Backlight low-light image enhancement method based on histogram peak analysis

Abstract

The invention relates to a backlight low-light image enhancement method based on histogram peak analysis, belongs to the technical field of image enhancement, solves the problems of single algorithm function, high complexity and long reasoning time in the prior art, and comprises the following steps: acquiring a peak point of a V-channel histogram of an image to obtain the category of the image; if the category is a low-light image, performing low-light image enhancement, including performing adaptive histogram equalization, gamma correction and morphological top-hat correction on V-channel data of the image; fusing the data after histogram equalization and the data after morphological top cap correction; if the category is a backlight image, carrying out backlight image enhancement, which comprises the following steps of: respectively carrying out gamma correction and histogram equalization and fusion on V-channel data of the image; segmenting V-channel data of the image, and generating a weight map; obtaining enhanced V channel data based on the weight map and the fused data; and an enhanced RGB image is obtained. The image enhancement method is multifunctional, low in complexity and fast.

Description

Technical Field

[0001] The present invention relates to the technical field of image enhancement, and in particular, to a backlight low-light image enhancement method based on histogram peak analysis. Background Art

[0002] With the rapid development of various image sensors and detector systems, the requirements for the clarity and signal-to-noise ratio of the acquired images are constantly increasing. However, in poor lighting environments such as low light and backlight, as well as in the case of weak signals and high background noise in positron emission tomography (PET) detection, problems such as low contrast and high noise often occur in the image information, seriously affecting the imaging quality and diagnostic accuracy.

[0003] Most of the existing image enhancement algorithms are for single scenarios. For example, the low-light enhancement algorithm only processes low-light images and cannot effectively solve the backlight problem with a large signal dynamic range; conversely, the backlight enhancement algorithm cannot enhance low-light images or images with weak global signals. In addition, although the enhancement algorithm based on deep learning can produce good results, the algorithm has high complexity and long inference time, and it is difficult to achieve real-time deployment on devices with limited hardware resources. Summary of the Invention

[0004] In view of the above analysis, the embodiments of the present invention aim to provide a backlight low-light image enhancement method based on histogram peak analysis to solve the problems of single function, high complexity, and long inference time of the existing image enhancement algorithms.

[0005] The embodiments of the present invention provide a backlight low-light image enhancement method based on histogram peak analysis, including:

[0006] Obtain the peak points of the V-channel histogram of the image to be enhanced, and obtain the category of the image based on the distribution of the peak points;

[0007] If the category of the image is a low-light image, perform low-light image enhancement on the V-channel data of the image; wherein, the low-light image enhancement includes: sequentially performing adaptive histogram equalization, gamma correction, and morphological top-hat correction on the V-channel data of the image; fusing the data after adaptive histogram equalization and the data after morphological top-hat correction to obtain an enhanced V-channel image;

[0008] If the category of the image is a backlight image, perform backlight image enhancement on the V-channel data of the image; wherein, the backlight image enhancement includes: respectively performing gamma correction and histogram equalization on the V-channel data of the image and fusing them to obtain fused data; segmenting the V-channel data of the image, and generating a weight map based on the segmented data; performing secondary fusion on the V-channel data of the image and the fused data based on the weight map to obtain an enhanced V-channel image;

[0009] Based on the enhanced V-channel image, an enhanced RGB image is obtained.

[0010] Based on a further improvement of the above method, obtaining the peak points of the V-channel histogram of the image to be enhanced includes:

[0011] Obtain the RGB image of the image to be enhanced;

[0012] Convert the RGB image of the image to be enhanced into an HSV image, and extract the V-channel histogram of the HSV image;

[0013] Traverse each point in the V-channel histogram to obtain all potential peak points; wherein, if the value of a point is greater than the value of its adjacent point, then this point is regarded as a potential peak point;

[0014] Based on the potential peak points, using a height threshold and a minimum distance threshold, obtain the peak points of the V-channel histogram.

