application, the electronic system 100 can be composed of one or more personal computers (personal computers, PCs), one or more server computers, and one or more workstation computer devices, which is not limited in the present application. In one embodiment of the present application, the electronic system 100 can include more processors for executing the image processing network module 121 and related instructions to implement the image processing method of the present application.

[0040] In one embodiment of the present application, the processor 110 may include a central processing unit (CPU), a graphic processing unit (GPU), or other programmable general or special microprocessors, digital signal processors (DSP), application specific integrated circuits (ASIC), programmable logic devices (PLD), other similar processing circuits or combinations of these devices. In one embodiment of the present application, the memory 120 may be a non-temporary computer-readable recording medium, such as a read-only memory (ROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM) or a non-volatile memory (NVM), which is not limited in this application. In one embodiment of the present application, the image processing network module 121 may also be stored in a non-temporary computer-readable recording medium of one device and executed by a processor of another device.

[0041] In one embodiment of the present application, the image processing network module 121 includes a neural network model, which can be implemented by an engineering convolutional neural network (U-NET) structure, and the present application does not limit this. In one embodiment of the present application, the above-mentioned neural network model can also be implemented by a context aggregation network (CAN) or other convolutional neural networks.

[0042] Figure 2 is a flowchart of an image processing method according to an embodiment of the present application. Figure 1 and Figure 2 , the electronic system 100 may perform the following steps S201 to S204 to implement the image processing method. In step S201, the processor 110 may obtain an initial image. In an embodiment of the present application, the initial image may be selected from an image database, wherein the image database includes a plurality of low-noise images having image quality higher than a threshold level, and the image database may be configured to provide a plurality of low-noise images for training the image processing network module 121. The image database may be set in an external storage device or memory 120.

[0043] In step S202, the processor 110 may perform degradation processing on the initial image according to the preset light source of the camera module to generate an input image. In one embodiment of the present application, the processor 110 may use the initial image to simulate the image captured by the camera module under the influence of a strong point light source, other noises and equipment. The camera module may be an under-screen camera, which is not limited in this application. In step S203, the processor 110 may merge the image of the preset light source into the light area of ​​the initial image to generate a ground-truth image. In one embodiment of the present application, the processor 110 may use the same initial image to simulate the ground-truth image to represent the desired result processed by the image processing network module 121.

[0044] In step S204, the processor 110 may process the input image using the image processing network module 121 to generate an output image; wherein, processing the input image using the image processing network module 121 includes training the image processing network module 121 based on the comparison of the output image with the ground truth image. In an embodiment of the present application, the processor 110 may generate training data (input image and ground truth image) based on the initial image provided by the image database. The processor 110 may input the input image into the image processing network module 121 so that the image processing network module 121 outputs the output image. The processor 110 may compare the output image with the ground truth image to determine whether the image quality comparison of the output image and the ground truth image meets the preset standard. In addition, if the image quality comparison of the output image and the ground truth image meets the preset standard, the processor 110 may complete the training of the image processing network module and associate the image processing network module with multiple weights and multiple biases. Therefore, the electronic system 100 and the image processing method in an embodiment of the present application may automatically generate training data and effectively train the image processing network module 121 by using the training data.

[0045] Figure 3 is a flowchart of an image processing method according to another embodiment of the present application. Figure 3 Examples can be used to illustrate Figure 2 Specific implementation details of the image processing method of the embodiment. Figure 1 and Figure 3 , the electronic system 100 may perform the following steps S301 to S310 to implement the image processing method. It should be noted that the processor 110 may execute the data simulation pipeline (i.e., execute steps S301 to S307) to implement, for example, under-screen camera image synthesis, and execute the neural network processing pipeline (i.e., steps S308 to S310) to implement normal image synthesis.

