Enhancement Method, Device, Storage Medium and Computer Equipment for Low-illumination Images

Abstract

The present invention discloses a method, apparatus, storage medium and computer device for enhancing low-illumination images, which relates to the technical field of image processing and mainly aims to improve the enhancement efficiency and enhancement effect of low-illumination images. The method includes: acquiring a low-illumination image to be enhanced; determining an adjustment parameter curve for dynamically adjusting the low-illumination image to be enhanced; and using the adjustment parameter curve to dynamically adjust the low-illumination image to be enhanced to obtain an enhanced low-illumination image. The present invention is mainly applicable to the application scenario of image enhancement.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a method, device, storage medium and computer equipment for enhancing low-illumination images. Background Art

[0002] As a storage medium, images record a wealth of important information, and the integrity of the image data is paramount. However, in real-world environments such as nighttime shooting and low lighting conditions, or when poor-quality equipment or lens blur is used, low-light images can suffer from significant information loss. This can lead to varying degrees of image quality issues, such as low visibility, low contrast, and color distortion. These issues severely impact image quality and complicate subsequent image processing. This is particularly true in machine vision applications such as online recognition, inspection, and measurement, where the degradation of low-light images can severely impact application accuracy. Therefore, low-light image quality enhancement is necessary.

[0003] Currently, low-light images are usually enhanced manually. However, this manual processing method is time-consuming and labor-intensive, and due to the uneven technical level of workers, over-enhancement or under-enhancement may occur. Summary of the Invention

[0004] The present invention provides a low-illumination image enhancement method, device, storage medium and computer equipment, which are mainly capable of improving the enhancement efficiency and enhancement effect of the image quality of low-illumination images.

[0005] According to a first aspect of the present invention, a method for enhancing a low-light image is provided, comprising:

[0006] Acquire a low-light image to be enhanced;

[0007] Determining an adjustment parameter curve for performing dynamic range adjustment on the low-illumination image to be enhanced;

[0008] The dynamic range of the low-light image to be enhanced is adjusted using the adjustment parameter curve to obtain an enhanced low-light image.

[0009] Optionally, determining an adjustment parameter curve for adjusting the dynamic range of the low-illumination image to be enhanced includes:

[0010] Obtaining a preset parameter curve prediction model, wherein the preset parameter curve prediction model includes an encoder for extracting image features and a decoder for predicting a parameter curve;

[0011] Input the low - illumination image to be enhanced into the preset parameter curve prediction model. Extract image features from the low - illumination image to be enhanced through the encoder to obtain image features. Predict a curve for the image features through the decoder to obtain an adjustment parameter curve for adjusting the dynamic range of the low - illumination image to be enhanced.

[0012] Optionally, the step of predicting a curve for the image features through the decoder to obtain an adjustment parameter curve for adjusting the dynamic range of the low - illumination image to be enhanced includes:

[0013] Determine the reference images corresponding to different degradation types The corresponding reference degraded image features , and determine the key of the image features , value , and feature dimension ;

[0014] Based on the reference degraded image features , the key , the value , and the feature dimension , respectively determine the matching weights between the low - illumination image to be enhanced and different degradation types , where , is the normalized exponential function;

[0015] Based on different matching weights and the value of the image features , determine the image features with the target degradation type as the degraded image features , where , is the total number of different degradation types;

[0016] Predict a curve for the degraded image features through the decoder to obtain an adjustment parameter curve corresponding to the target degradation type.

[0017] Optionally, before obtaining the preset parameter curve prediction model, the method further includes:

[0018] Construct a preset initial parameter curve prediction model;

[0019] Obtain a sample data set, where the sample data set includes multiple low - illumination sample images and the standard adjustment parameter curves corresponding to the low - illumination sample images;

[0020] Divide the sample data set into a training set and a test set, use the training set to train the preset initial parameter curve prediction model, use the test set to test the trained preset initial parameter curve prediction model, and use the preset initial parameter curve prediction model that meets the test conditions as the preset parameter curve prediction model.

[0021] Optionally, using the adjustment parameter curve to perform dynamic range adjustment on the low-light image to be enhanced to obtain an enhanced low-light image, including:

[0022] Determine each pixel value in the low-light image to be enhanced ;

[0023] Use the adjustment parameter curve to remap each pixel value to obtain each remapped pixel value , where , is the curve parameter of the adjustment parameter curve, and each remapped pixel value constitutes the enhanced low-light image.

[0024] Optionally, before using the adjustment parameter curve to perform dynamic range adjustment on the low-light image to be enhanced to obtain an enhanced low-light image, the method further includes:

[0025] Obtain a verification image set for verifying the image adjustment ability of the adjustment parameter curve, where the verification image set includes verification low-light images and standard enhanced images corresponding to the verification low-light images;

[0026] Use the adjustment parameter curve to perform dynamic range adjustment on the verification low-light image to obtain a predicted enhanced image;

[0027] Determine the brightness difference, color difference, and structural difference between the standard enhanced image and the predicted enhanced image respectively, and determine the brightness loss function corresponding to the brightness difference, the color loss function corresponding to the color difference, and the structural loss function corresponding to the structural difference;

[0028] Based on the brightness loss function, the color loss function, and the structural loss function, determine a joint loss function, and based on the joint loss function, iteratively update the curve parameters in the adjustment parameter curve to obtain the adjustment parameter curve with updated parameters;

[0029] The using the adjustment parameter curve to perform dynamic range adjustment on the low-light image to be enhanced to obtain an enhanced low-light image includes:

[0030] Performing dynamic range adjustment on the low - illumination image to be enhanced by using the adjusted adjustment parameter curve to obtain the enhanced low - illumination image.