[0015] Based on a further improvement of the above method, before extracting the V-channel histogram in the HSV image, it also includes downsampling the V-channel feature map in the HSV image, and obtaining the V-channel histogram based on the downsampled V-channel feature map.

[0016] Based on a further improvement of the above method, using the height threshold and the minimum distance threshold to obtain the peak points of the V-channel histogram includes: regarding the potential peak points with a height greater than the height threshold and an interval greater than the minimum distance threshold as the peak points of the V-channel histogram;

[0017] Among them, the height threshold is obtained by the following formula:

[0018] F1 = rows * cols / 250

[0019] Among them, F1 represents the height threshold, rows represents the height of the V-channel feature map, and cols represents the width of the V-channel feature map.

[0020] Based on a further improvement of the above method, obtaining the category of the image based on the distribution of the peak points includes:

[0021] If the peak points are only distributed in the low-light interval, then the category of the image is a low-light image;

[0022] If the peak points are distributed in both the low-light interval and the high-light interval at the same time, then the category of the image is a backlight image; wherein,

[0023] The brightness value of the low-light interval is between 0 - 100, and the brightness value of the high-light interval is between 155 - 255.

[0024] Based on the further improvement of the above method, the data after adaptive histogram equalization and the data after morphological top-hat correction are fused using the following formula:

[0025]

[0026] where, represents the fused data, represents the data after adaptive histogram equalization, represents the data after morphological top-hat correction; W 1 and W 2 represent weight values.

[0027] Based on the further improvement of the above method, the calculation steps of W 1 and W 2 include:

[0028] Concatenate the data after adaptive histogram equalization and the data after morphological top-hat correction to obtain new data;

[0029] Calculate the covariance matrix of the obtained new data;

[0030] Traverse the eigenvalues and eigenvectors corresponding to each eigenvalue of the covariance matrix to obtain the eigenvector corresponding to the largest eigenvalue;

[0031] Normalize the eigenvector corresponding to the largest eigenvalue to obtain the normalized eigenvector; W 1 and W 2 are respectively the first element value and the second element value of the normalized eigenvector.

[0032] Based on the further improvement of the above method, the segmentation of the V-channel data of the image and the generation of a weight map based on the segmented data include:

[0033] Use the maximum inter-class variance algorithm to obtain the segmentation threshold, and segment the V-channel data of the image into a binary image based on the segmentation threshold;

[0034] Based on the binary image, use guided filtering to generate a weight map.

[0035] Based on the further improvement of the above method, after obtaining the category of the image, before low-light image enhancement or backlight image enhancement, it also includes preprocessing the V-channel data of the image; where,

[0036] The preprocessing includes: using the method of cross-segmentation to segment the V-channel data of the image to obtain multiple V-channel sub-images;

[0037] Perform low-light image enhancement or backlight image enhancement on the multiple V-channel sub-images in parallel; splice the enhanced multiple V-channel sub-images along the original path to obtain the enhanced V-channel image.

[0038] Based on a further improvement of the above method, the number of V-channel sub-images is four; among them,

[0039] The first V-channel sub-image is spliced by the pixel points of odd rows and odd columns in the V-channel data of the image;

[0040] The second V-channel sub-image is spliced by the pixel points of odd rows and even columns in the V-channel data of the image;

[0041] The third V-channel sub-image is spliced by the pixel points of even rows and odd columns in the V-channel data of the image;

[0042] The fourth V-channel sub-image is spliced by the pixel points of even rows and even columns in the V-channel data of the image.

[0043] Compared with the prior art, the present invention can at least achieve one of the following beneficial effects:

[0044] 1. By analyzing the peak points of the histogram, this method can effectively distinguish low-light images and backlight images, and perform image enhancement respectively, so as to achieve adaptive enhancement;

[0045] 2. After the backlight image is enhanced by the backlight image enhancement, the bright part and the dark part in the backlight image are effectively balanced, the details of the foreground dark part area are restored, and the image is clearer and more natural; after the low-light image is enhanced by the low-light image enhancement, the overall brightness is significantly improved, the details are retained at the same time, the color is more saturated, and the visual effect is greatly improved;

[0046] 3. Existing enhancement algorithms based on deep learning often have problems such as high complexity and long inference time, and are not suitable for real-time deployment of devices with limited hardware resources. However, this method adopts a traditional method process, which not only significantly improves the visual effect, but also reduces the computational complexity, achieving a faster processing speed and lower resource consumption.

