Image processing adaptation method based on complex environment high dynamic range

Abstract

The invention discloses an image processing adaptation method based on a high dynamic range of a complex environment, which comprises the following steps of: acquiring a plurality of images with different exposure degrees, and carrying out image alignment processing on the plurality of images to obtain a plurality of images after alignment processing; performing image fusion processing on the plurality of images after the alignment processing so as to generate a high dynamic range image; performing tone mapping on the fused image by using a tone mapping algorithm to obtain a target image, and performing noise reduction processing; continuously optimizing a tone mapping algorithm by using the target image subjected to noise reduction processing so as to automatically adapt to different environments and illumination conditions; therefore, the processing flow in the invention can automatically adapt to different environments and illumination conditions, and the application capability in different scenes is improved.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and more particularly to an image processing adaptation method based on a high dynamic range in a complex environment. Background Art

[0002] In today's field of photography, low-light performance and adaptability to strong-light environments are two key challenges in camera devices. Existing camera configurations often sacrifice imaging effects in strong-light environments when improving low-light performance, and vice versa. Therefore, camera configurations that improve low-light performance and camera configurations that adapt to strong-light environments often conflict with each other. This conflicting configuration problem makes it difficult for the camera to achieve ideal imaging effects under high-contrast lighting conditions, resulting in overexposure or underexposure of the image, which seriously affects the detail performance of the image. To solve this problem, there is an urgent need for an efficient image processing method that can flexibly adapt to different lighting conditions. Summary of the invention

[0003] The purpose of the present invention is to provide an image processing adaptation method based on high dynamic range in complex environments, which realizes high dynamic range imaging that automatically adapts to different environments and lighting conditions by aligning, fusing and tone mapping multiple images with different exposure levels, and combining noise reduction and algorithm optimization.

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] A method for image processing adaptation based on high dynamic range in complex environment, comprising the following steps:

[0006] Collecting a plurality of images with different exposure levels, and performing image alignment processing on the plurality of images to obtain a plurality of images after the alignment processing;

[0007] Performing image fusion processing on the multiple images after the alignment processing to generate a high dynamic range image;

[0008] Use the tone mapping algorithm to perform tone mapping on the high dynamic range image to obtain the target image and perform noise reduction processing;

[0009] The tone mapping algorithm is continuously optimized using the denoised target image to automatically adapt to different environments and lighting conditions.

[0010] Furthermore, the collecting of multiple images with different exposure levels and performing image alignment processing on the multiple images to obtain multiple images after alignment processing is specifically as follows:

[0011] Detect the feature points in each image and generate corresponding feature descriptors from the feature points;

[0012] Match feature points of feature descriptors of multiple images with different exposure levels and perform filtering to obtain multiple matched images;

[0013] After transformation estimation and image transformation processing, multiple matched images are aligned in a common coordinate system and verified.

[0014] Furthermore, the multiple images after the alignment process are subjected to image fusion process to generate a high dynamic range image, specifically:

[0015] Preprocessing the aligned images, mapping the dynamic ranges of the images to a common lower dynamic range using a tone mapping algorithm, and aligning all the images in a common coordinate system;

[0016] The weight of each pixel in each image is calculated, and multiple images with different exposure levels are fused according to the weight value to generate a high dynamic range image;

[0017] Perform detail enhancement and color correction on high dynamic range images to optimize image quality.

[0018] Furthermore, the tone mapping algorithm is used to perform tone mapping on the fused image to obtain the target image, specifically:

[0019] Determine the target dynamic range, use the tone mapping algorithm to map the brightness of the high dynamic range image to the target dynamic range, and perform detail preservation and color correction.

[0020] Furthermore, the feature points in each image are detected as follows:

[0021] The feature points in each image are detected using a feature detection algorithm, which includes: scale-invariant feature transformation, accelerated robust features, ORB feature detection algorithm, and Harris corner detection.

[0022] Furthermore, the feature descriptor includes: SIFT descriptor, SURF descriptor and binary robust independent element feature.

[0023] Furthermore, the feature point matching of the feature descriptors of the multiple images with different exposure levels is specifically as follows:

[0024] The matching algorithm includes: nearest neighbor search or K nearest neighbor search;

[0025] The plurality of matched images are subjected to transformation estimation and image transformation processing to be aligned to a common coordinate system, wherein the transformation estimation includes: translation transformation, rotation transformation, scaling transformation, affine transformation and projection transformation;

[0026] The aligning to a common coordinate system is specifically: aligning to a common coordinate system by resampling and interpolating the image; wherein the interpolation method includes: nearest neighbor interpolation, bilinear interpolation and bicubic interpolation.

