Image processing method and device based on bit width compression and restoration and server

Abstract

The invention provides an image processing method and device based on bit width compression and restoration and a server, and relates to the technical field of image processing, and the method comprises the steps: mapping high-bit-depth original image data to a low-bit-depth quantization space through linear mapping to obtain compressed low-bit-depth first compressed image data; performing image processing on the first compressed image data through a pre-trained neural network algorithm, determining second compressed image data, and determining a first space enhancement feature corresponding to the quantization space with low bit depth according to the first compressed image data and the second compressed image data; and performing inverse mapping processing on the first spatial enhancement feature to restore the first spatial enhancement feature to a high-bit-depth quantization space, determining a second spatial enhancement feature corresponding to the high-bit-depth quantization space, and determining target image data according to the original image data and the second spatial enhancement feature. According to the invention, the computing resource consumption during image processing can be obviously saved.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular, to an image processing method, apparatus, and server based on bit-width compression and restoration. Background Art

[0002] Currently, when processing image data, it is usually chosen to directly optimize the image. However, if the resolution of the image is high, or there are many images to be processed, directly optimizing the image consumes a large amount of computing resources during image processing. Related technologies propose that the resolution of the image can be reduced and the low-resolution image can be processed to reduce the resources required during image processing. However, this solution will cause loss of image information and result in poor image processing effects. Summary of the Invention

[0003] In view of this, the purpose of the present invention is to provide an image processing method, apparatus, and server based on bit-width compression and restoration, which can significantly save the computing resource consumption during image processing.

[0004] In a first aspect, an embodiment of the present invention provides an image processing method based on bit-width compression and restoration. The method includes: mapping the original image data with a high bit depth to a quantization space with a low bit depth through linear mapping to obtain the first compressed image data with a low bit depth after compression; performing image processing on the first compressed image data through a pre-trained neural network algorithm to determine the second compressed image data, and determining the first spatial enhancement feature corresponding to the quantization space with a low bit depth according to the first compressed image data and the second compressed image data; performing inverse mapping processing on the first spatial enhancement feature to restore the first spatial enhancement feature to the quantization space with a high bit depth, determining the second spatial enhancement feature corresponding to the quantization space with a high bit depth, and determining the target image data according to the original image data and the second spatial enhancement feature.

[0005] In an implementation manner, the step of mapping the original image data with a high bit depth to a quantization space with a low bit depth through linear mapping to obtain the first compressed image data with a low bit depth after compression includes: obtaining the value range of the original image data, and mapping the original image data to the quantization space with a low bit depth based on the value range of the original image data through a preset linear mapping model to determine the first compressed image data, where the linear mapping model is expressed as:

[0006] where represents the original image data, represents the minimum value of the original image data, represents the maximum value of the original image data, Is the first compressed image data, representing the mapped 8-bit value, with the 8-bit range between 0 and 255.

[0007] In one embodiment, the step of mapping the original image data with high bit depth to the quantization space with low bit depth through linear mapping to obtain the first compressed image data with low bit depth after compression further includes: obtaining the brightness, sharpness, and sensor noise distribution of the original image data; if the brightness, sharpness, or sensor noise distribution is uneven, through non-linear mapping, setting different quantization precisions in different gray value ranges of the image, and compressing the original image data based on the different quantization precisions to obtain the first compressed image data.

[0008] In one embodiment, the step of setting different quantization precisions in different gray value ranges of the image through non-linear mapping includes: in the dark area with low pixel values, using a dense quantization distribution through non-linear mapping to retain the weak light noise characteristics; in the bright area with high pixel values, using a sparse quantization distribution through non-linear mapping to save bit space.

[0009] In one embodiment, the step of performing image processing on the first compressed image data through a pre-trained neural network algorithm to determine the second compressed image data includes: performing denoising processing on the first compressed image data through a pre-trained neural network algorithm, and determining the second compressed image data after optimization of the first compressed image data through wide dynamic range enhancement and low light enhancement operations.