[0046] In step S301, the processor 110 may acquire an initial image. The initial image may be a randomly sampled image from an image database. In step S302, the processor 110 may perform a haze simulation on the initial image. In step S304, the processor 110 may perform a bright light source diffraction simulation on the initial image. In step S305, the processor 110 may perform a noise simulation on the initial image. In step S306, the processor 110 may generate an input image (i.e., a degraded image). The processor 110 may perform a degradation process on the initial image according to a preset light source of the camera module, and apply a haze effect, a diffraction effect, and one or more types of noise to the initial image. For example, referring to Figure 4A , Figure 4A is a schematic diagram of an input image according to an embodiment of the present application. Figure 4AAs shown, the processor 110 may generate an input image 410 .

[0047] On the other hand, in step S303, the processor 110 may perform light source simulation in the initial image. In step S307, the processor 110 may generate a ground truth image (i.e., as a target sample). The processor 110 may generate the initial image according to the preset light source of the camera module. For example, referring to Figure 4B , Figure 4B is a schematic diagram of a ground truth image according to an embodiment of the present application. Figure 4B As shown, the processor 110 can generate a ground truth image 420. The ground truth image 420 does not have a fogging effect or one or more types of noise. The light source in the ground truth image 420 is convolved with the preset light source of the camera module.

[0048] Then, in step S308, the processor 110 may input the input image to the image processing network module 121 so that the image processing network module 121 generates an output image. In step S309, the processor 110 may compare the image quality of the output image and the ground truth image. In step S310, the processor 110 may determine whether the comparison of the image quality meets the preset standard. If the image quality comparison meets the preset standard, the processor 110 will complete the training of the image processing network module 121 and associate the image processing network module 121 with multiple weights and multiple biases. If the image quality comparison does not meet the preset standard, the processor 110 may generate the next training data and continue to train the image processing network module 121.

[0049] For example, the initial image may include a first initial image, the input image may include a first input image, and the ground truth image may include a first ground truth image. If the image quality comparison of the first output image and the first ground truth image does not meet the preset standard, the processor 110 may further acquire a second initial image, generate a second input image and a second ground truth image, and train the image processing network module 121 according to the second input image and the second ground truth image. By analogy, the processor 110 may repeatedly perform steps S301 to S307 to generate a large amount of training data, and repeatedly perform steps S308 to S310, and effectively train the image processing network module 121 by using a large amount of training data.

[0050] Figure 5A is a schematic diagram of measuring a point spread function according to an embodiment of the present application. Figure 5B is a schematic diagram of a set of reconstructed point spread functions according to an embodiment of the present application. Figure 1 , Figure 5A and Figure 5BIn one embodiment of the present application, a method for measuring a high dynamic range point source function (PSF) is provided below. Specifically, the preset light source may include an array of point light sources, each of which may include an optical fiber light source obscured by a pinhole. The preset light source may be used to approximate a point light source, that is, an optical fiber light source obscured by a pinhole having a radius less than 100 times the average wavelength. Figure 5A As shown, a fiber optic light source 510 obscured by a pinhole may be configured to provide incident light to a camera module 520. The processor 110 may represent each point light source with a point source function to generate an input image and a ground truth image.

[0051] In one embodiment of the present application, since multiple optical processes may include incident-angle-dependent diffraction, the PSF may have a complex dependency on the direction of the point light source relative to the optical center of the camera module. That is, each point light source may provide incident light with a certain range of incident angles. Therefore, Figure 5A As shown, the camera module 520 can be coupled to a data acquisition computer 530 (or Figure 1 The processor 110 in the embodiment of the present invention can be arranged on the electric pan-tilt head 540. The electric pan-tilt head 540 can be used and programmed to move the smartphone to a pan-tilt angle range of, for example, approximately 60 degrees by 60 degrees, so that the camera module 520 can generate multiple captured images to obtain the PSF at different incident angles. Therefore, the measurement of the PSF can cover the entire field of view of the camera. Therefore, the data acquisition computer 530 can collect the captured images generated by the camera module 520 to perform post-processing to generate the PSF. Figure 5B As shown in image 550, the data acquisition computer 530 can reconstruct a set of PSFs from the captured image, the PSFs corresponding to multiple sub-regions of the entire image area.

