Self-supervised multi-task underwater image restoration method based on underwater imaging model

Abstract

The invention provides a self-supervised multi-task underwater image restoration method based on an underwater imaging model, and the method comprises the following steps: carrying out the color cast correction of an underwater image collected by an underwater robot through an adaptive color balance algorithm, and obtaining an image after the color cast correction; obtaining the transmissivity of the image after the color cast correction through a transmissivity estimation network; enabling the image after color cast correction to pass through a background light estimation network to obtain background light; enabling the image after color cast correction to pass through a clear image reconstruction network to obtain a reconstructed clear image; and the background light and the transmissivity of the reconstructed clear image are synthesized through a Jaffe-McGammery underwater imaging model, and training is carried out in a self-supervision mode to obtain a final enhanced image. According to the method, an end-to-end self-supervised learning architecture is adopted, network training can be completed without large-scale annotation of a data set, the method has the advantages of low training cost and high generalization ability, and efficient image restoration can be realized in a complex and changeable underwater environment.

Description

Technical Field

[0001] This invention relates to the technical field of underwater image processing, and more particularly to a self-supervised multi-task underwater image restoration method based on an underwater imaging model. Background Technology

[0002] Underwater optical imaging technology, enabling underwater robots or autonomous underwater vehicles to acquire underwater images, has become a crucial tool for deep-sea exploration. However, dissolved substances and suspended particles in water exert complex effects on light through absorption, refraction, and scattering, leading to image degradation phenomena such as color shift, decreased contrast, and blurred details. These degradations not only affect visual perception but also pose significant challenges to tasks relying on optical imaging, such as pattern recognition, target detection, and feature extraction. Therefore, researching underwater image sharpening techniques is of great importance for improving the quality of optical images in complex aquatic environments and enhancing the performance of underwater vision tasks.

[0003] There are three main types of existing underwater image sharpening methods: physical model-based image restoration methods, model-free image enhancement methods, and deep learning-based image enhancement methods. Model-based image restoration methods are based on underwater imaging models and use prior knowledge and assumptions to estimate transmission patterns and background light to eliminate attenuation caused by light absorption and scattering, thereby improving image sharpness. Model-free image enhancement methods improve upon traditional image processing techniques for underwater scenes by adjusting pixel intensity distribution in the spatial or transform domains to enhance contrast and restore colors. Deep learning-based image enhancement methods employ an end-to-end data-driven strategy, training networks using large-scale labeled data to achieve quality enhancement of underwater images.

[0004] While model-based image restoration methods can demonstrate excellent performance under specific underwater conditions, their high dependence on imaging priors limits their applicability in complex and variable aquatic environments. Non-model-based augmentation methods, by ignoring the degradation mechanisms of underwater images and the imaging model, often result in noise amplification and color distortion. Deep learning-based image augmentation methods require end-to-end training with a large number of paired samples. However, in the field of underwater imaging, obtaining distortion-free real images as references is extremely difficult. Summary of the Invention

[0005] In view of the technical problems mentioned in the background section, a self-supervised multi-task underwater image restoration method based on an underwater imaging model is provided.

[0006] The technical means employed in this invention are as follows: A self-supervised multi-task underwater image restoration method based on an underwater imaging model includes the following steps: Step 1: Use an adaptive color balance algorithm to correct the color cast of underwater images acquired by an underwater robot to obtain the color cast corrected image; Step 2: Pass the color-corrected image through a transmittance estimation network to obtain the transmittance of the image; Step 3: Pass the color-corrected image through a background light estimation network to obtain the background light; Step 4: The color-corrected image is processed through a sharp image reconstruction network to obtain a reconstructed sharp image; Step 5: The background light and transmittance of the reconstructed clear image are synthesized using the Jaffe-McGlamery underwater imaging model and trained in a self-supervised manner to obtain the final enhanced image.

[0007] Furthermore, in step 1, the mathematical expression for color cast correction is: ; in, Represents the original underwater image At pixel The passage Pixel values; Indicates The local neighborhood centered on the center; Indicates in The pixel values ​​within the range are filtered for maximum values; This indicates that the maximum value is taken from the maximum value filtering results of the three channels; The generated maximum channel map; ; in, This represents the color-corrected image. This represents the compensation factor used to adjust brightness; This indicates a round-down operation; This represents the global average intensity of the original underwater image across the three channels.

[0008] Furthermore, in step 2, to preserve the detail information in the shallow features, the projection map estimation network adopts a non-degenerate same-size structure and does not introduce downsampling operations; the transmittance estimation network includes: alternating stacked multi-layer convolutional layers, batch normalization and linear correction units; and the last layer of the transmittance estimation network uses Sigmoid activation to normalize the output to the range [0,1].