[0047] In the present invention, the above technical solutions can also be combined with each other to achieve more preferred combination schemes. Other features and advantages of the present invention will be described in the subsequent specification, and some advantages can be made obvious from the specification, or understood by implementing the present invention. The objectives and other advantages of the present invention can be realized and obtained from the content specifically pointed out in the specification and the drawings. Description of the Drawings

[0048] The drawings are only for the purpose of showing specific embodiments, and are not considered as a limitation of the present invention. Throughout the drawings, the same reference signs represent the same components;

[0049] Figure 1 It is a flowchart of a backlight low-light image enhancement method based on histogram peak analysis shown in the embodiments of the present invention;

[0050] Figure 2 It is the distribution of peak points of low-light images shown in the embodiments of the present invention;

[0051] Figure 3 It is the distribution of peak points of backlight images shown in the embodiments of the present invention;

[0052] Figure 4 It is a schematic diagram of cross-segmentation shown in the embodiments of the present invention;

[0053] Figure 5 It is a schematic diagram of parallel image enhancement for multiple V-channel sub-images shown in the embodiments of the present invention. Detailed implementation manners

[0054] Next, the preferred embodiments of the present invention will be specifically described with reference to the accompanying drawings. The accompanying drawings form a part of this application and are used together with the embodiments of the present invention to explain the principles of the present invention, rather than to limit the scope of the present invention.

[0055] A specific embodiment of the present invention discloses a backlight low-light image enhancement method based on histogram peak analysis, as Figure 1 shown.

[0056] During implementation, it includes:

[0057] S10: Obtain the peak points of the V-channel histogram of the image to be enhanced, and obtain the category of the image based on the distribution of the peak points; the image to be enhanced refers to the input original image; the categories of the image include low-light images, backlight images, and other images;

[0058] S20: If the category of the image is a low-light image, perform low-light image enhancement on the V-channel data of the image; wherein, the low-light image enhancement includes: sequentially performing adaptive histogram equalization, gamma correction, and morphological top-hat correction on the V-channel data of the image; fusing the data after adaptive histogram equalization and the data after morphological top-hat correction to obtain the enhanced V-channel image;

[0059] S21: If the category of the image is a backlight image, perform backlight image enhancement on the V-channel data of the image; wherein, the backlight image enhancement includes: respectively performing gamma correction and histogram equalization on the V-channel data of the image and fusing them to obtain the fused data; segmenting the V-channel data of the image, generating a weight map based on the segmented data; performing secondary fusion on the V-channel data of the image and the fused data based on the weight map to obtain the enhanced V-channel image;

[0060] S30: Obtain the enhanced RGB image based on the enhanced V-channel image.

[0061] The method provided in this embodiment analyzes the peak points of the histogram, effectively discriminates low-light images and backlight images, and performs image enhancement separately, so as to achieve adaptive enhancement; after the backlight image is enhanced by the backlight image enhancement, the bright and dark parts in the backlight image are effectively balanced, the details of the foreground dark part area are restored, and the image is clearer and more natural; after the low-light image is enhanced by the low-light image enhancement, the overall brightness is significantly improved, the details are retained at the same time, the color is more saturated, and the visual effect is greatly improved.

[0062] Specifically, in step S10, obtaining the peak points of the V-channel histogram of the image to be enhanced includes steps S101 - S104:

[0063] S101: Obtain the RGB image of the image to be enhanced; this step clarifies the input form of the image to be enhanced, ensures the generality of the method, and provides a standardized image data source for subsequent processing.