[0052] In one embodiment of the present application, many optical processes of image degradation involve optical diffraction, and one feature of the diffraction pattern is that its power is spatially spread over many high-order features. Although high-order diffraction is not prominent under normal lighting conditions, it will be very obvious when the scene of interest contains a very strong light source. For example, the spotlights commonly found in shopping malls will produce strong diffraction patterns, which seriously reduce the clarity of other parts of the image. Therefore, in order to fully describe very high-order diffraction patterns, one embodiment of the present application can measure PSF within an ultra-high dynamic range with a brightness exceeding 1,000,000:1. From the lowest exposure time analog gain setting to the highest exposure time analog gain setting, more than 50 frames are taken in each incident direction, and then weighted averaged according to its exposure time analog gain setting to obtain a high dynamic range (HDR) result. Therefore, in one embodiment of the present application, the camera module 520 can also obtain a high dynamic ratio image for measuring PSF, so that each reconstructed PSF has a high dynamic range greater than the brightness threshold ratio. Furthermore, each reconstructed PSF may apply a series of exposure times for each of a plurality of incident angles and apply a weighted average in the corresponding input image or the corresponding ground truth image.

[0053] In one embodiment of the present application, the data acquisition computer 530 may further perform data enhancement to further enhance the sampling diversity of the discrete measurement PSF. The applied data enhancement may include spatial enhancement based on random affine transformation (covering spatial translation, scaling, shearing, rotation and reflection) and random color enhancement (covering the spectral range of common light sources).

[0054] Figure 6 is a data simulation pipeline flow chart according to an embodiment of the present application. Figure 1 and Figure 6 , the processor 110 may execute the following steps S601 to S607 to implement a data simulation pipeline. Figure 6 Examples can be used to illustrate the above Figure 3 Specific implementation details of the data simulation pipeline of the embodiment. In one embodiment of the present application, if the image database does not provide enough linear data (RAW image format), the processor 110 can quickly convert (simulate) linear RAW data from available tone mapping data (JPEG or similar format).

[0055] In step S601, the processor 110 may convert the initial image from a JPEG image format to a RAW image format. In one embodiment of the present application, the JPEG to RAW conversion process is an approximate inversion of an image signal processing (ISP) pipeline, which may include white balance adjustment, color correction, gamma correction, tone mapping, etc. The ISP parameters used in the JPEG to RAW conversion process are based on imaging statistics of the relevant camera module. In one embodiment of the present application, the initial image may include a plurality of saturated pixels and a plurality of unsaturated pixels. And unsaturated pixels are different from saturated pixels.

[0056] In step S602, the processor 110 may perform unsaturated pixel degradation processing on multiple unsaturated pixels in the initial image using a spatial average PSF. In one embodiment of the present application, for unsaturated pixels, the degradation process can be well approximated as a linear system, and the inverse problem of restoring an image with unsaturated pixels is generally easier than the inverse problem of restoring an image with saturated pixels. Taking the under-screen camera as an example, the degradation of its unsaturated pixels is manifested as a fog-like appearance, and the fog-like appearance has a loss of sharpness and contrast. In the under-screen camera, this degradation is called fog because of its foggy appearance. Therefore, in order to synthesize degraded unsaturated pixels, the processor 110 may convolve the pixels in the linear RAW data with the corresponding PSF. Although the PSF of the imaging system is usually spatially variable, the spatial average PSF can be used to speed up the simulation while capturing most of the features of the unsaturated pixel degradation process. In step S603, the processor 110 may perform saturated pixel degradation processing on multiple saturated pixels using a random affine transform PSF. In other words, the processor 110 may change the preset light source (ie, PSF) of a plurality of unsaturated pixels and a plurality of saturated pixels in the initial image to generate a simulated noise-free degraded image.

[0057] In step S605, the processor 110 may add noise to the simulated noiseless degraded image. In step S606, the processor 110 may generate an input image. On the other hand, in step S604, the processor 110 may perform saturated pixel degradation using a normal PSF. The processor 110 may apply an enhanced light source to multiple saturated pixels of the initial image using a normal PSF. The enhanced light source may be represented by multiple PSFs having different shapes and intensity distributions. Therefore, in step S607, the processor 110 may generate a ground truth image. Therefore, the electronic system 100 may achieve efficient and accurate image synthesis by separating the degradation modeling of unsaturated pixels and saturated pixels.