[0010] In one embodiment, the step of determining the first spatial enhancement feature corresponding to the quantization space with low bit depth according to the first compressed image data and the second compressed image data includes: determining the difference between the first compressed image data and the second compressed image data as the first spatial enhancement feature corresponding to the quantization space with low bit depth, where the first spatial enhancement feature is the noise of the image in the quantization space with low bit depth.

[0011] In one embodiment, the step of determining the target image data according to the original image data and the second spatial enhancement feature includes: determining the difference between the original image data and the second spatial enhancement feature as the optimized image data with high bit depth; performing Gamma correction, demosaicing processing, and color space conversion processing on the optimized image data to determine the target image data.

[0012] In a second aspect, an embodiment of the present invention further provides an image processing apparatus based on bit-width compression and restoration. The apparatus includes: a bit-width compression module that maps original image data with a high bit depth to a quantization space with a low bit depth through linear mapping to obtain first compressed image data with a low bit depth after compression; a network processing module that performs image processing on the first compressed image data through a pre-trained neural network algorithm to determine second compressed image data, and determines a first spatial enhancement feature corresponding to the quantization space with a low bit depth based on the first compressed image data and the second compressed image data; a feature restoration module that restores the first spatial enhancement feature to the quantization space with a high bit depth through inverse mapping processing, determines a second spatial enhancement feature corresponding to the quantization space with a high bit depth, and determines target image data based on the original image data and the second spatial enhancement feature.

[0013] In a third aspect, an embodiment of the present invention further provides a server, including a processor and a memory. The memory stores computer-executable instructions that can be executed by the processor, and the processor executes the computer-executable instructions to implement the method according to any one of the first aspect.

[0014] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium. The computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are called and executed by a processor, the computer-executable instructions cause the processor to implement the method according to any one of the first aspect.

[0015] The embodiments of the present invention bring the following beneficial effects: The image processing method, apparatus, and server based on bit-width compression and restoration provided by the embodiments of the present invention map original image data with a high bit depth to a quantization space with a low bit depth through linear mapping to obtain first compressed image data with a low bit depth after compression, perform image processing on the first compressed image data through a pre-trained neural network algorithm to determine second compressed image data, and determine a first spatial enhancement feature corresponding to the quantization space with a low bit depth based on the first compressed image data and the second compressed image data. Finally, the first spatial enhancement feature is restored to the quantization space with a high bit depth through inverse mapping processing, a second spatial enhancement feature corresponding to the quantization space with a high bit depth is determined, and target image data is determined based on the original image data and the second spatial enhancement feature. The embodiments of the present invention can improve the quality of image processing while significantly saving the consumption of computing resources during image processing.

[0016] Other features and advantages of the present invention will be set forth in the following description, and in part will be obvious from the description, or may be learned by practice of the present invention. The objectives and other advantages of the present invention are realized and attained by the structure particularly pointed out in the specification, claims as well as the drawings.

[0017] To make the above objectives, features and advantages of the present invention more comprehensible, the following provides preferred embodiments in conjunction with the accompanying drawings and detailed descriptions are as follows. Description of the Drawings

[0018] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0019] Figure 1 It is a schematic flowchart of an image processing method based on bit-width compression and restoration provided by an embodiment of the present invention; Figure 2 It is a schematic flowchart of an image compression method provided by an embodiment of the present invention; Figure 3 It is a schematic structural diagram of an image processing device based on bit-width compression and restoration provided by an embodiment of the present invention; Figure 4 It is a schematic structural diagram of a server provided by an embodiment of the present invention. Detailed Embodiments

[0020] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions of the present invention in conjunction with the embodiments. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0021] Currently, when processing image data, it is usually chosen to directly optimize the image. However, if the resolution of the image is high, or there are many images to be processed, directly optimizing the image consumes a large amount of computing resources during image processing. Related technologies propose that the resolution of the image can be reduced and the low-resolution image can be processed to reduce the resources required during image processing. However, this solution will cause loss of image information and result in poor image processing effects.