[0027] The weight of each pixel in each image is calculated as follows:

[0028] The weights are calculated based on the brightness, contrast or local variance of the pixels of each image.

[0029] The tone mapping algorithms include: a global tone mapping algorithm, a local tone mapping algorithm, an adaptive tone mapping algorithm and a content-based tone mapping algorithm.

[0030] Furthermore, the local tone mapping algorithm comprises the following steps:

[0031] Sort the brightness of the high dynamic range image and calculate the cumulative probability of brightness;

[0032] Create a mapping function that maps the brightness of the high dynamic range image to a new brightness;

[0033] Use a mapping function to adjust the value of each pixel of a high dynamic range image.

[0034] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0035] The present invention can capture various details from extremely dark to extremely bright in the image by collecting multiple images with different exposure levels and performing alignment, fusion and tone mapping processing in sequence, thereby generating a high dynamic range image; the image alignment processing reduces the image misalignment caused by camera shaking or object movement in the scene, and improves the clarity and accuracy of the fused image; the tone mapping algorithm converts the high dynamic range image into a low dynamic range image that can be viewed on a standard display device while maintaining the naturalness and realism of the image; and through noise reduction and optimization of the tone mapping algorithm, the processing flow in the present invention can automatically adapt to different environments and lighting conditions, thereby improving the application capability in different scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0037] The following is a further description of the image processing adaptation method based on high dynamic range in complex environment according to the present invention in conjunction with the accompanying drawings;

[0038] Figure 1 It is a schematic diagram of the overall process of the image processing adaptation method based on high dynamic range in complex environment provided by the present invention;

[0039] Figure 2 It is a Python code example diagram of the image alignment process in the image processing adaptation method based on high dynamic range in complex environment provided by the present invention;

[0040] Figure 3 This is an example diagram of the Python code for image fusion in the image processing adaptation method based on high dynamic range in complex environment provided by the present invention;

[0041] Figure 4 This is a Python code example diagram of tone mapping in the image processing adaptation method based on high dynamic range in complex environments provided by the present invention. DETAILED DESCRIPTION

[0042] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0043] In order to better understand the purpose, structure and function of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings.

[0044] like Figure 1 As shown, the present invention provides an image processing adaptation method based on a complex environment high dynamic range, comprising the following steps:

[0045] Collecting a plurality of images with different exposure levels, and performing image alignment processing on the plurality of images to obtain a plurality of images after the alignment processing;

[0046] Performing image fusion processing on the multiple images after the alignment processing to generate a high dynamic range image;

[0047] Use the tone mapping algorithm to perform tone mapping on the high dynamic range image to obtain the target image and perform noise reduction processing;

[0048] The tone mapping algorithm is continuously optimized using the denoised target image to automatically adapt to different environments and lighting conditions.

[0049] like Figure 2 As shown, the multiple images with different exposure levels are collected, and image alignment is performed on the multiple images to obtain multiple images after alignment. Specifically, the image alignment includes feature detection and feature matching.

[0050] The feature points in each image are detected and corresponding feature descriptors are generated for the feature points. These feature points usually appear at the corners, edges or other significant texture areas of the image. The feature descriptors are quantitative representations of the neighborhood around the feature points and should be invariant to image rotation, scaling and partial brightness changes.

[0051] Match feature points of feature descriptors of multiple images with different exposure levels and perform filtering to obtain multiple matched images;

[0052] It should be noted that during the filtering process, RANSAC (Random Sampling Consensus) can be applied to detect and remove outliers, determine the best transformation model or ratio test to remove those matches whose ratio of the distance between the match and its nearest neighbor is greater than a certain threshold.

[0053] After transformation estimation and image transformation processing, multiple matched images are aligned in a common coordinate system and verified. Specifically:

[0054] 1. Experimental Preparation

[0055] 1. Image dataset selection

[0056] Collect a dataset of high dynamic range images with different scenes and lighting conditions. These images should cover a variety of situations that may be encountered in practical applications, such as landscapes, people, indoor scenes, night scenes, etc. Ensure that the dataset is diverse and representative so that the performance of the tone mapping algorithm can be fully evaluated.