[0064] S102: Convert the RGB image of the image to be enhanced into an HSV image, and extract the V-channel histogram of the HSV image, that is, the brightness channel histogram.

[0065] It should be noted that in another implementable manner, before extracting the V-channel histogram in the HSV image, the V-channel feature map in the HSV image can also be downsampled first, and the normalized V-channel histogram is obtained based on the downsampled V-channel feature map; exemplarily, the height and width of the downsampled V-channel feature map are both one-fourth of the original. The downsampling operation can significantly reduce the computational complexity and improve the efficiency of image processing; normalization is used to improve the stability, robustness and efficiency of the calculation.

[0066] The V-channel histogram plots the number of pixels corresponding to each brightness value in the image, so the exposure level of the image and whether there are overexposed or underexposed areas in the image can be judged; among them, the range of the brightness value is 0 - 255.

[0067] S103: Traverse each point in the V-channel histogram to obtain all potential peak points; among them, if the value of a point is greater than the value of its adjacent point, then this point is regarded as a potential peak point.

[0068] The value refers to the brightness value corresponding to this point.

[0069] S104: Based on the potential peak points, use the height threshold and the minimum distance threshold to obtain the peak points of the V-channel histogram.

[0070] The ordinate of the V-channel histogram represents the total number of occurrences of each pixel brightness value in the image, the height threshold represents the ordinate value of the potential peak point; the minimum distance threshold represents the horizontal coordinate distance between a potential peak point and its adjacent potential peak point in the V-channel histogram.

[0071] Specifically, it includes: regarding potential peak points with a height greater than the height threshold and an interval greater than the minimum distance threshold as the peak points of the V-channel histogram; filtering out smaller and closer peak points through the height threshold and the minimum distance threshold; among them, the height threshold is obtained by the following formula:

[0072] F1 = rows * cols / 250

[0073] Wherein, F1 represents the height threshold, rows represents the height of the V-channel feature map, and cols represents the width of the V-channel feature map. Exemplarily, through experimental summary, the minimum distance threshold is taken as 30.

[0074] Furthermore, in step S10, the categories of the image obtained based on the distribution of the peak points include:

[0075] As Figure 2 shown, if the peak points are only distributed in the low-light interval, the category of the image is a low-light image; the characteristic of a low-light image is the low-light distribution in the entire image range, so the peak points are concentrated in the low-light interval;

[0076] As Figure 3 shown, if the peak points are distributed in both the low-light interval and the high-light interval, the category of the image is a backlight image; the characteristic of a backlight image is a brighter background area and a darker foreground area, so the brightness of a backlight image is concentrated in the low-light interval and the high-light interval, and thus the peak points appear concentrated in the low-light interval and the high-light interval;

[0077] Otherwise, it is considered that the category of the image belongs to other images and no enhancement processing is performed; among them,

[0078] In the brightness value range of 0 - 255, the brightness value in the low-light interval is 0 - 100, and the brightness value in the high-light interval is 155 - 255.

[0079] This step can extract the significant features of the V-channel histogram through the distribution of the peak points, effectively distinguish between low-light and backlight scenes, and enhance the adaptive ability of the method; at the same time, this step uses the prior knowledge of the brightness interval to quickly classify the image category, realizes the distinction between low-light images and backlight images, provides clear input conditions for the subsequent targeted enhancement algorithm, and further improves the processing efficiency and accuracy.