[0022] Referring to CN201610195410.5, an image processing method and a mobile terminal. When performing image processing, multiple frames of original images are synthesized through a preset algorithm to obtain a first intermediate image with a 10-bit width. Then, image processing is performed on the first intermediate image to obtain a second intermediate image with an 8-bit width. Finally, image compression processing is performed on the second intermediate image to obtain a final image with an 8-bit width. That is to say, the image is compressed through synthesis processing and then image processing is performed, thereby reducing the resources required in the image processing process. However, the image generated by this solution is a compressed image, which will cause loss of image information.

[0023] In addition, referring to CN202310085818.7, an image processing method, apparatus, electronic device, storage medium, and imaging device. The first image is processed to obtain a second image with a resolution less than or equal to the resolution of the first image and a bit width less than or equal to the bit width of the first image. Then, the second image is processed based on a neural network to obtain a third image. Finally, based on the first image and the third image, a fourth image is generated such that the resolution of the fourth image is equal to the resolution of the first image and the bit width of the fourth image is less than or equal to the bit width of the first image. That is to say, this solution is equivalent to performing image processing after compressing the image, and then weighted fusion of the processed image and the original image is performed to restore the image resolution to the initial state. However, this solution may still cause loss of image information when generating the second image by image compression and when restoring the resolution of the third image to the initial resolution.

[0024] Based on this, the image processing method based on bit width compression and restoration provided by the embodiments of the present invention can significantly save the consumption of computing resources while retaining the original image information through efficient denoising. Moreover, since the present invention only processes noise whether it is compressing the image or restoring the image to a high-bit-depth quantization space, when determining the target image, the completely uncompressed or restored original image and the restored noise are used. Therefore, it is possible to avoid the problem of image information loss fundamentally while reducing the computing power.

[0025] Referring to Figure 1 The schematic flowchart of an image processing method based on bit width compression and restoration as shown. This method mainly includes the following steps S102 to step S106: Step S102, mapping the original image data with a high bit depth to a quantization space with a low bit depth through linear mapping to obtain the first compressed image data with a compressed low bit depth. Among them, the high bit depth can include 16bitRAW, 12bit RAW, etc., and can also be other bit depths such as 10bit and 14bit. The image format is not limited to RAW and is also applicable to RGB format and ISP stream format. In practical applications, image data with a high bit depth can be received and map the high-bit data into an 8-bit quantization space in a linear or non-linear manner to obtain 8-bit image data:

[0026] Step S104: Perform image processing on the first compressed image data through a pre-trained neural network algorithm to determine the second compressed image data, and determine the first spatial enhancement feature corresponding to the quantization space with a low bit depth according to the first compressed image data and the second compressed image data. In one implementation, perform denoising processing on the first compressed image data through a pre-trained neural network algorithm, and determine the optimized second compressed image data of the first compressed image data through wide dynamic range enhancement and low-light enhancement operations, and determine the difference between the first compressed image data and the second compressed image data as the first spatial enhancement feature corresponding to the quantization space with a low bit depth, where the first spatial enhancement feature is the noise of the image in the quantization space with a low bit depth.

[0027] In practical applications, after receiving the image data compressed to 8 bits perform denoising, wide dynamic range enhancement, or low-light enhancement operations on the compressed image through a pre-trained neural network algorithm to generate a processed image :

[0028] Obtain the estimated 8-bit spatial enhancement feature (estimated noise map or enhanced details) by taking the difference between the image data before and after processing :

[0029] Step S106: Restore the first spatial enhancement feature to the quantization space with a high bit depth through inverse mapping processing of the first spatial enhancement feature to determine the second spatial enhancement feature corresponding to the quantization space with a high bit depth, and determine the target image data according to the original image data and the second spatial enhancement feature. When performing inverse mapping, the first spatial enhancement feature can be adaptively enlarged or scaled in combination with the Zhuang Anqi characteristic. In one implementation, the difference between the original image data and the second spatial enhancement feature can be determined as the optimized image data with a high bit depth, and Gamma correction, demosaicing processing, and color space conversion processing are performed on the optimized image data to determine the target image data.