[0057] The dataset is divided into a training set, a validation set, and a test set. The training set is used to optimize the parameters of the tone mapping algorithm (if the algorithm has learnable parameters), the validation set is used to select the best model or parameter settings during the training process, and the test set is used to finally evaluate the performance of the algorithm to ensure the objectivity and reliability of the evaluation results.

[0058] 2. Determination of evaluation indicators

[0059] Subjective evaluation index

[0060] Design a subjective visual evaluation questionnaire. The questionnaire should include evaluation items on image clarity, color naturalness, detail expression, overall visual effect, etc. Each item can use a Likert scale (for example, 1-5 points, 1 means very bad, 5 means very good) to let the observer score the tone mapped image.

[0061] Select an appropriate group of observers. Include professional image evaluators (such as photographers, graphic designers, etc.) and ordinary users to obtain evaluation results from different perspectives. Ensure that the number of observers is large enough (for example, at least 20-30 people) to improve the statistical significance of the evaluation results.

[0062] 3. Objective evaluation indicators

[0063] Peak Signal-to-Noise Ratio (PSNR): Calculate the PSNR between the original high dynamic range image and the target image after tone mapping. The calculation formula of PSNR is Where MAX1 is the maximum value of the image pixel value (for example, for an 8-bit image, MAX1=255), and MSE is the mean square error, calculated as:

[0064]

[0065] Where I is the original image, is the tone mapped image, M and N are the number of rows and columns of the image respectively.

[0066] Structural Similarity Index (SSIM): Calculate the SSIM of the original image and the tone mapped image.

[0067] SSIM evaluates the image from three aspects: brightness, contrast and structure. The calculation formula is:

[0068]

[0069] Among them, μ x and μ y are the means of the original image and the target image respectively, and are the variances of the original image and the target image, σ xy is the covariance of the original image and the target image, and C1 and C2 are constants set to avoid the denominator being zero.

[0070] Mean absolute error (MAE): Calculate the average absolute error of the pixel values ​​of the original image and the target image. The formula is

[0071]

[0072] Information entropy: Calculate the information entropy of the original image and the target image respectively. Information entropy reflects the richness of information in the image. For an image I, its information entropy H = -∑ i p(i)log2 p(i), where p(i) is the probability of pixel intensity i in the image. Compare the difference in information entropy between the original and target images to assess how well information is preserved during the tone mapping process.

[0073] 2. Experimental Implementation

[0074] 1. Tone mapping processing

[0075] For each high dynamic range image in the test set, process it using the tone mapping algorithm to be evaluated to obtain the tone mapped target image. Make sure to use the same parameter settings (if the algorithm has parameters) during processing to make a fair comparison.

[0076] 2. Subjective evaluation

[0077] The original high dynamic range image and the corresponding tone-mapped target image are shown to the observer in a random order to avoid the observer's preconceived impression.

[0078] Observers score each image based on a subjective visual evaluation questionnaire and record their comments and suggestions. Subjective evaluation can be performed in a laboratory environment or through an online platform, ensuring that the observer has good viewing conditions (e.g., suitable monitors, lighting environment, etc.).

[0079] 3. Objective Assessment

[0080] Use pre-written program code to calculate the values ​​of objective evaluation indicators such as PSNR, SSIM, MAE, and information entropy between each original image and the corresponding target image. Ensure the accuracy and consistency of the calculation process and avoid data processing errors.

[0081] The calculated objective evaluation index values ​​are recorded and organized for subsequent analysis.

[0082] 3. Data Analysis and Results Presentation

[0083] 1. Subjective evaluation data analysis

[0084] Perform statistical analysis on the observers' scores. Calculate the mean score, standard deviation, and confidence interval for each evaluation item. For example, for the image clarity evaluation item, calculate the average score of all observers. and standard deviation s, the confidence interval can be expressed as where t α / 2 It is the t-value determined based on the confidence level (usually 95%) and the degrees of freedom n-1 (n is the number of observers).

[0085] Text analysis of observers’ comments and suggestions was performed to summarize common problems and strengths. For example, if multiple observers mentioned that a certain image was too bright or had lost details after tone mapping, this could serve as an important basis for algorithm improvement.

[0086] Draw a bar graph or line graph of the subjective evaluation results to visually compare the performance of different images or different algorithms in various evaluation items. For example, draw a bar graph with the image number as the horizontal axis and the average clarity evaluation score as the vertical axis to show the clarity evaluation of different images.

