Exposure machine image exposure correction method, system, readable storage medium and computer

Abstract

This invention provides a method, system, readable storage medium, and computer for image exposure correction using an exposure machine. The method includes: acquiring a set of exposed images collected by the exposure machine under different exposure parameters; filtering the underexposed images and inputting the overexposed images and the filtered underexposed images into a pre-trained deep residual network for feature extraction to obtain corresponding multi-level depth features; calculating the information content and corresponding image weights of the overexposed and filtered underexposed images based on the multi-level depth features, constructing a fusion loss function, and performing weighted fusion based on the image weights to obtain a fusion optimization function; optimizing the brightness channels of the overexposed and underexposed images using the fusion optimization function as the optimization objective to obtain the corrected exposed image. This invention, by constructing a fusion optimization function, can enhance adaptability to different exposure conditions and image content.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to an exposure correction method, system, readable storage medium, and computer for an exposure machine image. Background Technology

[0002] As a key piece of equipment in semiconductor manufacturing, the exposure machine's function is to precisely expose the circuit patterns on the photomask onto the wafer surface using an optical system. The exposure quality of the image directly determines the accuracy and resolution of the pattern transfer, significantly impacting subsequent process steps and the final chip performance.

[0003] However, in actual exposure processes, due to factors such as fluctuations in light source intensity, characteristics of the optical path system, environmental interference, and adjustments to process parameters, exposed images often exhibit overexposure or underexposure. Overexposed areas suffer from loss of detail and blurred edges due to excessive illumination, while underexposed areas suffer from low signal-to-noise ratio and unclear textures due to insufficient illumination. These exposure anomalies reduce the contrast and clarity of the pattern, leading to defects such as linewidth deviation and image distortion, severely impacting chip manufacturing yield and reliability.

[0004] Currently, the correction of exposed images mostly employs traditional image enhancement methods, such as histogram equalization, gamma correction, and local contrast stretching. While these methods can improve the visual effect of images to some extent, they often struggle to simultaneously restore details in overexposed areas and suppress noise in underexposed areas. Furthermore, they are prone to introducing artifacts or losing important structural information during the adjustment process. In addition, traditional methods rely heavily on manually setting parameters and lack the ability to adapt to image content, making it difficult to meet the high-precision, high-consistency exposure image correction requirements in semiconductor manufacturing. Summary of the Invention

[0005] Based on this, the purpose of the present invention is to provide an exposure correction method, system, readable storage medium, and computer for an exposure machine, so as to at least solve the shortcomings of the above-mentioned technologies.

[0006] This invention proposes a method for image exposure correction using an exposure machine, comprising: The set of exposure images acquired by the exposure machine under different exposure parameters includes at least one overexposed image and one underexposed image. The underexposed image is filtered, and the overexposed image and the filtered underexposed image are input into a pre-trained deep residual network for feature extraction to obtain corresponding multi-level deep features. Based on the multi-level deep features, the information content and corresponding image weights of the overexposed image and the filtered underexposed image are calculated, and a fusion loss function is constructed. The image weights are then used for weighted fusion to obtain a fusion optimization function. Using the fusion optimization function as the optimization target, the brightness channels of the overexposed image and the underexposed image are optimized to obtain the corrected exposure image.

[0007] Furthermore, the steps of filtering the underexposed image and inputting the overexposed image and the filtered underexposed image into a pre-trained deep residual network for feature extraction to obtain the corresponding multi-level deep features include: A corresponding guide image is constructed based on the underexposed image, and linear processing is performed on the guide image to obtain the corresponding linear coefficients. The linear coefficients are then used to construct a local window of the guide image. A linear transformation is performed on the guide image within the local window to obtain the corresponding output image; The output image and the overexposed image are processed by a pre-trained deep residual network to capture the shallow edge texture and deep semantic structure of the image, so as to form corresponding multi-level deep features.

[0008] Furthermore, the steps of calculating the information content and corresponding image weights of the overexposed image and the filtered underexposed image based on the multi-level depth features, constructing a fusion loss function, and performing weighted fusion based on the image weights to obtain a fusion optimization function include: The information content of the overexposed image and the underexposed image after filtering is calculated based on the multi-level depth features. After scaling each information content, it is normalized by a normalization function to obtain the corresponding image weight. A fusion loss function is constructed, and the image weights are used as weighting coefficients to train and optimize the fusion loss function to obtain the corresponding fusion optimization function.