[0030] In practical applications, the features estimated in the 8-bit space can be restored to the original high-bit space through the inverse mapping or extension function corresponding to the previous mapping function to obtain the features of the original high-bit space :

[0031] The extended functions adopted can correspond one by one to the compression functions, or can be adaptively adjusted according to the sensor characteristics. For example, different extension strategies can be used in the highlight area or the dark area, and the features restored from different channels of the RAW image can be processed separately to better meet the requirements of the actual scene.

[0032] Furthermore, when generating the high-bit optimized result, the original high-bit depth image can be subtracted from the restored enhanced feature to obtain the high-bit depth optimized image (i.e., the target image data):

[0033] Among them, the result can be further post-processed by ISP (Image Signal Processor) according to the application scenario in combination with the brightness and color characteristics, such as sharpening, color correction, etc. In terms of the processing flow, there are two choices for the input data: one is to directly input the data in the stream format into the ISP processing pipeline to complete the entire processing process; the other is to first store the data in an independent storage space such as DDR (Double Data Rate), and then input it into the ISP processing pipeline for processing as needed.

[0034] The above image processing method based on bit-width compression and restoration provided by the embodiments of the present invention can significantly save the consumption of computing resources during image processing.

[0035] See Figure 2 As shown in the flowchart of an image compression method, the embodiments of the present invention also provide an implementation manner for compressing an image from a high-bit depth to a low-bit depth, specifically as follows (1) to (3): (1) Obtain the brightness, sharpness, and sensor noise distribution of the original image data. Among them, the mapping function of linear or non-linear bit-width compression can set different quantization precisions in different gray ranges according to the image brightness, sharpness, and sensor noise distribution, etc.

[0036] (2) If uneven brightness, sharpness, or sensor noise distribution is detected, image compression is performed through non-linear mapping. In one implementation, the non-linear mapping can be adjusted according to information such as image brightness, noise distribution, and image sharpness. For example, in the dark area (low pixel value), a denser quantization distribution is used to retain the characteristics of low-light noise; in the bright area, a moderately sparse quantization is used to save bit space. Specifically, if the brightness, sharpness, or sensor noise distribution is uneven, through non-linear mapping, different quantization precisions are set in different gray-scale ranges of the image, and the original image data is compressed based on the different quantization precisions to obtain the first compressed image data: in the dark area with low pixel values, a dense quantization distribution is used through non-linear mapping to retain the characteristics of low-light noise; in the bright area with high pixel values, a sparse quantization distribution is used through non-linear mapping to save bit space.

[0037] In one implementation, the high-bit-depth image data includes, but is not limited to, Bayer array data in RAW format. If the original image has multiple channels (such as the R, Gr, Gb, and B channels of a Bayer image), different bit-width compression and inverse mapping strategies can also be adopted for different sub-channels (R, Gr, Gb, B) in the Bayer array.

[0038] (3) If uneven brightness, sharpness, and sensor noise distribution are not detected, image compression is performed through linear mapping. In one implementation, the value range of the original image data is obtained, and through a preset linear mapping model, based on the value range of the original image data, the original image data is mapped to a low-bit-depth quantization space to determine the first compressed image data, where the linear mapping model is expressed as:

[0039] where, represents the original image data, which is a value within the range of , represents the minimum value of the original image data, represents the maximum value of the original image data, is the first compressed image data, representing the mapped 8-bit value, and the 8-bit range is between 0 and 255.