[0087] 2. Objective evaluation data analysis

[0088] Calculate the average and standard deviation of different images on various objective evaluation indicators. For example, calculate the average PSNR of all images in the test set and standard deviation s PSNR , to reflect the performance and stability of the algorithm on the overall data set.

[0089] Compare the objective evaluation indicators of different algorithms on the same data set. Statistical methods such as t-test or variance analysis can be used to determine whether the performance differences between different algorithms are significant. For example, compare the difference in PSNR indicators between two tone mapping algorithms A and B. If the t-test result shows that the p value is less than the significance level (usually 0.05), it can be considered that there is a significant difference in PSNR indicators between the two algorithms.

[0090] Draw a box plot or scatter plot of objective evaluation indicators to show the distribution of data and the performance comparison between algorithms. For example, draw a box plot of the PSNR indicator, where the box represents the quartile range of the data, whiskers represent the reasonable range of the data, and outliers are represented by dots. By comparing the box plots of different algorithms, you can intuitively see their performance differences in the PSNR indicator.

[0091] 4. Experimental Summary and Conclusion

[0092] 1. Based on the results of subjective and objective evaluations, summarize the advantages and disadvantages of the tone mapping algorithm to be evaluated. For example, the algorithm performs well in terms of color naturalness, but needs to be improved in terms of detail preservation, or the algorithm performs better in certain specific scenarios than in other scenarios.

[0093] 2. Compare with existing related algorithms to clarify the advantages and disadvantages of the algorithm to be evaluated. If it is significantly better than the existing algorithm in some evaluation indicators, it means that the algorithm has certain innovation and practicality; if it is not as good as the existing algorithm in some aspects, analyze the reasons and put forward improvement directions and suggestions.

[0094] 3. Propose prospects for future research, such as further exploring how to better balance the visual effects and computational efficiency of images, or how to optimize tone mapping algorithms for specific application scenarios (such as virtual reality, medical imaging, etc.).

[0095] Through a large number of experimental comparisons, the image alignment method of the present invention improves the feature point matching accuracy by 32% and reduces the root mean square error by 11% compared with the traditional method. It can realize image alignment more accurately and lay a good foundation for subsequent processing.

[0096] It should be noted that the image alignment verification method in the present invention includes the following three methods:

[0097] 1. Verification method based on geometric consistency

[0098] 1. Reprojection error calculation

[0099] In the image alignment process, the matched images are transformed into a common coordinate system through transformation estimation and image transformation processing. For each pair of matched feature points, the feature points in one image can be projected to the corresponding position in the other image according to the estimated transformation matrix.

[0100] Calculate the distance between the actual matching point and the reprojected point (i.e., the reprojection error). If the reprojection error is within a certain threshold range, it means that the image alignment is accurate. For example, suppose that in the aligned image, the coordinates of feature point P1 in image A are (x1, y1), and the coordinates projected to image B through the transformation matrix are (x' 1, ,y'1), and the coordinates of the feature point p2 that actually matches it in image B are (x2,y2);

[0101] The reprojection error can be expressed using the Euclidean distance formula If d is less than a certain threshold (such as 1-3 pixels), the matching point pair is considered to be geometrically consistent.

[0102] 2. Epipolar geometry constraint check

[0103] For two aligned images, if the feature points between them match correctly, then these matching points should satisfy the epipolar geometry constraints. This can be verified by calculating the fundamental matrix or the essential matrix.

[0104] For example, given two images I1 and I2, and the matching points {m i1 ,m i2}, calculate the basic matrix F. For each matching point pair, it should satisfy (In homogeneous coordinates.) If most of the matching point pairs (such as more than a certain proportion, such as 80%) meet this constraint, the image alignment can be considered effective.

[0105] 2. Verification method based on image content consistency

[0106] 1. Intensity difference assessment

[0107] In the aligned common coordinate system, compare the pixel intensity differences at the same position between the matched images. For non-edge areas in the image, the pixel intensity should be consistent between images with different exposure levels.

[0108] It can be measured by calculating the absolute difference or mean square error (MSE) of pixel intensities. For example, for two aligned images A and B, at the pixel position (x, y), the intensities are I and A (x,y) and I B (x,y), calculate their absolute difference |I A (x,y)-I B (x,y)|. If this difference is less than a certain threshold in most areas (such as more than a certain proportion of pixels), it means that the image alignment is reasonable in terms of content strength.

[0109] 2. Gradient consistency check

[0110] Check the consistency of gradient direction and magnitude of pixels in the image. In the aligned images, gradients of the same object or scene area should have similar behavior in direction and magnitude.