[0009] Furthermore, the formula for calculating the information content is as follows:

[0010] In the formula, The number of layers representing multi-level deep features. , , They represent the first and second digits of the pre-trained deep residual network, respectively. The height, width, and number of channels of all feature maps in each feature layer. Overexposed image In the pre-trained deep residual network On the feature layer, the th Each feature map Represents the gradient operator. This represents calculating the square root of the sum of the squares of all elements in a matrix; The formula for calculating the image weight is:

[0011] In the formula, Represents the normalization function. , The overexposed images are respectively Information content and the underexposed image after filtering Information level , They are respectively Scaling factor and The scaling factor.

[0012] Furthermore, the step of optimizing the brightness channels of the overexposed and underexposed images using the fusion optimization function as the optimization target to obtain the corrected exposure image includes: Convert the RGB color space of the overexposed image and the underexposed image to the YCbCr color space; Using the Y color space of the overexposed image and the underexposed image as input, and the fusion optimization function as the optimization objective, end-to-end training is performed in the image fusion network to output the corresponding fused image; The Cb and Cr color spaces of the exposed image and the underexposed image are fused, and the fused image is combined with the fused image and the color space is converted to obtain the corrected exposed image.

[0013] The present invention also proposes an image exposure correction system for an exposure machine, comprising: The image acquisition module is used to acquire a set of exposed images acquired by the exposure machine under different exposure parameters, wherein the set of exposed images includes at least one overexposed image and one underexposed image; The feature extraction module is used to filter the underexposed image and input the overexposed image and the filtered underexposed image into a pre-trained deep residual network for feature extraction to obtain the corresponding multi-level deep features. The weighted fusion module is used to calculate the information content and corresponding image weights of the overexposed image and the filtered underexposed image based on the multi-level depth features, and to construct a fusion loss function. The image weights are then used for weighted fusion to obtain a fusion optimization function. The image correction module is used to optimize the brightness channels of the overexposed image and the underexposed image using the fusion optimization function as the optimization target, so as to obtain a corrected exposure image.

[0014] Furthermore, the feature extraction module is specifically used for: A corresponding guide image is constructed based on the underexposed image, and linear processing is performed on the guide image to obtain the corresponding linear coefficients. The linear coefficients are then used to construct a local window of the guide image. A linear transformation is performed on the guide image within the local window to obtain the corresponding output image; The output image and the overexposed image are processed by a pre-trained deep residual network to capture the shallow edge texture and deep semantic structure of the image, so as to form corresponding multi-level deep features.

[0015] Furthermore, the weighted fusion module is specifically used for: The information content of the overexposed image and the underexposed image after filtering is calculated based on the multi-level depth features. After scaling each information content, it is normalized by a normalization function to obtain the corresponding image weight. A fusion loss function is constructed, and the image weights are used as weighting coefficients to train and optimize the fusion loss function to obtain the corresponding fusion optimization function.

[0016] Furthermore, the image correction module is specifically used for: Convert the RGB color space of the overexposed image and the underexposed image to the YCbCr color space; Using the Y color space of the overexposed image and the underexposed image as input, and the fusion optimization function as the optimization objective, end-to-end training is performed in the image fusion network to output the corresponding fused image; The Cb and Cr color spaces of the exposed image and the underexposed image are fused, and the fused image is combined with the fused image and the color space is converted to obtain the corrected exposed image.

[0017] The present invention also proposes a storage medium storing a computer program that, when executed by a processor, implements the above-described exposure correction method for exposure machine images.

[0018] The present invention also proposes a computer, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described exposure correction method for exposure machine images.