[0040] In summary, in the efficient denoising solution provided by the present invention, the pre - processing of the entire process is equivalent to only processing the noise. The image information is still the original high - bit - depth image, retaining the effective information of the original high - bit - depth. If it is compressed to a low - bit - depth and then image processing is performed, linear compression will seriously affect the image quality, and non - linear compression will affect the noise distribution. Therefore, the present invention can significantly save the consumption of computing resources while ensuring the quality of image processing. In addition, the present invention can also be deployed on mobile devices or embedded systems to reduce the computing power requirements and take into account the high - quality optimization effect.

[0041] For the image processing method based on bit - width compression and restoration provided in the foregoing embodiments, the embodiments of the present invention provide an image processing apparatus based on bit - width compression and restoration. Refer to Figure 3 the structural schematic diagram of an image processing apparatus based on bit - width compression and restoration shown in The bit - width compression module 302 maps the original high - bit - depth image data to a low - bit - depth quantization space through linear mapping to obtain the first compressed image data with a low bit - depth after compression; The network processing module 304 performs image processing on the first compressed image data through a pre - trained neural network algorithm, determines the second compressed image data, and determines the first spatial enhancement feature corresponding to the low - bit - depth quantization space according to the first compressed image data and the second compressed image data; The feature restoration module 306 restores the first spatial enhancement feature to a high - bit - depth quantization space through inverse mapping processing, determines the second spatial enhancement feature corresponding to the high - bit - depth quantization space, and determines the target image data according to the original image data and the second spatial enhancement feature.

[0042] The above - mentioned image processing apparatus based on bit - width compression and restoration provided by the embodiments of the present application can significantly save the consumption of computing resources during image processing.

[0043] In one implementation, when performing the step of mapping the original high - bit - depth image data to a low - bit - depth quantization space through linear mapping to obtain the first compressed image data with a low bit - depth after compression, the above - mentioned bit - width compression module 302 is further configured to: obtain the value range of the original image data, and based on the value range of the original image data, map the original image data to the low - bit - depth quantization space through a preset linear mapping model to determine the first compressed image data, where the linear mapping model is expressed as:

[0044] where represents the original image data, represents the minimum value of the original image data, represents the maximum value of the original image data, is the first compressed image data, representing the mapped 8-bit value, and the 8-bit range is between 0 and 255.

[0045] In one implementation, when performing the step of mapping the original image data with high bit depth to the quantization space with low bit depth through linear mapping to obtain the first compressed image data with low bit depth after compression, the above bit width compression module 302 is further configured to: obtain the brightness, sharpness, and sensor noise distribution of the original image data; if the brightness, sharpness, or sensor noise distribution is uneven, through non-linear mapping, set different quantization precisions in different gray ranges of the image, and compress the original image data based on the different quantization precisions to obtain the first compressed image data.

[0046] In one implementation, when performing the step of setting different quantization precisions in different gray ranges of the image through non-linear mapping, the above bit width compression module 302 is further configured to: in the dark area with low pixel values, use a dense quantization distribution through non-linear mapping to retain the weak light noise characteristics; in the bright area with high pixel values, use a sparse quantization distribution through non-linear mapping to save bit space.

[0047] In one implementation, when performing the step of performing image processing on the first compressed image data through a pre-trained neural network algorithm to determine the second compressed image data, the above network processing module 304 is further configured to: perform denoising processing on the first compressed image data through a pre-trained neural network algorithm, and determine the optimized second compressed image data of the first compressed image data through wide dynamic range enhancement and low light enhancement operations.

[0048] In one implementation, when performing the step of determining the first spatial enhancement feature corresponding to the quantization space with low bit depth according to the first compressed image data and the second compressed image data, the above network processing module 304 is further configured to: determine the difference between the first compressed image data and the second compressed image data as the first spatial enhancement feature corresponding to the quantization space with low bit depth, where the first spatial enhancement feature is the noise of the image in the quantization space with low bit depth.