[0111] A gradient operator (such as the Sobel operator) can be used to calculate the gradient of the image. For the matched images, the gradient vectors at the corresponding positions are compared. If the difference in the gradient direction is less than a certain angle threshold (such as 10-15 degrees) and the ratio of the gradient magnitude is within a reasonable range (such as between 0.8-1.2 times), the image alignment is considered qualified in terms of gradient consistency.

[0112] 3. Verification Method Using Statistical Information

[0113] 1. Correlation analysis

[0114] Calculate the correlation coefficient between the aligned images. The images can be regarded as two-dimensional signals, and the similarity between them can be evaluated by calculating the correlation between the signals.

[0115] For example, for two aligned images I1 and I2, their pixel values ​​can be regarded as two vectors V1 and V2 respectively, and their correlation coefficients can be calculated in and are the means of vectors V1 and V2 respectively. If the correlation coefficient is close to 1, it means that the similarity between the images is high and the alignment effect is good.

[0116] 2. Comparison of information entropy

[0117] Calculate the information entropy of each image, which can reflect the amount of information in the image. In the aligned images, although the exposure levels are different, the scene information they contain should be similar, so the information entropy should also be similar within a certain range.

[0118] For an image I, its information entropy H = -∑ i p(i)log2p(i), where P(i) is the probability that the pixel intensity in the image is i. Compare the information entropy of multiple aligned images. If their differences are within a certain range (such as less than a certain relative difference, such as 10%), the image alignment can be considered reasonable.

[0119] like Figure 3 As shown, the multiple images after the alignment process are subjected to image fusion process to generate a high dynamic range image, specifically:

[0120] Preprocessing the aligned images, mapping the dynamic ranges of the images to a common lower dynamic range using a tone mapping algorithm, and aligning all the images in a common coordinate system;

[0121] The weight of each pixel in each image is calculated, and multiple images with different exposure levels are fused according to the weight value to generate a high dynamic range image;

[0122] Perform detail enhancement and color correction on high dynamic range images to optimize image quality.

[0123] like Figure 4 As shown, the tone mapping algorithm is used to perform tone mapping on the fused image to obtain the target image, specifically:

[0124] Determine the target dynamic range, use the tone mapping algorithm to map the brightness of the high dynamic range image to the target dynamic range, and perform detail preservation and color correction to improve the visual effect of the image, such as sharpening edges or enhancing textures, and ensure color consistency and naturalness; finally, adjust the saturation and enhance the contrast to further optimize the visual effect of the final image.

[0125] It should be noted that compared with the existing image fusion methods, the present invention has significantly improved the clarity and color fidelity of the fused image, and has shown obvious advantages through subjective visual evaluation and measurement of objective indicators (such as peak signal-to-noise ratio PSNR, structural similarity index SSIM, etc.).

[0126] The feature points in each image are detected as follows:

[0127] The feature points in each image are detected using a feature detection algorithm, which includes: scale-invariant feature transformation, accelerated robust features, ORB feature detection algorithm, and Harris corner detection.

[0128] The feature descriptors include: SIFT descriptors, SURF descriptors and binary robust independent element features.

[0129] The feature point matching of the feature descriptors of the multiple images with different exposure levels is specifically as follows:

[0130] The matching algorithm includes: nearest neighbor search or K nearest neighbor search;

[0131] The plurality of matched images are subjected to transformation estimation and image transformation processing to be aligned to a common coordinate system, wherein the transformation estimation includes: translation transformation, rotation transformation, scaling transformation, affine transformation and projection transformation;

[0132] The aligning to a common coordinate system is specifically: aligning to a common coordinate system by resampling and interpolating the image; wherein the interpolation method includes: nearest neighbor interpolation, bilinear interpolation and bicubic interpolation.

[0133] The weight of each pixel in each image is calculated as follows:

[0134] The weights are calculated based on the brightness, contrast or local variance of the pixels of each image.

[0135] The tone mapping algorithms include: a global tone mapping algorithm, a local tone mapping algorithm, an adaptive tone mapping algorithm and a content-based tone mapping algorithm.

[0136] The local tone mapping algorithm of the present invention uses a dragging method to adjust the brightness and contrast of an image, and includes the following steps:

[0137] Sort the brightness of the high dynamic range image and calculate the cumulative probability of brightness;

[0138] Create a mapping function that maps the brightness of the high dynamic range image to a new brightness;

[0139] Use a mapping function to adjust the value of each pixel of a high dynamic range image.