[0019] The exposure correction method, system, readable storage medium, and computer of this invention acquire overexposed and underexposed images, and effectively separate and extract effective information from the two types of images using filtering and depth feature extraction. This achieves accurate identification and recovery of exposure-abnormal areas, improving the overall visual effect and information integrity of the image. A pre-trained deep residual network is used for multi-level depth feature extraction, which can simultaneously capture the shallow edge texture and deep semantic structure of the image, providing richer and more discriminative feature representations for exposure correction. Through information content calculation and adaptive weight allocation, intelligent fusion of overexposed and underexposed image information is achieved, avoiding information loss or artifact problems. By constructing a fusion optimization function, the adaptability to different exposure conditions and image content is enhanced, maintaining good correction effects even under complex exposure conditions, and exhibiting strong robustness and generalization ability. Attached Figure Description

[0020] Figure 1 This is a flowchart of the exposure correction method for an exposure machine image according to the first embodiment of the present invention; Figure 2 This is a structural block diagram of the exposure correction system for an exposure machine in the second embodiment of the present invention; Figure 3 This is a structural block diagram of the computer in the third embodiment of the present invention.

[0021] The following detailed description, in conjunction with the accompanying drawings, will further illustrate the present invention. Detailed Implementation

[0022] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Several embodiments of the invention are illustrated in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.

[0023] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the specification of this invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0024] Example 1 Please see Figure 1 The image shown is an exposure correction method for an exposure machine in the first embodiment of the present invention, the method specifically including steps S101 to S104: S101, Collect a set of exposure images acquired by the exposure machine under different exposure parameters, wherein the set of exposure images includes at least one overexposed image and one underexposed image; In practice, the exposure set of the exposure machine is collected from the same mask or wafer under different exposure parameters. The exposure set includes at least one underexposed image and one overexposed image.

[0025] S102, the underexposed image is filtered, and the overexposed image and the filtered underexposed image are input into a pre-trained deep residual network for feature extraction to obtain the corresponding multi-level deep features; Furthermore, step S102 specifically includes steps S1021 to S1023: S1021, construct a corresponding guide image based on the underexposed image, and perform linear processing on the guide image to obtain the corresponding linear coefficients, and use the linear coefficients to construct a local window of the guide image; S1022, A linear transformation is performed on the guide image in the local window to obtain the corresponding output image; S1023, The output image and the overexposed image are processed by a pre-trained deep residual network to capture the shallow edge texture and deep semantic structure of the image to form corresponding multi-level deep features.

[0026] In practice, a corresponding guide image is constructed based on the underexposed image, and linear processing is performed on the guide image to obtain corresponding linear coefficients. A local window of the guide image is then constructed using these linear coefficients, and a linear transformation is performed on the guide image within this local window to obtain the corresponding output image. It can be understood that this output image is a pixel-based representation of the guide image. The central window Linear transformation within:

[0027] In the formula, Indicates pixel index, , for The linear coefficients in Indicates that it contains pixel indexes The output image, Indicates that it contains pixel indexes The guide image.

[0028] To ensure that the difference between the output image and the underexposed image is minimized, a cost function is constructed to solve for the linear coefficients mentioned above:

[0029] In the formula, This is a regularization parameter used to prevent coefficients from being regularized. Too large Indicates that it contains pixel indexes An underexposed image.

[0030] Furthermore, a pre-trained deep residual network (in this embodiment, the deep residual network adopts the ResNet-50 architecture, ResNet-18 architecture, ResNet-34 architecture, Inception-ResNet-v2 architecture, etc.) is used to process the output image and the overexposed image respectively, capturing the shallow edge texture and deep semantic structure of the image to form corresponding multi-level deep features.

[0031] S103, calculate the information content and corresponding image weights of the overexposed image and the underexposed image after filtering based on the multi-level depth features, and construct a fusion loss function. Then, perform weighted fusion based on the image weights to obtain a fusion optimization function. Furthermore, step S103 specifically includes steps S1031 to S1032: S1031, the information content of the overexposed image and the underexposed image after filtering is calculated based on the multi-level depth features, and the information content is scaled and normalized by a normalization function to obtain the corresponding image weight. S1032, Construct a fusion loss function, and train and optimize the fusion loss function using the image weights as weighting coefficients to obtain the corresponding fusion optimization function.

[0032] In practice, to measure the information content of the depth features obtained above, the information content of the overexposed image and the underexposed image after filtering is calculated from the obtained multi-level depth features:

[0033] In the formula, The number of layers representing multi-level deep features. , , They represent the first and second digits of the pre-trained deep residual network, respectively. The height, width, and number of channels of all feature maps in each feature layer. Overexposed image In the pre-trained deep residual network On the feature layer, the th Each feature map Represents the gradient operator. This represents calculating the square root of the sum of the squares of all elements in a matrix; Furthermore, since information content is an absolute value, it cannot reflect the difference between overexposed and underexposed images. Therefore, after scaling each information content, a normalization function is used to normalize it, comparing and quantifying the relative magnitudes of information content between underexposed and overexposed images to obtain image weights.