[0058] Figure 7 is a flow chart for generating a simulated noise-free degraded image according to an embodiment of the present application. Figure 1 and Figure 7 , Figure 7 Examples can be used to illustrate the above Figure 6 The embodiment in (i.e. Figure 6 The specific implementation details of the saturated pixel degradation synthesis of step S603) are as follows. In step S701, the processor 110 may generate multiple enhanced PSFs (enhanced spatial variant PSFs). In step S702, the processor 110 may simulate multiple light sources with various shapes and intensity distributions. In step S703, the processor 110 may convolve the light source with the enhanced PSF. That is, the non-degraded light source may be convolved with the enhanced spatial variant PSF to obtain a degraded image of the light source. In step S704, the processor 110 may generate a linearized image with simulated unsaturated pixel degradation through multiple spatially averaged PSFs. In step S705, the processor 110 may generate a simulated noise-free degraded image. That is, the degraded image of the light source is superimposed on the linearized image (a synthetic degraded image with only an unsaturated image) to generate a simulated noise-free degraded image.

[0059] In other words, in one embodiment of the present application, the processor 110 may use multiple spatially averaged PSFs to represent a preset light source, which is used to degrade multiple unsaturated pixels in the initial image; and use multiple PSFs with different shapes and intensity distributions to represent a preset light source, which is used to degrade multiple saturated pixels in the initial image, thereby generating a simulated noise-free degraded image.

[0060] Figure 8 is a noise synthesis flow chart according to an embodiment of the present application. Figure 1 and Figure 8 , Figure 8 Examples can be used to illustrate the above Figure 6 The specific implementation details of the noise synthesis in the embodiment (ie Figure 6 In step S605). In step S801, the processor may obtain a simulated noise-free degraded image. In step S802, the processor 110 may obtain shot noise. The processor may superimpose the initial image with the image with shot noise. In step S803, the processor 110 may obtain read noise. The processor may superimpose the previously superimposed image with the image with read noise. In one embodiment of the present application, the processor 110 may determine the shot noise and the read noise based on the camera characteristics (camera statistics), and the shot noise and the read noise are modeled using measured noise statistics. In step S804, the processor 110 may generate a simulated degraded image with noise. In this way, the real noise synthesis of the camera module is integrated into the training data generation pipeline to enhance the noise reduction capability of the neural network. In addition, the image processing network module 121 may include adjustable parameters to control the denoising intensity.

[0061] In summary, the electronic system and non-transitory computer-readable medium of the present application can use the existing image database to automatically generate a large amount of reliable and useful training data, thereby effectively training the image processing network module to learn to remove the degradation of unsaturated pixels and saturated pixels and suppress noise. In other words, the present application does not require a lot of manpower, time and cost of experimental data collection, and can generate real training data in a digital way, greatly reducing the time and cost required to train the image processing network module. Therefore, the trained image processing network module can effectively process the images taken by the camera module (such as the under-screen camera) to generate corresponding optimized images.

[0062] Reference Numbers List

[0063] 110: Processor

[0064] 120: Memory

[0065] 121: Image processing network module

[0066] 410: Input image

[0067] 420: Ground truth image

[0068] 510: Fiber Optic Light Source

[0069] 520: Camera module

[0070] 530: Data Acquisition Computer

[0071] 540: Electric PTZ

[0072] 550: Image

[0073] S201~S204, S301~310, S601~S607, S701~S705, S801~S804: Steps

Claims

1. An image processing method, applied to an electronic system, comprising: Get the initial image; Performing degradation processing on the initial image according to a preset light source of the camera module to generate an input image; Merging the image of the preset light source into the light region of the initial image to generate a ground truth image; as well as The input image is processed using an image processing network module to generate an output image, wherein the processing of the input image using the image processing network module includes training the image processing network module based on a comparison of the output image with the ground truth image.