[0031] Optionally, determining the adjustment parameter curve for performing dynamic range adjustment on the low - illumination image to be enhanced includes:

[0032] Determining the illumination feature information of the low - illumination image to be enhanced, where the illumination feature information includes the underexposure degree and illumination uniformity of the low - illumination image to be enhanced;

[0033] Based on the illumination feature information, determining the curve constraint conditions for the adjustment parameter curve to be generated, where the curve constraint conditions include the pixel value range of the image adjusted by the adjustment parameter curve to be generated, the monotonicity and reversibility of the adjustment parameter curve to be generated;

[0034] Based on the curve constraint conditions, constructing the adjustment parameter curve for performing dynamic range adjustment on the low - illumination image to be enhanced.

[0035] According to the second aspect of the present invention, there is provided an enhancement device for low - illumination images, including:

[0036] An acquisition unit for acquiring the low - illumination image to be enhanced;

[0037] A determination unit for determining the adjustment parameter curve for performing dynamic range adjustment on the low - illumination image to be enhanced;

[0038] An adjustment unit for performing dynamic range adjustment on the low - illumination image to be enhanced by using the adjustment parameter curve to obtain the enhanced low - illumination image.

[0039] According to the third aspect of the present invention, there is provided a computer - readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the above - mentioned enhancement method for low - illumination images is implemented.

[0040] According to the fourth aspect of the present invention, there is provided a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, and when the processor executes the program, the above - mentioned enhancement method for low - illumination images is implemented.

[0041] An enhancement method, device, storage medium, and computer equipment for low-illumination images provided by the present invention, compared with the current method of manually enhancing low-illumination images, can achieve automatic adjustment of low-illumination images by determining an adjustment parameter curve for dynamically adjusting the low-illumination image to be enhanced and using the adjustment parameter curve to perform dynamic range adjustment on the low-illumination image to be enhanced, avoiding the problem of time-consuming and laborious manual adjustment, thereby improving the enhancement efficiency of low-illumination images; the parameter curve can accurately adjust parameters such as the brightness and contrast of the image, avoiding errors or subjective biases that may occur during manual adjustment, thereby improving the enhancement accuracy of low-illumination images; the results of parameter curve adjustment have high consistency, and the same set of parameters can be applied to different images to ensure the consistency of processing effects; the parameter curve can be saved and reused, facilitating subsequent processing of the same or similar images, while manual adjustment may vary due to different operators. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The drawings described herein are used to provide a further understanding of the present invention, form a part of this application, and the schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0043] Figure 1 The flowchart of an enhancement method for low-illumination images provided by an embodiment of the present invention is shown;

[0044] Figure 2 The flowchart of another enhancement method for low-illumination images provided by an embodiment of the present invention is shown;

[0045] Figure 3 The schematic diagram of a method for generating an adjustment parameter curve provided by an embodiment of the present invention is shown;

[0046] Figure 4 The schematic structural diagram of an enhancement device for low-illumination images provided by an embodiment of the present invention is shown;

[0047] Figure 5 The schematic structural diagram of another enhancement device for low-illumination images provided by an embodiment of the present invention is shown;

[0048] Figure 6 The schematic physical structure diagram of a computer equipment provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0049] The present invention will be described in detail below with reference to the drawings and in combination with embodiments. It should be noted that, without conflict, the embodiments in this application and the features in the embodiments can be combined with each other.

[0050] At present, the method of manually enhancing low - illumination images is time - consuming and laborious. Moreover, due to the uneven technical levels of staff, situations of over - enhancement or under - enhancement may occur.

[0051] To solve the above problems, an embodiment of the present invention provides a method for enhancing low - illumination images, as Figure 1 shown. The method includes:

[0052] 101. Obtain the low - illumination image to be enhanced.

[0053] Among them, the low - illumination image to be enhanced is an image taken under sub - optimal illumination conditions such as insufficient light, low light, strong light, backlight, etc. during night shooting, or when the shooting device has poor quality or the lens is blurred. The low - illumination image has problems such as loss of detail information, low contrast, and narrow dynamic range.

[0054] Specifically, an image taken by a device such as a low - dynamic - range visible - light camera, or a low - illumination image taken under special light conditions such as at night will be stored in the database. When it is necessary to enhance the image quality of the low - illumination image, the corresponding low - illumination image can be directly obtained from the database.

[0055] 102. Determine the adjustment parameter curve for adjusting the dynamic range of the low - illumination image to be enhanced.