[0049] In one implementation, when performing the step of determining the target image data according to the original image data and the second spatial enhancement feature, the above feature restoration module 306 is further configured to: determine the difference between the original image data and the second spatial enhancement feature as the optimized image data with high bit depth; perform Gamma correction, demosaicing processing, and color space conversion processing on the optimized image data to determine the target image data.

[0050] The device provided by the embodiment of the present invention has the same implementation principle and technical effects as those of the foregoing method embodiment. For the sake of brief description, for the parts not mentioned in the device embodiment, reference may be made to the corresponding content in the foregoing method embodiment.

[0051] The embodiment of the present invention provides a server. Specifically, the server includes a processor and a storage device; a computer program is stored on the storage device, and when the computer program is run by the processor, it executes the method described in any one of the foregoing embodiments.

[0052] Figure 4 FIG. is a schematic structural diagram of a server provided by an embodiment of the present invention. The server 100 includes: a processor 40, a memory 41, a bus 42, and a communication interface 43. The processor 40, the communication interface 43, and the memory 41 are connected through the bus 42. The processor 40 is used to execute an executable module stored in the memory 41, such as a computer program.

[0053] Among them, the memory 41 may include a high-speed random access memory (RAM, Random Access Memory), and may also include a non-volatile memory, such as at least one disk memory. Through at least one communication interface 43 (which can be wired or wireless), a communication connection is realized between the system network element and at least one other network element, and the Internet, wide area network, local area network, metropolitan area network, etc. can be used.

[0054] The bus 42 may be an ISA bus, a PCI bus, an EISA bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. For the sake of convenience of representation, Figure 4 only a bidirectional arrow is used in FIG., but it does not mean that there is only one bus or one type of bus.

[0055] Among them, the memory 41 is used to store a program. After receiving an execution instruction, the processor 40 executes the program. The method executed by the device defined by the flow process disclosed in any one of the foregoing embodiments of the present invention can be applied to the processor 40 or implemented by the processor 40.

[0056] The processor 40 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the integrated logic circuit of the hardware in the processor 40 or the instructions in the form of software. The above-mentioned processor 40 may be a general-purpose processor, including a central processing unit (CPU for short), a network processor (NP for short), etc.; it may also be a digital signal processor (DSP for short), an application specific integrated circuit (ASIC for short), a field-programmable gate array (FPGA for short), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. It can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of the present invention. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in combination with the embodiments of the present invention can be directly embodied as being executed and completed by a hardware decoding processor, or executed and completed by a combination of the hardware and software modules in the decoding processor. The software module may be located in a mature storage medium in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. This storage medium is located in the memory 41, and the processor 40 reads the information in the memory 41 and combines its hardware to complete the steps of the above method.

[0057] The computer program product of the readable storage medium provided by the embodiments of the present invention includes a computer-readable storage medium storing program code, and the instructions included in the program code can be used to execute the method described in the foregoing method embodiments. For the specific implementation, reference can be made to the foregoing method embodiments, and details are not described herein again.

[0058] If the above-described functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs.

[0059] Finally, it should be noted that the above-described embodiments are only specific implementation manners of the present invention, used to illustrate the technical solutions of the present invention, rather than limiting it. The protection scope of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that any person skilled in the art within the technical scope disclosed by the present invention can still modify the technical solutions described in the foregoing embodiments, or can easily think of changes, or perform equivalent replacements on some of the technical features; and these modifications, changes, or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. An image processing method based on bit width compression and restoration, characterized in that: The method comprises: Mapping the original image data of high bit depth to a quantization space of low bit depth by linear mapping to obtain compressed first compressed image data of low bit depth; Performing image processing on the first compressed image data by a pre-trained neural network algorithm to determine second compressed image data, and determining a first spatial enhancement feature corresponding to a quantization space of a low bit depth according to the first compressed image data and the second compressed image data; By performing inverse mapping processing on the first spatial enhancement feature, the first spatial enhancement feature is restored to a quantization space with a high bit depth, a second spatial enhancement feature corresponding to the quantization space with a high bit depth is determined, and the target image data is determined based on the original image data and the second spatial enhancement feature.