[0009] Furthermore, step 3 includes the following steps: Step 31: The encoder maps the input color-corrected image to a latent space through multiple convolutional layers. The latent space outputs hidden layer variables. Used to approximate the true posterior distribution ; Step 32: Adjust hidden layer variables through intermediate blocks The mapping is performed to generate two 100-dimensional vectors representing the underlying Gaussian distribution. ;in, Represents the mean vector. Represents the variance vector of the Gaussian model; Step 33: Use the reparameterization technique to sample from the standard normal distribution to obtain... ; Step 34, through Derive the hidden variables of the reconstruction That is ; Step 35: The decoder will resample the obtained data. As input, the background light is reconstructed inversely through convolutional layers, and the pixel values ​​are constrained to the range [0,1] by the Sigmoid function; Step 36: Introduce a hierarchical search strategy based on quadtree subdivision to obtain... Further refine the initial background light estimate.

[0010] Furthermore, step 36 includes the following steps: Step 361: Initial division: Divide the color-corrected image into four sub-regions of equal size; Step 362, Region Scoring: Assign a score to each sub-region; the score is obtained by subtracting the standard deviation of the pixel values ​​within the region from the average pixel value; Step 363, Iterative subdivision: Select the region with the highest score from all sub-regions in the current layer as a candidate region, continue to divide it into four sub-regions, and repeat the scoring and selection process; Step 364: Repeat steps 361-363 until the pixel value in the candidate region is less than [a certain value]. This yields the final background light.

[0011] Furthermore, in step 4, the clear image reconstruction network includes: a downsampling module and an upsampling module; the downsampling module consists of several sets of Conv-ReLU; the upsampling module amplifies the upsampled high-level feature map through bilinear interpolation and spatially aligns it to the corresponding low-level features; The working process of the sharp image reconstruction network is represented as follows: ; ; in, and These represent the upsampling module and the downsampling module, respectively. Represents a join operation; in addition, , and These represent the input, upsampled, and feature maps after feature enhancement, respectively.

[0012] Furthermore, step 5 includes the following steps: Step 51: During underwater optical imaging, the absorption and scattering of light by suspended particles leads to a significant attenuation of light intensity. Therefore, the underwater image is represented as follows: ; in, This refers to the light that reaches the camera directly after physical reflection, i.e., the direct reflection component; This represents the light scattered back toward the camera by suspended particles in the water, i.e., the backscattering component. The forward scattering component refers to the blurry halo formed by forward scattering of light during its propagation. Step 52: Since forward scattering is usually much smaller than direct reflection and backscattering, it can be ignored, thus simplifying the underwater imaging model to: ; in, Indicates a clear image. This represents a transmission diagram, reflecting the degree of attenuation of light as it propagates through water. This represents background light, which originates from ambient light components scattered in all directions from the water body. Step 53: Establish a self-supervised multi-task underwater image restoration method based on an underwater imaging model, and train it in a self-supervised training manner; by comparing the input underwater image with the image reconstructed from the clear image and the background light based on the network estimation, calculate the reconstruction error, and use this error to update the network parameters of the projection map estimation through backpropagation; Step 54: Introduce the loss function Underwater images after reconstructing the unwrapped layers :

[0013] in, The calculation of the value involves merging the outputs of the three sub-networks, and the joint training objective of the entire network can be expressed as the sum of the losses of each sub-network. Step 55: The total loss during network training is summarized as follows: .

[0014] Furthermore, the total loss of the background light estimation network includes: reconstruction loss, KL divergence loss, and regularization loss; the total loss is: ; ; ; ; in, Indicates reconstruction loss and measures reconstruction background light. and initial background light Differences between them: Describes the KL third-order loss, which makes the posterior distribution... Close to the prior distribution : The neighborhood regularization loss is represented by a smooth estimation using pixel neighborhood consistency constraints: Represents pixels The set of pixels in the neighborhood, This indicates the number of pixels in the neighborhood.

[0015] Furthermore, the loss function used by the sharp image reconstruction network is: ; in, Denotes the p-norm of a matrix. express brightness, Indicates its saturation; loss For an ideal, distortion-free image, its brightness and saturation should be highly consistent in value. It should approach 0, that is .