[0056] Among them, the dynamic range refers to the range of brightness or gray - scale value differences between the brightest part and the darkest part of an image; through dynamic range adjustment, the lost details in the low - illumination image can be restored, making the image clearer and more natural; dynamic range adjustment includes brightness adjustment, contrast adjustment, hue adjustment, saturation adjustment, etc.

[0057] For the embodiment of the present invention, after obtaining the low - illumination image to be enhanced, it is necessary to determine the parameter curve for dynamic range adjustment of the low - illumination image to be enhanced. Based on this, step 102 specifically includes: determining the illumination feature information of the low - illumination image to be enhanced, where the illumination feature information includes the underexposure degree and illumination uniformity of the low - illumination image to be enhanced; based on the illumination feature information, determining the curve constraint conditions for the adjustment parameter curve to be generated, where the curve constraint conditions include the pixel value range of the image adjusted based on the adjustment parameter curve to be generated, the monotonicity and reversibility of the adjustment parameter curve to be generated; based on the curve constraint conditions, constructing the adjustment parameter curve for adjusting the dynamic range of the low - illumination image to be enhanced.

[0058] Specifically, first, the low-light image to be enhanced is subjected to grayscale histogram statistics. According to the grayscale histogram, the total number of pixels at each grayscale level in the low-light image to be enhanced is counted, and the number of low pixels whose pixel values in the histogram are lower than a preset threshold (where the preset threshold is set according to actual requirements, such as 50) is calculated. The ratio of the number of low pixels to the total number of pixels is used as the underexposure ratio of the low-light image to be enhanced, and the underexposure degree corresponding to this underexposure ratio is determined in a preset underexposure degree configuration table, where different underexposure ratio ranges and their corresponding underexposure degrees are stored in the preset underexposure degree configuration table. For example, if the underexposure ratio is 40%, and the underexposure degree corresponding to the underexposure ratio range of 30% - 50% in the preset underexposure degree configuration table is intermediate level, then the underexposure degree of the low-light image to be enhanced is finally determined to be intermediate level. The higher the level, the more severe the underexposure degree. When determining the illumination uniformity of the low-light image to be enhanced, first, the low-light image to be enhanced is divided into multiple non-overlapping image blocks. The grayscale mean of each image block is calculated, and the variance corresponding to each grayscale mean is determined. Finally, the illumination uniformity of the low-light image to be enhanced is determined based on the variance. For example, the larger the variance, the more uneven the illumination of the low-light image to be enhanced. Further, after determining the illumination feature information such as the underexposure degree and illumination uniformity of the low-light image to be enhanced, according to the illumination feature information, the pixel value range, monotonicity, and reversibility of the adjusted image are set. For example, if the low-light image to be enhanced is underexposed and has uneven illumination, the minimum brightness of the output image is constrained to be 10, and the maximum brightness is 240 (8-bit image) to ensure that the details in the dark part of the image are visible and the bright part is not overexposed. The histogram of the low-light image to be enhanced is adaptively equalized to generate a target histogram distribution, and the output range of the adjustment parameter curve is constrained accordingly. At the same time, the adjustment parameter curve needs to satisfy the monotonicity that the input pixel value and the output pixel value are strictly positively correlated, to avoid situations such as brightness inversion or artifacts. At the same time, it is necessary to ensure that the slope of each segment of the curve in the adjustment parameter curve is greater than zero, and the overall continuity is ensured through interpolation. At the same time, for any output pixel value y, there is a unique input pixel value x corresponding to it to ensure reversible image processing, so as to achieve the reversibility of the curve. Finally, according to the curve constraint conditions, an adjustment parameter curve for dynamically adjusting the low-light image to be enhanced is constructed. To ensure that the adjustment parameter curve is monotonic and gradient reversible, a monotonic curve is used to ensure the contrast of the original image, and gradient reversibility can ensure the optimization of the parameters of the parameter curve, and finally, each pixel value of the image adjusted by the adjustment parameter curve can fall within the required range. For example, each pixel value in the enhancement result falls within the range of [0, 1]. The purpose is to avoid information loss caused by overflow or truncation. The adjustment parameter curve in the embodiment of the present invention can be a quadratic curve.

[0059] Further, after determining the adjustment parameter curve, in order to achieve a higher-order curve and thus improve the dynamic range adjustment ability, it is also necessary to continuously perform parameter iterative updates on the adjustment parameter curve. Based on this, the method includes: obtaining a verification image set for verifying the image adjustment ability of the adjustment parameter curve, where the verification image set includes verification low-illumination images and standard enhanced images corresponding to the verification low-illumination images; using the adjustment parameter curve to perform dynamic range adjustment on the verification low-illumination images to obtain predicted enhanced images; respectively determining the brightness difference, color difference, and structural difference between the standard enhanced images and the predicted enhanced images, and determining a brightness loss function corresponding to the brightness difference, a color loss function corresponding to the color difference, and a structural loss function corresponding to the structural difference; based on the brightness loss function, the color loss function, and the structural loss function, determining a joint loss function, and based on the joint loss function, performing iterative updates on the curve parameters in the adjustment parameter curve to obtain the adjustment parameter curve after parameter update; the using the adjustment parameter curve to perform dynamic range adjustment on the low-illumination image to be enhanced to obtain an enhanced low-illumination image includes: using the adjustment parameter curve after parameter update to perform dynamic range adjustment on the low-illumination image to be enhanced to obtain an enhanced low-illumination image.