[0034] In the formula, Represents the normalization function. , The first Zhang Guo's exposed image Information level and the first Underexposed image after Zhang filtering Information level , They are respectively Scaling factor and The scaling factor.

[0035] Specifically, a fusion loss function is constructed by weighted fusion based on image weights to obtain a fusion optimization function. In this embodiment, the fusion loss function is constructed by combining structural similarity constraints and pixel intensity constraints.

[0036] In the formula, Represents the fusion loss function. This represents the set of training parameters in an image fusion network. Represents structural similarity loss. This represents the weighted mean square error loss. , They are respectively , The weighting coefficients.

[0037] The formula for calculating structural similarity loss is as follows:

[0038]

[0039] In the formula, This represents the fused image output by the fusion network. Represents fused images With the Zhang Guo's exposed image Structural similarity index between them Represents fused images With the Zhang Qian's exposed image Structural similarity index between them Represents fused images With the Zhang Guo's exposed image The mean square error between them Represents fused images With the Zhang Qian's exposed image The mean square error between them.

[0040] S104, using the fusion optimization function as the optimization target, optimize the brightness channels of the overexposed image and the underexposed image to obtain the corrected exposure image.

[0041] Furthermore, step S104 specifically includes steps S1041 to S1043: S1041, convert the RGB color space of the overexposed image and the underexposed image to the YCbCr color space; S1042, using the Y color space of the overexposed image and the underexposed image as input, and the fusion optimization function as the optimization target, end-to-end training is performed in the image fusion network to output the corresponding fused image; S1043, the Cb color space and Cr color space of the exposed image and the underexposed image are fused, and the fused image is combined with the fused image and the color space is converted to obtain the corrected exposed image.

[0042] In practical implementation, DenseNet is used to construct an image fusion network. The GB color space of overexposed and underexposed images is converted to the YCbCr color space, and the Y color space of the two images is used as input. The image fusion network is trained end-to-end with the fusion optimization function as the optimization objective. The output of each layer in the image fusion network is passed to the next layer as input and is also directly connected to the input of all subsequent layers. The dense connection ensures that each layer can receive input from all previous layers, thereby promoting feature propagation and reuse to output the corresponding fused image. It is understandable that since the Y color space contains more structural details and brightness change information required in the image fusion task, using DenseNet can increase the reusability of features and improve the performance of the network. Specifically, the Cb and Cr color spaces of the exposed and underexposed images are fused, and the fused image is combined with the original image and its color space is converted to obtain the corrected exposed image.

[0043] In summary, the exposure correction method for exposure machines described in the above embodiments of the present invention acquires overexposed and underexposed images, and effectively separates and extracts effective information from the two types of images using filtering and depth feature extraction. This achieves accurate identification and recovery of exposure-abnormal areas, improving the overall visual effect and information integrity of the image. The method employs a pre-trained deep residual network for multi-level depth feature extraction, which can simultaneously capture the shallow edge texture and deep semantic structure of the image, providing richer and more discriminative feature representations for exposure correction. Through information content calculation and adaptive weight allocation, intelligent fusion of overexposed and underexposed image information is achieved, avoiding information loss or artifact problems. By constructing a fusion optimization function, the method enhances adaptability to different exposure conditions and image content, maintaining good correction effects even under complex exposure conditions, and exhibiting strong robustness and generalization ability.

[0044] Example 2 In another aspect, this invention also proposes an image exposure correction system for an exposure machine; please refer to [link / reference needed]. Figure 2 The image shown is an exposure correction system for an exposure machine according to a second embodiment of the present invention. The system includes: Image acquisition module 11 is used to acquire a set of exposed images acquired by the exposure machine under different exposure parameters, wherein the set of exposed images includes at least one overexposed image and one underexposed image; The feature extraction module 12 is used to filter the underexposed image and input the overexposed image and the filtered underexposed image into a pre-trained deep residual network for feature extraction to obtain corresponding multi-level deep features. The weighted fusion module 13 is used to calculate the information content and corresponding image weights of the overexposed image and the filtered underexposed image based on the multi-level depth features, and to construct a fusion loss function. The weighted fusion is performed based on the image weights to obtain a fusion optimization function. The image correction module 14 is used to optimize the brightness channels of the overexposed image and the underexposed image by using the fusion optimization function as the optimization target, so as to obtain the corrected exposure image.