[0016] Compared with the prior art, the present invention has the following advantages: This invention addresses the problems of color cast, low visibility, and blurriness in underwater images. Based on an underwater imaging model, it proposes a multi-task self-supervised neural network-based underwater image restoration method. This method significantly improves the clarity and contrast of underwater images, effectively supporting applications such as seabed target detection, target recognition, dock maintenance, and intelligent vessel navigation. Furthermore, this method employs an end-to-end self-supervised learning architecture, eliminating the need for large-scale labeled datasets for network training. It boasts advantages such as low training cost and strong generalization ability, enabling efficient image restoration in complex and ever-changing underwater environments. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the underwater imaging model of the present invention.

[0019] Figure 2 This is a schematic diagram of the multi-task self-supervised network of the present invention.

[0020] Figure 3 This is a schematic diagram of the multi-level dense feature enhancement unit of the present invention.

[0021] Figure 4 This is a schematic diagram of the initial background light estimation method of the present invention.

[0022] Figure 5 This is a schematic diagram of the overall process of the present invention. Detailed Implementation

[0023] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0024] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0025] like Figure 1-4 As shown, this invention provides a self-supervised multi-task underwater image restoration method based on an underwater imaging model, comprising the following steps: Step 1: Apply an adaptive color balance algorithm to correct the color cast of underwater images acquired by the underwater robot, resulting in a color-corrected image. The mathematical expression for color cast correction in Step 1 is as follows: ; in, Represents the original underwater image At pixel The passage Pixel values; Indicates The local neighborhood centered on the center; Indicates in The pixel values ​​within the range are filtered for maximum values; This indicates that the maximum value is taken from the maximum value filtering results of the three channels; The generated maximum channel map; ; in, This represents the color-corrected image. This represents the compensation factor used to adjust brightness; This indicates a round-down operation; This represents the global average intensity of the original underwater image across the three channels.

[0026] Step 2: The color-corrected image is passed through a transmittance estimation network to obtain the transmittance of the image. In Step 2, to preserve the detail information in the shallow features, the projection map estimation network adopts a non-degenerate same-size structure and does not introduce downsampling operations. The transmittance estimation network includes: alternating stacked multi-layer convolutional layers, batch normalization, and linear correction units. The last layer of the transmittance estimation network uses Sigmoid activation to normalize the output to the range [0,1].

[0027] Step 3: Pass the color-corrected image through a background light estimation network to obtain the background light; Step 3 includes the following steps: Step 31: The encoder maps the input color-corrected image to a latent space through multiple convolutional layers. The latent space outputs hidden layer variables. Used to approximate the true posterior distribution ; Step 32: Adjust hidden layer variables through intermediate blocks The mapping is performed to generate two 100-dimensional vectors representing the underlying Gaussian distribution. ;in, Represents the mean vector. Represents the variance vector of the Gaussian model; Step 33: Use the reparameterization technique to sample from the standard normal distribution to obtain... ; Step 34, through Derive the hidden variables of the reconstruction That is ; Step 35: The decoder will resample the obtained data. As input, the background light is reconstructed inversely through convolutional layers, and the pixel values ​​are constrained to the range [0,1] by the Sigmoid function; Step 36: Introduce a hierarchical search strategy based on quadtree subdivision to obtain... Further refine the initial background light estimation; step 36 includes the following steps: Step 361: Initial division: Divide the color-corrected image into four sub-regions of equal size; Step 362, Region Scoring: Assign a score to each sub-region; the score is obtained by subtracting the standard deviation of the pixel values ​​within the region from the average pixel value; Step 363, Iterative subdivision: Select the region with the highest score from all sub-regions in the current layer as a candidate region, continue to divide it into four sub-regions, and repeat the scoring and selection process; Step 364: Repeat steps 361-363 until the pixel value in the candidate region is less than [a certain value]. This yields the final background light.

[0028] Step 4: The color-corrected image is processed through a sharp image reconstruction network to obtain a reconstructed sharp image. In Step 4, the sharp image reconstruction network includes a downsampling module and an upsampling module. The downsampling module consists of several sets of Conv-ReLU. The upsampling module enlarges the upsampled high-level feature map through bilinear interpolation and spatially aligns it to the corresponding low-level features. The working process of the sharp image reconstruction network is represented as follows: ; ; in, and These represent the upsampling module and the downsampling module, respectively. Represents a join operation; in addition, , and These represent the input, upsampled, and feature maps after feature enhancement, respectively.