[0060] Among them, the verification image set contains multiple verification low-illumination images in different forms and standard enhanced images corresponding to the verification low-illumination images. The standard enhanced image refers to an image whose quality such as image brightness, image color, and contrast meets the requirements.

[0061] Specifically, the dynamic range of the verified low-light image is adjusted using the adjustment parameter curve, and the adjusted image is used as the predicted enhanced image. Then, for the predicted enhanced image and the standard enhanced image corresponding to the same verified low-light image, the brightness difference, color difference, structural difference, etc. between the two images are determined. When determining the brightness difference, the brightness histograms of the two images can be determined respectively, and the difference between the two brightness histograms can be measured. When determining the color difference, the two images can be converted to the color space, and based on the color space, the color difference between the corresponding pixels in the two images can be determined. When determining the structural difference, it is necessary to determine the differences in edges, textures, geometric shapes, etc. between the two images. The core of the structural difference is to measure the differences in the spatial layout and local features of the two images. For example, the image features of the two images can be extracted respectively, and the difference between the two image features can be used as the structural difference. Then, the brightness loss function corresponding to the brightness difference, the color loss function corresponding to the color difference, and the structural loss function corresponding to the structural difference are determined, and the weight coefficients corresponding to the brightness loss function, the color loss function, and the structural loss function are determined respectively. Based on the weight coefficients, the brightness loss function, the color loss function, and the structural loss function are added together to obtain the combined loss function. The combined loss function is the key to optimizing the curve parameters in the adjustment parameter curve. It measures the difference between the image adjusted by the parameter curve and the standard image. By minimizing the combined loss function, the optimal curve parameters can be found for the adjustment parameter curve. Finally, the dynamic range of the low-light image to be enhanced is adjusted using the adjustment parameter curve with updated parameters, and the enhanced low-light image is obtained. Thus, the adjustment parameter curve is continuously iterated through the combined loss function, and a higher-order curve is achieved. The higher-order curve has a larger curvature, so it has a stronger dynamic range adjustment ability, thereby improving the enhancement effect of the adjustment parameter curve on the low-light image.

[0062] 103. Use the adjustment parameter curve to perform dynamic range adjustment on the low-light image to be enhanced, and obtain the enhanced low-light image.

[0063] For the embodiment of the present invention, after determining the adjustment parameter curve, it is necessary to use the adjustment parameter curve to perform dynamic range adjustment on the low-light image to be enhanced. Based on this, step 103 specifically includes: determining each pixel value in the low-light image to be enhanced ; respectively remap each pixel value using the adjustment parameter curve to obtain each remapped pixel value , where , is the curve parameter of the adjustment parameter curve, and the enhanced low-light image is composed of each remapped pixel value .

[0064] Specifically, each pixel value in the low-light image to be enhanced is respectively substituted into the above adjustment parameter curve to obtain the mapped pixel value corresponding to each pixel value. Finally, the enhanced low-light image is composed of each mapped pixel value. In the embodiment of the present invention, the curve parameters in the adjustment parameter curve can be controlled to change the dynamic range of the low-light image to be enhanced. It can either reduce the dynamic range of the low-light image to be enhanced or increase the dynamic range of the low-light image to be enhanced. By adjusting the dynamic range of the low-light image to be enhanced through the adjustment parameter curve, the embodiment of the present invention can achieve automatic adjustment of the low-light image, avoiding the problem of time-consuming and laborious manual adjustment, thereby improving the enhancement efficiency of the low-light image; the parameter curve can accurately adjust parameters such as the brightness and contrast of the image, avoiding errors or subjective deviations that may occur during manual adjustment, thereby improving the enhancement accuracy of the low-light image; the results of parameter curve adjustment have high consistency, and the same set of parameters can be applied to different images to ensure the consistency of the processing effect; the parameter curve can be saved and reused, facilitating subsequent processing of the same or similar images, while manual adjustment may vary due to different operators.

[0065] According to an enhancement method for low-light images provided by the present invention, compared with the current method of manually enhancing low-light images, by determining the adjustment parameter curve for adjusting the dynamic range of the low-light image to be enhanced and using the adjustment parameter curve to adjust the dynamic range of the low-light image to be enhanced, the present invention can achieve automatic adjustment of the low-light image, avoiding the problem of time-consuming and laborious manual adjustment, thereby improving the enhancement efficiency of the low-light image; the parameter curve can accurately adjust parameters such as the brightness and contrast of the image, avoiding errors or subjective deviations that may occur during manual adjustment, thereby improving the enhancement accuracy of the low-light image; the results of parameter curve adjustment have high consistency, and the same set of parameters can be applied to different images to ensure the consistency of the processing effect; the parameter curve can be saved and reused, facilitating subsequent processing of the same or similar images, while manual adjustment may vary due to different operators.