[0080] In an alternative embodiment, after obtaining the category of the image and before low-light image enhancement or backlight image enhancement, it may further include preprocessing the V-channel data of the image;

[0081] Among them, as Figure 4 shown, the preprocessing includes: using the method of cross-segmentation to segment the V-channel data of the image to obtain multiple V-channel sub-images;

[0082] As Figure 5 shown, perform low-light image enhancement or backlight image enhancement on the multiple V-channel sub-images in parallel; splice the enhanced multiple V-channel sub-images along the original path to obtain the enhanced V-channel image. Using the method of cross-segmentation can avoid the boundary effect in image enhancement, and at the same time speed up the image processing through multi-threaded processing, significantly improving the real-time performance of the algorithm and meeting the application requirements of devices with limited hardware resources;

[0083] Exemplarily, the number of the V-channel sub-images is four, and the length and width of each sub-image are both half of the original image; among them,

[0084] The first V-channel sub-image is spliced by the pixel points of odd rows and odd columns in the V-channel data of the image;

[0085] The second V-channel sub-image is spliced by the pixel points of odd rows and even columns in the V-channel data of the image;

[0086] The third V-channel sub-image is spliced by the pixel points of even rows and odd columns in the V-channel data of the image;

[0087] The fourth V-channel sub-image is spliced by the pixel points of even rows and even columns in the V-channel data of the image.

[0088] Furthermore, in step S20, before performing adaptive histogram equalization on the V-channel data of the image, it further includes performing brightness inversion on the V-channel data of the image using the following formula:

[0089]

[0090] Among them, V ij represents the V-channel data of the image, i represents the abscissa of the image in the two-dimensional space, j represents the ordinate of the image in the two-dimensional space, represents the V-channel data after brightness inversion. Brightness inversion can prevent the noise enhancement of dark images and improve the accuracy of image processing.

[0091] Further, perform adaptive histogram equalization, gamma correction, and morphological top-hat correction on the V-channel data with brightness inversion in sequence; exemplarily, the gamma coefficient is 5. Adaptive histogram equalization can adjust the contrast of local regions of the V-channel data with brightness inversion, while avoiding the over-enhancement problems that may be caused by traditional histogram equalization, such as noise amplification or detail loss; gamma correction changes the brightness curve distribution of the image after adaptive histogram equalization by adjusting the non-linear mapping relationship of pixel values. For low-light images, the dark pixel values can be exponentially increased to significantly improve their brightness; morphological top-hat correction can correct the uneven distribution of image brightness values after gamma correction to prepare for subsequent processing.

[0092] Further, use the following formula to fuse the data after adaptive histogram equalization and the data after morphological top-hat correction in step S20:

[0093]

[0094] where, represents the fused data, represents the data after adaptive histogram equalization, represents the data after morphological top-hat correction; W 1 and W 2 represent weight values; the calculation steps of W 1 and W 2 include:

[0095] Concatenate the data after adaptive histogram equalization and the data after morphological top-hat correction to obtain new data; among them, the data matrix after adaptive histogram equalization, the data after morphological top-hat correction, and the new data are all in matrix form;

[0096] Calculate the covariance matrix of the new data;

[0097] Traverse the eigenvalues of the covariance matrix and the eigenvectors corresponding to each eigenvalue to obtain the eigenvector corresponding to the largest eigenvalue;

[0098] Normalize the eigenvector corresponding to the largest eigenvalue to obtain the normalized eigenvector; W 1 and W 2They are respectively the first element value and the second element value of the normalized eigenvector. This method can obtain the main direction of data change, determine the contribution of the data after adaptive histogram equalization and the data after morphological top-hat correction in the main direction of data change, and can dynamically adjust the weight according to the characteristics of the input image, thereby avoiding the performance instability caused by the fixed weight set artificially. The weight value obtained by this method not only retains the contrast enhancement effect of adaptive histogram equalization, but also corrects the non-uniformity brought by gamma correction and top-hat correction, and optimizes the quality of the fusion result by extracting the main change direction, significantly improving the enhancement effect of the image.

[0099] The V-channel image after enhancing the low-light image is obtained by using the following formula:

[0100]

[0101] where represents the V-channel image after enhancing the low-light image.

[0102] Preferably, in step S21, the data after fusing gamma correction and histogram equalization is obtained by using the following formula:

[0103]

[0104] where O ij represents the data after fusing gamma correction and histogram equalization, represents the data of gamma correction, represents the data of histogram equalization, and α represents the weight; exemplarily, the gamma coefficient is 2 and α is 0.7.