2. The image processing method based on bit width compression and restoration according to claim 1, characterized in that: The step of mapping the original image data of high bit depth to a quantization space of low bit depth by linear mapping to obtain compressed first compressed image data of low bit depth comprises: Obtaining a value range of the original image data, and mapping the original image data to a low bit depth quantization space based on the value range of the original image data through a preset linear mapping model to determine the first compressed image data, wherein the linear mapping model is expressed as: in, represents the original image data, Represents the minimum value of the original image data, Indicates the maximum value of the original image data, It is the first compressed image data, which represents the mapped 8-bit value, and the 8-bit range is between 0 and 255.

3. The image processing method based on bit width compression and restoration according to claim 1, characterized in that: The step of mapping the original image data of high bit depth to a quantization space of low bit depth by linear mapping to obtain compressed first compressed image data of low bit depth also includes: Acquire the brightness, clarity and sensor noise distribution of the original image data; If the brightness, clarity or sensor noise is unevenly distributed, differentiated quantization precision is set in different grayscale ranges of the image through nonlinear mapping, and the original image data is compressed based on the differentiated quantization precision to obtain the first compressed image data.

4. The image processing method based on bit width compression and restoration according to claim 3, characterized in that: The steps of setting differentiated quantization precision in different grayscale ranges of the image through nonlinear mapping include: In dark areas with low pixel values, dense quantization distribution is used through nonlinear mapping to preserve low-light noise characteristics; In bright areas with high pixel values, a sparse quantization distribution is used through nonlinear mapping to save bit space.

5. The image processing method based on bit width compression and restoration according to claim 1, characterized in that: The step of performing image processing on the first compressed image data by using a pre-trained neural network algorithm to determine the second compressed image data comprises: The first compressed image data is denoised by using a pre-trained neural network algorithm, and the second compressed image data after the first compressed image data is optimized is determined by performing wide dynamic range enhancement and dark light enhancement operations.

6. The image processing method based on bit width compression and restoration according to claim 1, characterized in that: The step of determining a first spatial enhancement feature corresponding to a quantization space of a low bit depth according to the first compressed image data and the second compressed image data comprises: The difference between the first compressed image data and the second compressed image data is determined as a first spatial enhancement feature corresponding to the low bit depth quantization space, wherein the first spatial enhancement feature is the noise of the image in the low bit depth quantization space.

7. The image processing method based on bit width compression and restoration according to claim 1, characterized in that: The step of determining target image data according to the original image data and the second spatial enhancement feature comprises: Determine the difference between the original image data and the second spatial enhancement feature as optimized image data with a high bit depth; Gamma correction, de-mosaicing and color space conversion are performed on the optimized image data to determine the target image data.

8. An image processing device based on bit width compression and restoration, characterized in that: The device comprises: A bit width compression module maps the original image data of high bit depth to a quantization space of low bit depth through linear mapping to obtain compressed first compressed image data of low bit depth; a network processing module, performing image processing on the first compressed image data through a pre-trained neural network algorithm to determine second compressed image data, and determining a first spatial enhancement feature corresponding to a quantization space of a low bit depth according to the first compressed image data and the second compressed image data; The feature restoration module restores the first spatial enhancement feature to a quantization space with a high bit depth by performing inverse mapping processing on the first spatial enhancement feature, determines a second spatial enhancement feature corresponding to the quantization space with a high bit depth, and determines the target image data based on the original image data and the second spatial enhancement feature.

9. A server, characterized in that: The method comprises a processor and a memory, wherein the memory stores computer executable instructions that can be executed by the processor, and the processor executes the computer executable instructions to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are called and executed by a processor, the computer-executable instructions prompt the processor to implement the method according to any one of claims 1 to 7.

Images