[0066] Further, in order to better illustrate the above process of enhancing low-light images, as a refinement and extension of the above embodiments, the embodiment of the present invention provides another enhancement method for low-light images, as Figure 2 shown, the method includes:

[0067] 201. Obtain the low-light image to be enhanced.

[0068] 202. Obtain a preset parameter curve prediction model, where the preset parameter curve prediction model includes an encoder for extracting image features and a decoder for predicting the parameter curve.

[0069] For the embodiments of the present invention, in order to improve the prediction accuracy of the preset parameter curve prediction model, it is first necessary to train and construct the preset parameter curve prediction model. Based on this, the method includes: constructing a preset initial parameter curve prediction model; obtaining a sample data set, where the sample data set includes multiple low-light sample images and the standard adjustment parameter curves corresponding to the low-light sample images; dividing the sample data set into a training set and a test set, using the training set to train the preset initial parameter curve prediction model, and using the test set to test the trained preset initial parameter curve prediction model, and taking the preset initial parameter curve prediction model that meets the test conditions as the preset parameter curve prediction model.

[0070] Among them, the standard adjustment parameter curve refers to a parameter curve whose image enhancement effect can meet the requirements.

[0071] Specifically, first construct a preset initial parameter curve prediction model, and the model structure can include an initial encoder and an initial decoder. Secondly, download the sample data set from the official website or the specified data source. Ensure that the sample data set contains all necessary files, including image files and the corresponding standard parameter adjustment curves. Finally, train and test the model. Specifically, the sample data set can be divided first: use a random or specific strategy (such as stratified sampling) to divide the sample data set into a training set and a test set. Then use the training set to train the model, monitor indicators such as the loss value during the training process to evaluate the model performance. Adjust the training parameters as needed, such as the learning rate, optimizer, regularization, etc., to optimize the training effect. Finally, test the model: use the test set to test the trained model and evaluate its performance on unseen images. Calculate and record indicators such as accuracy and recall rate on the test set. If the model performance does not meet the requirements, it can return to the training stage for more iterations or adjustments. In this way, a preset parameter curve prediction model with requirements such as prediction accuracy met is obtained.

[0072] 203. Input the low-light image to be enhanced into the preset parameter curve prediction model, extract image features from the low-light image to be enhanced through the encoder to obtain image features, and perform curve prediction on the image features through the decoder to obtain an adjustment parameter curve for dynamically adjusting the low-light image to be enhanced.

[0073] Specifically, as Figure 3 shown, input the low-light image to be enhanced into the preset parameter curve prediction model, extract the global and local features of the image through the encoder, then gradually restore the spatial information through the decoder and generate curve parameters, and finally obtain the adjustment parameter curve based on the curve parameters.

[0074] Further, in order to achieve refined enhancement of images, different adjustment parameter curves can be generated for images of different degradation types, and the degradation types of the images can be enhanced specifically through different adjustment parameter curves, so as to improve the enhancement effect of the images. Based on this, step 203 includes: determining reference images corresponding to different degradation types corresponding reference degraded image features , and determining the keys of the image features , values , and feature dimensions ; based on the reference degraded image features , the keys , the values , and the feature dimensions , respectively determine the matching weights between the low-light image to be enhanced and different degradation types , where , is the normalized exponential function; based on different matching weights and the values of the image features , determine the image features with the target degradation type as the degraded image features , where , is the total number of different degradation types; through the decoder, perform curve prediction on the degraded image features to obtain an adjustment parameter curve corresponding to the target degradation type.

[0075] Among them, different degradation types include low-light degradation type, backlight degradation type, local low-light degradation type, overexposure degradation type, blur degradation type, etc. Specifically, the reference images corresponding to different degradation types and the low-light image to be enhanced can be input into the encoder of the preset parameter curve prediction model at the same time. Through the encoder, extract the reference degraded image features corresponding to the reference images of different degradation types, determine the keys and values corresponding to the image features of the low-light image to be enhanced, and determine the length of the key vector as the feature dimension. The reference images corresponding to different degradation types Using the corresponding reference degraded image features as a query, matching the query with the keys and values in the image features, based on the matching results, generating the matching weights between the low-light image to be enhanced and different degradation types, and finally determining the target degradation type of the low-light image to be enhanced based on the matching weights. After that, determining the degradation features of the target degradation type information, and fusing the degradation features with the image features of the low-light image to be enhanced to obtain the degraded image features. Finally, inputting the degraded image features into the decoder of the preset parameter curve prediction model for curve prediction, and the decoder can output the adjustment parameter curve for adjusting the dynamic range of the low-light image to be enhanced. In the embodiment of the present invention, by adaptively matching the degradation type of the image and generating the image features with degradation type information, finally based on the image features, the adjustment parameter curve for a specific task (specific degradation type) can be generated, such as the dark light dynamic range adjustment curve, the backlight dynamic range adjustment curve, the local dark light dynamic range adjustment curve, the overexposure dynamic range adjustment curve, the blur dynamic range adjustment curve, etc. Through at least one targeted adjustment parameter curve, the targeted enhancement of the image quality can be achieved, thereby further improving the enhancement effect of the image quality.