[0045] Furthermore, the feature extraction module 12 is specifically used for: A corresponding guide image is constructed based on the underexposed image, and linear processing is performed on the guide image to obtain the corresponding linear coefficients. The linear coefficients are then used to construct a local window of the guide image. A linear transformation is performed on the guide image within the local window to obtain the corresponding output image; The output image and the overexposed image are processed by a pre-trained deep residual network to capture the shallow edge texture and deep semantic structure of the image, so as to form corresponding multi-level deep features.

[0046] Furthermore, the weighted fusion module 13 is specifically used for: The information content of the overexposed image and the underexposed image after filtering is calculated based on the multi-level depth features. After scaling each information content, it is normalized by a normalization function to obtain the corresponding image weight. A fusion loss function is constructed, and the image weights are used as weighting coefficients to train and optimize the fusion loss function to obtain the corresponding fusion optimization function.

[0047] Furthermore, the image correction module 14 is specifically used for: Convert the RGB color space of the overexposed image and the underexposed image to the YCbCr color space; Using the Y color space of the overexposed image and the underexposed image as input, and the fusion optimization function as the optimization objective, end-to-end training is performed in the image fusion network to output the corresponding fused image; The Cb and Cr color spaces of the exposed image and the underexposed image are fused, and the fused image is combined with the fused image and the color space is converted to obtain the corrected exposed image.

[0048] The functions or operation steps implemented by the above modules and units are largely the same as those in the above method embodiments, and will not be repeated here.

[0049] The exposure correction system for an exposure machine provided in this embodiment of the invention has the same implementation principle and technical effect as the aforementioned method embodiment. For the sake of brevity, any parts not mentioned in the system embodiment can be referred to the corresponding content in the aforementioned method embodiment.

[0050] Example 3 This invention also proposes a computer, please refer to [link / reference]. Figure 3 The computer shown in the third embodiment of the present invention includes a memory 10, a processor 20, and a computer program 30 stored in the memory 10 and executable on the processor 20. When the processor 20 executes the computer program 30, it implements the above-described exposure correction method for exposure machine images.

[0051] The memory 10 includes at least one type of storage medium, such as flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 10 can be an internal storage unit of a computer, such as the computer's hard disk. In other embodiments, the memory 10 can be an external storage device, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. Furthermore, the memory 10 can include both internal and external storage units of the computer. The memory 10 can be used not only to store application software and various types of data installed on the computer, but also to temporarily store data that has been output or will be output.

[0052] In some embodiments, the processor 20 may be an electronic control unit (ECU), a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip, used to run program code stored in the memory 10 or process data, such as executing access restriction programs.

[0053] It should be pointed out that, Figure 3 The structure shown does not constitute a limitation on the computer. In other embodiments, the computer may include fewer or more components than shown, or combine certain components, or have different component arrangements.

[0054] This invention also proposes a storage medium storing a computer program that, when executed by a processor, implements the exposure correction method for exposure machine images as described above.

[0055] Those skilled in the art will understand that the logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can mean any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0056] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0057] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0058] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0059] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for image exposure correction using an exposure machine, characterized in that, include: The set of exposure images acquired by the exposure machine under different exposure parameters includes at least one overexposed image and one underexposed image. The underexposed image is filtered, and the overexposed image and the filtered underexposed image are input into a pre-trained deep residual network for feature extraction to obtain corresponding multi-level deep features. Based on the multi-level deep features, the information content and corresponding image weights of the overexposed image and the filtered underexposed image are calculated, and a fusion loss function is constructed. The image weights are then used for weighted fusion to obtain a fusion optimization function. Using the fusion optimization function as the optimization target, the brightness channels of the overexposed image and the underexposed image are optimized to obtain the corrected exposure image.