[0105] Further, in step S21, segmenting the V-channel data of the image and generating a weight map based on the segmented data includes:

[0106] Obtaining a segmentation threshold by using the maximum inter-class variance algorithm, and segmenting the V-channel data of the image into a binary map W ij ;

[0107] Using the following formula, based on the binary map, generating a weight map by using guided filtering:

[0108]

[0109]

[0110] where represents the weight map, V ij represents the V-channel data of the image, represents the normalized V-channel data, W ij represents the binary map, aij and b ij represent the parameters of the guided filter, i represents the ordinate of the pixel point, j represents the abscissa of the pixel point, N represents the total number of pixel points in the local window, Ω represents the pixel point coordinates of the local window, ∈ represents the penalty term, which is used to prevent the denominator from being 0; exemplarily, ∈ is 0.005.

[0111] Further, based on the weight map, the V-channel data of the image and the fused data are fused again, and the enhanced V-channel image is obtained by the following formula:

[0112]

[0113] wherein, represents the V-channel image after the backlight image is enhanced, represents the image obtained by fusing the V-channel data of the image and the fused data again based on the weight map.

[0114] Further, in step S30, based on the enhanced V-channel image, obtaining the enhanced RGB image includes:

[0115] Replacing the V-channel image of the HSV image of the image to be enhanced with the enhanced V-channel image to obtain the replaced HSV image;

[0116] Restoring the replaced HSV image to the RGB image to obtain the enhanced RGB image.

[0117] Compared with the prior art, the method provided in this embodiment can effectively distinguish low-light images and backlight images by analyzing the peak points of the histogram, and perform image enhancement respectively, so as to achieve adaptive enhancement; after the backlight image is enhanced by the backlight image enhancement, the bright part and the dark part in the backlight image are effectively balanced, the details of the foreground dark part area are restored, and the image is clearer and more natural; after the low-light image is enhanced by the low-light image enhancement, the overall brightness is significantly improved, the details are retained at the same time, the color is more saturated, and the visual effect is greatly improved; existing enhancement algorithms based on deep learning often have problems of high complexity and long inference time, and are not suitable for real-time deployment of devices with limited hardware resources. However, this method adopts the traditional method process, which not only significantly improves the visual effect, but also reduces the computational complexity, achieving faster processing speed and lower resource consumption.

[0118] Those skilled in the art can understand that all or part of the processes of implementing the method of the above embodiment can be completed by instructing relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium. Among them, the computer-readable storage medium is a disk, an optical disc, a read-only memory or a random access memory, etc.

[0119] As described above, it is only the preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A backlit low-light image enhancement method based on histogram peak analysis, characterized in that: include: Obtaining the peak point of the V channel histogram of the image to be enhanced, and obtaining the category of the image based on the distribution of the peak point; If the category of the image is a low-light image, the V channel data of the image is enhanced in the low-light image; wherein the low-light image enhancement includes: sequentially performing adaptive histogram equalization, gamma correction and morphological top-hat correction on the V channel data of the image; fusing the data after adaptive histogram equalization and the data after morphological top-hat correction to obtain an enhanced V channel image; If the category of the image is a backlight image, the backlight image enhancement is performed on the V channel data of the image; wherein the backlight image enhancement includes: performing gamma correction and histogram equalization on the V channel data of the image respectively and fusing them to obtain fused data; segmenting the V channel data of the image, and generating a weight map based on the segmented data; performing secondary fusion on the V channel data of the image and the fused data based on the weight map to obtain an enhanced V channel image; Based on the enhanced V channel image, an enhanced RGB image is obtained.

2. The backlight and low-light image enhancement method based on histogram peak analysis according to claim 1, characterized in that: The step of obtaining the peak point of the V channel histogram of the image to be enhanced comprises: Obtain the RGB image of the image to be enhanced; Convert the RGB image of the image to be enhanced into an HSV image, and extract the V channel histogram of the HSV image; Traversing each point in the V channel histogram to obtain all potential peak points; wherein, if the value of a point is greater than the value of its adjacent point, the point is regarded as a potential peak point; Based on the potential peak point, the peak point of the V channel histogram is obtained by using a height threshold and a minimum distance threshold.