[0076] 204. Using the adjustment parameter curve to adjust the dynamic range of the low-light image to be enhanced to obtain the enhanced low-light image.

[0077] Specifically, by remapping each pixel in the low-light image to be enhanced through the adjustment parameter curve, the dynamic range adjustment of the low-light image to be enhanced is realized, that is, the image quality enhancement of the low-light image to be enhanced is realized.

[0078] Furthermore, in another embodiment of the present invention, if a more refined dynamic range adjustment is to be performed on the low-light image to be enhanced, the degradation type requirement information and the low-light image to be enhanced can be input into the preset parameter curve prediction model at the same time. The preset parameter curve prediction model can output a parameter adjustment curve corresponding to the degradation type of the low-light image to be enhanced. For example, if the degradation type of the low-light image to be enhanced is the dark light degradation type, the dark light dynamic range adjustment curve is output. Finally, the dark light dynamic range adjustment curve is used to perform enhancement processing on the low-light image to be enhanced, so that the brightness distribution of the processed image is uniform. Therefore, in the embodiment of the present invention, independent parameter curves are designed for each degradation type, which can more accurately adjust the dynamic range of the image, avoid the problem that the unified parameter curve is difficult to adapt to the complex distribution of different degradation types, and may lead to poor enhancement effects in some scenarios.

[0079] Another method for enhancing low-light images provided by the present invention can, compared with the current manual method for enhancing low-light images, achieve automatic adjustment of low-light images by determining an adjustment parameter curve for dynamic range adjustment of the low-light image to be enhanced and using the adjustment parameter curve to perform dynamic range adjustment on the low-light image to be enhanced. This can avoid the time-consuming and laborious problem of manual adjustment, thereby improving the enhancement efficiency of low-light images. The parameter curve can precisely adjust parameters such as the brightness and contrast of the image, avoiding errors or subjective biases that may occur during manual adjustment, thereby improving the enhancement accuracy of low-light images. The results of parameter curve adjustment are highly consistent, and the same set of parameters can be applied to different images to ensure the consistency of processing effects. The parameter curve can be saved and reused, facilitating subsequent processing of the same or similar images, while manual adjustment may vary due to different operators.

[0080] Further, as a Figure 1 specific implementation, an embodiment of the present invention provides an apparatus for enhancing low-light images, as Figure 4 shown. The apparatus includes: an acquisition unit 31, a determination unit 32, and an adjustment unit 33.

[0081] The acquisition unit 31 can be used to acquire a low-light image to be enhanced.

[0082] The determination unit 32 can be used to determine an adjustment parameter curve for dynamic range adjustment of the low-light image to be enhanced.

[0083] The adjustment unit 33 can be used to perform dynamic range adjustment on the low-light image to be enhanced by using the adjustment parameter curve to obtain an enhanced low-light image.

[0084] In a specific application scenario, in order to determine the adjustment parameter curve, as Figure 5 shown, the determination unit 32 includes an acquisition module 321 and a prediction module 322.

[0085] The acquisition module 321 can be used to acquire a preset parameter curve prediction model, where the preset parameter curve prediction model includes an encoder for extracting image features and a decoder for predicting the parameter curve.

[0086] The prediction module 322 can be used to input the low-light image to be enhanced into the preset parameter curve prediction model, extract image features from the low-light image to be enhanced through the encoder to obtain image features, and predict a curve for the image features through the decoder to obtain an adjustment parameter curve for dynamic range adjustment of the low-light image to be enhanced.

[0087] In a specific application scenario, to determine the adjustment parameter curve, the prediction module 322 may specifically be used to determine reference images corresponding to different degradation types corresponding reference degraded image features , and determine the keys values , and feature dimensions ; based on the reference degraded image features , the keys , the values , and the feature dimensions , respectively determine the matching weights between the low - illumination image to be enhanced and different degradation types , where , is the normalization exponential function; based on different matching weights and the values of the image features , determine the image features with the target degradation type as the degraded image features , where , is the total number of different degradation types; through the decoder, perform curve prediction on the degraded image features to obtain the adjustment parameter curve corresponding to the target degradation type.

[0088] In a specific application scenario, to train and construct a preset parameter curve prediction model, the determination unit 32 further includes a construction module 323.

[0089] The construction module 323 may be used to construct a preset initial parameter curve prediction model; obtain a sample data set, where the sample data set includes multiple low - illumination sample images and the standard adjustment parameter curves corresponding to the low - illumination sample images; divide the sample data set into a training set and a test set, use the training set to train the preset initial parameter curve prediction model, and use the test set to test the trained preset initial parameter curve prediction model, and use the preset initial parameter curve prediction model that meets the test conditions as the preset parameter curve prediction model.

[0090] In a specific application scenario, to perform dynamic range adjustment on the low - illumination image to be enhanced, the adjustment unit 33 includes a first determination module 331 and a mapping module 332.

[0091] The first determination module 331 may be used to determine each pixel value in the low - illumination image to be enhanced.