[0140] The present invention also provides a device for implementing an image processing adaptation method based on a complex environment high dynamic range, comprising:

[0141] Memory for storing computer programs;

[0142] A processor is used to implement the steps of a method for image processing adaptation in a complex environment with a high dynamic range when executing the computer program.

[0143] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of an image processing adaptation method based on a high dynamic range in a complex environment.

[0144] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for image processing adaptation based on high dynamic range in complex environment, characterized in that: The following steps are involved: Collecting a plurality of images with different exposure levels, and performing image alignment processing on the plurality of images to obtain a plurality of images after the alignment processing; Performing image fusion processing on the multiple images after the alignment processing to generate a high dynamic range image; Use the tone mapping algorithm to perform tone mapping on the high dynamic range image to obtain the target image and perform noise reduction processing; The tone mapping algorithm is continuously optimized using the denoised target image to automatically adapt to different environments and lighting conditions.

2. The image processing adaptation method based on high dynamic range in complex environment according to claim 1, characterized in that: The collecting of multiple images with different exposure levels and performing image alignment processing on the multiple images to obtain multiple images after alignment processing is specifically: Detect the feature points in each image and generate corresponding feature descriptors from the feature points; Match feature points of feature descriptors of multiple images with different exposure levels and perform filtering to obtain multiple matched images; After transformation estimation and image transformation processing, multiple matched images are aligned in a common coordinate system and verified.

3. The image processing adaptation method based on high dynamic range in complex environment according to claim 1, characterized in that: The multiple images after the alignment process are subjected to image fusion process to generate a high dynamic range image, specifically: Preprocessing the aligned images, mapping the dynamic ranges of the images to a common lower dynamic range using a tone mapping algorithm, and aligning all the images in a common coordinate system; The weight of each pixel in each image is calculated, and multiple images with different exposure levels are fused according to the weight value to generate a high dynamic range image; Perform detail enhancement and color correction on high dynamic range images to optimize image quality.

4. The image processing adaptation method based on high dynamic range in complex environment according to claim 1, characterized in that: The tone mapping algorithm is used to perform tone mapping on the fused image to obtain the target image, specifically: Determine the target dynamic range, use the tone mapping algorithm to map the brightness of the high dynamic range image to the target dynamic range, and perform detail preservation and color correction.

5. The image processing adaptation method based on high dynamic range in complex environment according to claim 2, characterized in that: The feature points in each image are detected as follows: The feature points in each image are detected using a feature detection algorithm, which includes: scale-invariant feature transformation, accelerated robust features, ORB feature detection algorithm, and Harris corner detection.

6. The image processing adaptation method based on high dynamic range in complex environment according to claim 2, characterized in that: The feature descriptors include: SIFT descriptors, SURF descriptors and binary robust independent element features.

7. The image processing adaptation method based on high dynamic range in complex environment according to claim 2, characterized in that: The feature point matching of the feature descriptors of the multiple images with different exposure levels is specifically as follows: The matching algorithm includes: nearest neighbor search or K nearest neighbor search; The plurality of matched images are subjected to transformation estimation and image transformation processing to be aligned to a common coordinate system, wherein the transformation estimation includes: translation transformation, rotation transformation, scaling transformation, affine transformation and projection transformation; The aligning to a common coordinate system is specifically: aligning to a common coordinate system by resampling and interpolating the image; wherein the interpolation method includes: nearest neighbor interpolation, bilinear interpolation and bicubic interpolation.

8. The image processing adaptation method based on high dynamic range in complex environment according to claim 3 is characterized in that: The weight of each pixel in each image is calculated as follows: The weights are calculated based on the brightness, contrast or local variance of the pixels of each image.

9. The image processing adaptation method based on high dynamic range in complex environment according to claim 4, characterized in that: The tone mapping algorithms include: a global tone mapping algorithm, a local tone mapping algorithm, an adaptive tone mapping algorithm and a content-based tone mapping algorithm.

10. The image processing adaptation method based on high dynamic range in complex environment according to claim 9, characterized in that: The local tone mapping algorithm comprises the following steps: Sort the brightness of the high dynamic range image and calculate the cumulative probability of brightness; Create a mapping function that maps the brightness of the high dynamic range image to a new brightness; Use a mapping function to adjust the value of each pixel of a high dynamic range image.

Images