2. The exposure correction method for an exposure machine image according to claim 1, characterized in that, The steps of filtering the underexposed image and inputting the overexposed image and the filtered underexposed image into a pre-trained deep residual network for feature extraction to obtain the corresponding multi-level deep features include: A corresponding guide image is constructed based on the underexposed image, and linear processing is performed on the guide image to obtain the corresponding linear coefficients. The linear coefficients are then used to construct a local window of the guide image. A linear transformation is performed on the guide image within the local window to obtain the corresponding output image; The output image and the overexposed image are processed by a pre-trained deep residual network to capture the shallow edge texture and deep semantic structure of the image, so as to form corresponding multi-level deep features.

3. The exposure correction method for an exposure machine image according to claim 1, characterized in that, The steps of calculating the information content and corresponding image weights of the overexposed image and the filtered underexposed image based on the multi-level deep features, constructing a fusion loss function, and performing weighted fusion based on the image weights to obtain the fusion optimization function include: The information content of the overexposed image and the underexposed image after filtering is calculated based on the multi-level depth features. After scaling each information content, it is normalized by a normalization function to obtain the corresponding image weight. A fusion loss function is constructed, and the image weights are used as weighting coefficients to train and optimize the fusion loss function to obtain the corresponding fusion optimization function.

4. The exposure correction method for an exposure machine image according to claim 3, characterized in that, The formula for calculating the information content is: In the formula, The number of layers representing multi-level deep features. , , They represent the first and second digits of the pre-trained deep residual network, respectively. The height, width, and number of channels of all feature maps in each feature layer. Overexposed image In the pre-trained deep residual network On the feature layer, the th Each feature map Represents the gradient operator. This represents calculating the square root of the sum of the squares of all elements in a matrix; The formula for calculating the image weight is: In the formula, Represents the normalization function. , The overexposed images are respectively Information content and the underexposed image after filtering Information level , They are respectively Scaling factor and The scaling factor.

5. The exposure correction method for an exposure machine image according to claim 1, characterized in that, The steps of optimizing the brightness channels of the overexposed and underexposed images using the fusion optimization function as the optimization target to obtain the corrected exposure image include: Convert the RGB color space of the overexposed image and the underexposed image to the YCbCr color space; Using the Y color space of the overexposed image and the underexposed image as input, and the fusion optimization function as the optimization objective, end-to-end training is performed in the image fusion network to output the corresponding fused image; The Cb and Cr color spaces of the exposed image and the underexposed image are fused, and the fused image is combined with the fused image and the color space is converted to obtain the corrected exposed image.

6. An image exposure correction system for an exposure machine, characterized in that, include: The image acquisition module is used to acquire a set of exposed images acquired by the exposure machine under different exposure parameters, wherein the set of exposed images includes at least one overexposed image and one underexposed image; The feature extraction module is used to filter the underexposed image and input the overexposed image and the filtered underexposed image into a pre-trained deep residual network for feature extraction to obtain the corresponding multi-level deep features. The weighted fusion module is used to calculate the information content and corresponding image weights of the overexposed image and the filtered underexposed image based on the multi-level depth features, and to construct a fusion loss function. The image weights are then used for weighted fusion to obtain a fusion optimization function. The image correction module is used to optimize the brightness channels of the overexposed image and the underexposed image using the fusion optimization function as the optimization target, so as to obtain a corrected exposure image.

7. The image exposure correction system for an exposure machine according to claim 6, characterized in that, The feature extraction module is specifically used for: A corresponding guide image is constructed based on the underexposed image, and linear processing is performed on the guide image to obtain the corresponding linear coefficients. The linear coefficients are then used to construct a local window of the guide image. A linear transformation is performed on the guide image within the local window to obtain the corresponding output image; The output image and the overexposed image are processed by a pre-trained deep residual network to capture the shallow edge texture and deep semantic structure of the image, so as to form corresponding multi-level deep features.

8. The image exposure correction system for an exposure machine according to claim 6, characterized in that, The weighted fusion module is specifically used for: The information content of the overexposed image and the underexposed image after filtering is calculated based on the multi-level depth features. After scaling each information content, it is normalized by a normalization function to obtain the corresponding image weight. A fusion loss function is constructed, and the image weights are used as weighting coefficients to train and optimize the fusion loss function to obtain the corresponding fusion optimization function.

9. A readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the exposure correction method for an exposure machine image as described in any one of claims 1 to 5.

10. A computer comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the exposure correction method for an exposure machine image as described in any one of claims 1 to 5.

Images