[0092] The mapping module 332 can be used to determine each pixel value in the low-light image to be enhanced. ; respectively remap each pixel value using the adjustment parameter curve to obtain each remapped pixel value , where , is the curve parameter of the adjustment parameter curve, and each remapped pixel value constitutes the enhanced low-light image.

[0093] In a specific application scenario, in order to iteratively update the adjustment parameter curve, the device further includes an iterative update unit 34.

[0094] The iterative update unit 34 can be used to obtain a set of verification images for verifying the image adjustment ability of the adjustment parameter curve, where the set of verification images includes verification low-light images and the standard enhanced images corresponding to the verification low-light images; perform dynamic range adjustment on the verification low-light images using the adjustment parameter curve to obtain predicted enhanced images; respectively determine the brightness difference, color difference, and structural difference between the standard enhanced images and the predicted enhanced images, and determine the brightness loss function corresponding to the brightness difference, the color loss function corresponding to the color difference, and the structural loss function corresponding to the structural difference; based on the brightness loss function, the color loss function, and the structural loss function, determine a combined loss function, and based on the combined loss function, iteratively update the curve parameters in the adjustment parameter curve to obtain the adjustment parameter curve with updated parameters.

[0095] The adjustment unit 33 can also be used to perform dynamic range adjustment on the low-light image to be enhanced using the adjustment parameter curve with updated parameters to obtain the enhanced low-light image.

[0096] In a specific application scenario, in order to determine the adjustment parameter curve for performing dynamic range adjustment on the low-light image to be enhanced, the determination unit 32 further includes a second determination module 324.

[0097] The second determination module 324 can be used to determine the illumination feature information of the low-light image to be enhanced, where the illumination feature information includes the underexposure degree and illumination uniformity of the low-light image to be enhanced.

[0098] Specifically, the second determination module 324 can be used to determine the curve constraint conditions of the adjustment parameter curve to be generated based on the illumination feature information, where the curve constraint conditions include the pixel value range of the image adjusted by the adjustment parameter curve to be generated, the monotonicity and reversibility of the adjustment parameter curve to be generated.

[0099] The building block 323 can also be used to construct an adjustment parameter curve for dynamically adjusting the dynamic range of the low-light image to be enhanced based on the curve constraint condition.

[0100] It should be noted that for other corresponding descriptions of each functional module involved in the low-light image enhancement device provided in the embodiments of the present invention, reference can be made to Figure 1 the corresponding description of the method shown, which will not be elaborated here.

[0101] Based on the above as Figure 1 shown in the method, correspondingly, the embodiments of the present invention also provide a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the following steps are implemented: obtaining a low-light image to be enhanced; determining an adjustment parameter curve for dynamically adjusting the dynamic range of the low-light image to be enhanced; using the adjustment parameter curve to dynamically adjust the dynamic range of the low-light image to be enhanced to obtain an enhanced low-light image.

[0102] Based on the above as Figure 1 shown in the method and the embodiments of the device as Figure 4 shown, the embodiments of the present invention also provide a physical structure diagram of a computer device, as Figure 6 shown. The computer device includes: a processor 41, a memory 42, and a computer program stored on the memory 42 and executable on the processor. The memory 42 and the processor 41 are both arranged on a bus 43. When the processor 41 executes the program, the following steps are implemented: obtaining a low-light image to be enhanced; determining an adjustment parameter curve for dynamically adjusting the dynamic range of the low-light image to be enhanced; using the adjustment parameter curve to dynamically adjust the dynamic range of the low-light image to be enhanced to obtain an enhanced low-light image.

[0103] Through the technical solution of the present invention, by determining an adjustment parameter curve for dynamically adjusting the dynamic range of the low-light image to be enhanced and using the adjustment parameter curve to dynamically adjust the dynamic range of the low-light image to be enhanced, the automatic adjustment of the low-light image can be realized, avoiding the problem of time-consuming and laborious manual adjustment, thereby improving the enhancement efficiency of the low-light image; the parameter curve can accurately adjust parameters such as the brightness and contrast of the image, avoiding errors or subjective deviations that may occur during manual adjustment, thereby improving the enhancement accuracy of the low-light image; the results of the parameter curve adjustment have high consistency, and the same set of parameters can be applied to different images to ensure the consistency of the processing effect; the parameter curve can be saved and reused, which is convenient for subsequent processing of the same or similar images, while manual adjustment may vary due to different operators.

[0104] Obviously, those skilled in the art should understand that the above-mentioned modules or steps of the present invention can be implemented by a general-purpose computing device. They can be concentrated on a single computing device or distributed over a network composed of multiple computing devices. Optionally, they can be implemented by program codes executable by the computing device. Thus, they can be stored in a storage device and executed by the computing device. And in some cases, the steps shown or described can be executed in a sequence different from that here, or they can be separately fabricated into individual integrated circuit modules, or multiple modules or steps among them can be fabricated into a single integrated circuit module for implementation. In this way, the present invention is not limited to any specific combination of hardware and software.