3. The backlight and low-light image enhancement method based on histogram peak analysis according to claim 2, characterized in that: Before extracting the V channel histogram in the HSV image, the method further includes downsampling the V channel feature map in the HSV image, and obtaining the V channel histogram based on the downsampled V channel feature map.

4. The backlight and low-light image enhancement method based on histogram peak analysis according to claim 3, characterized in that: The step of obtaining the peak point of the V channel histogram by using the height threshold and the minimum distance threshold comprises: considering potential peak points whose height is greater than the height threshold and whose interval is greater than the minimum distance threshold as the peak points of the V channel histogram; The height threshold is obtained using the following formula: F1 = rows*cold / 250 Among them, F1 represents the height threshold, rows represents the height of the V channel feature map, and cols represents the width of the V channel feature map.

5. The backlight and low-light image enhancement method based on histogram peak analysis according to claim 1, characterized in that: The category of the image obtained based on the distribution of the peak points includes: If the peak points are only distributed in the low-light range, the image is classified as a low-light image; If the peak points are distributed in both the low light interval and the high light interval, the image is classified as a backlit image; wherein, The brightness value of the low light range is 0-100, and the brightness value of the highlight range is 155-255.

6. The backlight and low-light image enhancement method based on histogram peak analysis according to claim 1, characterized in that: The following formula is used to fuse the data after adaptive histogram equalization and the data after morphological top hat correction: in, represents the fused data, Represents the data after adaptive histogram equalization, represents the data after morphological top-hat correction; W1 and W2 represent weight values.

7. The backlight and low-light image enhancement method based on histogram peak analysis according to claim 6, characterized in that: The calculation steps of W1 and W2 include: The data after adaptive histogram equalization and the data after morphological top-hat correction are spliced ​​to obtain new data; Calculate and obtain the covariance matrix of the new data; Traversing the eigenvalues ​​of the covariance matrix and the eigenvectors corresponding to the eigenvalues, and obtaining the eigenvector corresponding to the maximum eigenvalue; The eigenvector corresponding to the maximum eigenvalue is normalized to obtain a normalized eigenvector; W1 and W2 are respectively the first element value and the second element value of the normalized eigenvector.

8. The backlight and low-light image enhancement method based on histogram peak analysis according to claim 1, characterized in that: The step of segmenting the V channel data of the image and generating a weight map based on the segmented data includes: A segmentation threshold is obtained by using a maximum inter-class difference algorithm, and V channel data of the image is segmented into a binary image based on the segmentation threshold; Based on the binary image, a weight map is generated using guided filtering.

9. The backlight and low-light image enhancement method based on histogram peak analysis according to claim 1, characterized in that: After obtaining the category of the image, before low-light image enhancement or backlight image enhancement, the V channel data of the image is preprocessed; wherein, The preprocessing includes: using a cross-segmentation method to segment the V channel data of the image to obtain a plurality of V channel sub-images; Low-light image enhancement or backlight image enhancement is performed on the multiple V-channel sub-images in parallel; and the multiple enhanced V-channel sub-images are spliced ​​in the original way to obtain an enhanced V-channel image.

10. The backlight and low-light image enhancement method based on histogram peak analysis according to claim 9, characterized in that: The number of the V channel sub-images is four; wherein, The first V channel sub-image is formed by splicing the pixels of odd rows and odd columns in the V channel data of the image; The second V channel sub-image is formed by splicing the pixels in odd rows and even columns in the V channel data of the image; The third V channel sub-image is formed by splicing the pixels in the even rows and odd columns of the V channel data of the image; The fourth V channel sub-image is formed by splicing the pixel points in the even-numbered rows and even-numbered columns in the V channel data of the image.

Images