2. The image processing method according to claim 1, characterized in that: The camera module is an under-screen camera.

3. The image processing method according to claim 1 or 2, characterized in that: Also includes: determining whether a comparison of the image quality of the output image and the ground truth image satisfies a preset criterion; as well as If the comparison of the image quality of the output image and the ground truth image meets the preset criteria, the training of the image processing network module is completed and the image processing network module is associated with a plurality of weights and biases.

4. The image processing method according to any one of claims 1 to 3, characterized in that: The initial image includes a first initial image, the input image includes a first input image, and the ground truth image includes a first ground truth image; and the image processing method further includes: If a comparison of the image quality of the first output image and the first ground truth image does not meet a preset criterion: acquiring a second initial image; generating a second input image and a second ground truth image; and The image processing network module is trained based on the second input image and the second ground truth image.

5. The image processing method according to any one of claims 1 to 4, characterized in that: The initial image is selected from an image database, the image database includes a plurality of low-noise images whose image quality is higher than a threshold level, and the image database is configured to provide the plurality of low-noise images for training the image processing network module.

6. The image processing method according to any one of claims 1 to 5, characterized in that: The step of performing degradation processing on the initial image comprises: A fog effect, a diffraction effect, and one or more types of noise are applied to the initial image.

7. The image processing method according to any one of claims 1 to 5, wherein: The one or more types of noise include shooting noise and reading noise, and the image processing method further includes: The shooting noise and the reading noise are determined according to the camera characteristics; wherein the step of performing degradation processing on the initial image further includes: The shooting noise and the read noise are added to the original image.

8. The image processing method according to claim 6, characterized in that: The ground truth image does not have the fogging effect or the one or more types of noise, and the light source in the ground truth image is convolved with the preset light source of the camera module.

9. The image processing method according to any one of claims 1 to 8, characterized in that: The step of obtaining the initial image comprises: Acquiring a digital image in a JPEG image format; and The digital image is converted from a JPEG image format to a RAW image format to generate an initial image in the RAW image format.

10. The image processing method according to any one of claims 1 to 9, characterized in that: The preset light source comprises an array of point light sources, each of the point light sources comprises a fiber optic light source shielded by a pinhole, and each of the point light sources is configured to provide incident light having a range of incident angles; Wherein, the image processing method further includes: Each point light source is represented by a point source function to generate the input image and the ground truth image.

11. The image processing method according to claim 10, characterized in that: Each of the point source functions has a high dynamic range with brightness greater than a threshold ratio, each of the point source functions applies a series of exposure times for each of a plurality of incident angles, and applies a weighted average in a corresponding input image or a corresponding ground truth image.

12. The image processing method according to any one of claims 1 to 11, characterized in that: The initial image includes a plurality of saturated pixels and a plurality of unsaturated pixels, wherein the unsaturated pixels are different from the saturated pixels; The step of performing degradation processing on the initial image include: The step of performing degradation processing on the initial image includes: changing the preset light sources of the plurality of unsaturated pixels and the plurality of saturated pixels; The step of merging the image of the preset light source into the light area of ​​the initial image comprises: An enhanced light source is applied to the plurality of saturated pixels.

13. The image processing method according to claim 12, characterized in that: The enhanced light source is represented by a plurality of point source functions with different shapes and intensity distributions.

14. The image processing method according to claim 12, characterized in that: The step of changing the preset light source comprises: Using a plurality of spatially averaged point source functions to represent the preset light source to degrade the plurality of unsaturated pixels; and The preset light source is represented by a plurality of point source functions having different shapes and intensity distributions to degrade the plurality of saturated pixels.

15. An electronic system comprising: one or more processors; as well as A memory, the memory is coupled to the one or more processors, the memory is configured to store a plurality of instructions, when the instructions are executed by the one or more processors, the one or more processors perform the image processing method according to any one of claims 1 to 14. 16 . A non-transitory computer-readable medium configured to store a plurality of instructions, wherein when the instructions are executed by one or more processors, the one or more processors perform the image processing method according to any one of claims 1 to 14.

Images