[0105] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. An enhancement method for low-light images, characterized in that, Including: Obtain a low - illumination image to be enhanced; Determine an adjustment parameter curve for dynamically adjusting the dynamic range of the low - illumination image to be enhanced; Determine each pixel value in the low - illumination image to be enhanced , and respectively remap each pixel value by using the adjustment parameter curve to obtain each remapped pixel value , where , is the curve parameter of the adjustment parameter curve, and the enhanced low - illumination image is composed of each remapped pixel value ; Among them, the method for determining the adjustment parameter curve for dynamically adjusting the dynamic range of the low - illumination image to be enhanced includes: Obtain a preset parameter curve prediction model, where the preset parameter curve prediction model includes an encoder for extracting image features and a decoder for predicting the parameter curve; Input the low - illumination image to be enhanced into the preset parameter curve prediction model. Extract image features from the low - illumination image to be enhanced through the encoder to obtain image features, and determine the reference images corresponding to different degradation types The corresponding reference degraded image features , and determine the key value and feature dimension ; Based on the reference degraded image features and the key and the value and the feature dimension , respectively determine the matching weights between the low-light image to be enhanced and different degradation types , where , is the normalized exponential function; Based on different matching weights and the values of the image features , determine the image features with the target degradation type as the degraded image features , where , is the total number of different degradation types; perform curve prediction on the degraded image features through the decoder to obtain an adjustment parameter curve corresponding to the target degradation type.

2. The method for enhancing a low-light image according to claim 1, wherein Before obtaining the preset parameter curve prediction model, the method further includes: Construct a preset initial parameter curve prediction model; Obtain a sample data set, where the sample data set includes multiple low - illumination sample images and the standard adjustment parameter curves corresponding to the low - illumination sample images; Divide the sample data set into a training set and a test set, use the training set to train the preset initial parameter curve prediction model, and use the test set to test the trained preset initial parameter curve prediction model, and use the preset initial parameter curve prediction model that meets the test conditions as the preset parameter curve prediction model.

3. The enhancement method of the low-illumination image according to claim 1, characterized in that, Before remapping each pixel value respectively using the adjustment parameter curve to obtain each remapped pixel value the method further includes: Obtain a verification image set for verifying the image adjustment ability of the adjustment parameter curve, where the verification image set includes verification low - illumination images and the standard enhanced images corresponding to the verification low - illumination images; Use the adjustment parameter curve to perform dynamic range adjustment on the verification low - illumination images to obtain predicted enhanced images; Respectively determine the brightness difference, color difference, and structure difference between the standard enhanced image and the predicted enhanced image, and determine the brightness loss function corresponding to the brightness difference, the color loss function corresponding to the color difference, and the structure loss function corresponding to the structure difference; Based on the brightness loss function, the color loss function, and the structure loss function, determine a combined loss function, and based on the combined loss function, iteratively update the curve parameters in the adjustment parameter curve to obtain the adjustment parameter curve with updated parameters; Remap each pixel value using the described adjustment parameter curve to obtain each remapped pixel value , including: Remap each pixel value respectively using the adjusted parameter curve after parameter update to obtain each remapped pixel value .

4. The enhancement method of the low-illumination image according to claim 1, wherein Determining the adjustment parameter curve for dynamically adjusting the dynamic range of the low - illumination image to be enhanced includes: Determine the illumination feature information of the low - illumination image to be enhanced, where the illumination feature information includes the underexposure degree and illumination uniformity of the low - illumination image to be enhanced; Based on the illumination feature information, determine the curve constraint conditions for the adjustment parameter curve to be generated, where the curve constraint conditions include the pixel value range of the image adjusted based on the adjustment parameter curve to be generated, the monotonicity and reversibility of the adjustment parameter curve to be generated; Based on the curve constraint conditions, construct an adjustment parameter curve for dynamically adjusting the dynamic range of the low - illumination image to be enhanced.

5. An enhancement device for low-illumination images, characterized in that, Including: An acquisition unit for obtaining a low - illumination image to be enhanced; A determination unit for determining an adjustment parameter curve for dynamically adjusting the dynamic range of the low - illumination image; An adjustment unit for determining each pixel value in the low - illumination image to be enhanced , and respectively remapping each pixel value by using the adjustment parameter curve to obtain each remapped pixel value , where , is the curve parameter of the adjustment parameter curve, and each remapped pixel value constitutes the enhanced low - illumination image; Determination unit, configured to obtain a preset parameter curve prediction model, where the preset parameter curve prediction model includes an encoder for extracting image features and a decoder for predicting a parameter curve; input the low-light image to be enhanced into the preset parameter curve prediction model, extract image features from the low-light image to be enhanced through the encoder to obtain image features, and determine reference images corresponding to different degradation types Corresponding reference degraded image features , and determine the key of the image features , value , and feature dimension ; based on the reference degraded image features , the key , the value , and the feature dimension , respectively determine the matching weights between the low-light image to be enhanced and different degradation types , where , is a normalization exponential function; based on different matching weights and the value of the image features , determine the image features with the target degradation type as the degraded image features , where , is the total number of different degradation types; perform curve prediction on the degraded image features through the decoder to obtain an adjustment parameter curve corresponding to the target degradation type.​ 6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 4.

7. A computer device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 4.

Images