[0029] Step 5: The background light and transmittance of the reconstructed sharp image are synthesized using the Jaffe-McGlamery underwater imaging model, and trained in a self-supervised manner to obtain the final enhanced image. Step 5 includes the following steps: Step 51: During underwater optical imaging, the absorption and scattering of light by suspended particles leads to a significant attenuation of light intensity. Therefore, the underwater image is represented as follows: ; in, This refers to the light that reaches the camera directly after physical reflection, i.e., the direct reflection component; This represents the light scattered back toward the camera by suspended particles in the water, i.e., the backscattering component. The forward scattering component refers to the blurry halo formed by forward scattering of light during its propagation. Step 52: Since forward scattering is usually much smaller than direct reflection and backscattering, it can be ignored, thus simplifying the underwater imaging model to: ; in, Indicates a clear image. This represents a transmission diagram, reflecting the degree of attenuation of light as it propagates through water. This represents background light, which originates from ambient light components scattered in all directions from the water body. Step 53: Establish a self-supervised multi-task underwater image restoration method based on an underwater imaging model, and train it in a self-supervised training manner; by comparing the input underwater image with the image reconstructed from the clear image and the background light based on the network estimation, calculate the reconstruction error, and use this error to update the network parameters of the projection map estimation through backpropagation; Step 54: Introduce the loss function Underwater images after reconstructing the unwrapped layers :

[0030] in, The calculation of the value involves merging the outputs of the three sub-networks, and the joint training objective of the entire network can be expressed as the sum of the losses of each sub-network. Step 55: The total loss during network training is summarized as follows: .

[0031] Furthermore, the total loss of the background light estimation network includes: reconstruction loss, KL divergence loss, and regularization loss; the total loss is: ; ; ; ; in, Indicates reconstruction loss and measures reconstruction background light. and initial background light Differences between them: Describes the KL third-order loss, which makes the posterior distribution... Close to the prior distribution : The neighborhood regularization loss is represented by a smooth estimation using pixel neighborhood consistency constraints: Represents pixels The set of pixels in the neighborhood, This indicates the number of pixels in the neighborhood.

[0032] Furthermore, the loss function used by the sharp image reconstruction network is: ; in, Denotes the p-norm of a matrix. express brightness, Indicates its saturation; loss For an ideal, distortion-free image, its brightness and saturation should be highly consistent in value. It should approach 0, that is .

[0033] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0034] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0035] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0036] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0037] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0038] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0039] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A self-supervised multi-task underwater image restoration method based on an underwater imaging model, characterized in that, Includes the following steps: Step 1: Use an adaptive color balance algorithm to correct the color cast of underwater images acquired by an underwater robot to obtain the color cast corrected image; Step 2: Pass the color-corrected image through a transmittance estimation network to obtain the transmittance of the image; Step 3: Pass the color-corrected image through a background light estimation network to obtain the background light; Step 4: The color-corrected image is processed through a sharp image reconstruction network to obtain a reconstructed sharp image; Step 5: The background light and transmittance of the reconstructed clear image are synthesized using the Jaffe-McGlamery underwater imaging model and trained in a self-supervised manner to obtain the final enhanced image.

2. The self-supervised multi-task underwater image restoration method based on an underwater imaging model according to claim 1, characterized in that, In step 1, the mathematical expression for color cast correction is: ； in, Represents the original underwater image At pixel The passage Pixel values; Indicates The local neighborhood centered on the center; Indicates in The pixel values ​​within the range are filtered for maximum values; This indicates that the maximum value is taken from the maximum value filtering results of the three channels; The generated maximum channel map; ； in, This represents the color-corrected image. This represents the compensation factor used to adjust brightness; This indicates a round-down operation; This represents the global average intensity of the original underwater image across the three channels.

3. The self-supervised multi-task underwater image restoration method based on an underwater imaging model according to claim 1, characterized in that, In step 2, to preserve the detail information in the shallow features, the projection map estimation network adopts a non-degenerate same-size structure and does not introduce downsampling operations. The transmittance estimation network includes: alternating stacked multilayer convolutional layers, batch normalization, and linear correction units; and the last layer of the transmittance estimation network uses Sigmoid activation to normalize the output to the range [0,1].

4. The self-supervised multi-task underwater image restoration method based on an underwater imaging model according to claim 1, characterized in that, Step 3 includes the following steps: Step 31: The encoder maps the input color-corrected image to a latent space through multiple convolutional layers. The latent space outputs hidden layer variables. Used to approximate the true posterior distribution ; Step 32: Adjust hidden layer variables through intermediate blocks The mapping is performed to generate two 100-dimensional vectors representing the underlying Gaussian distribution. ;in, Represents the mean vector. Represents the variance vector of the Gaussian model; Step 33: Use the reparameterization technique to sample from the standard normal distribution to obtain... ; Step 34, through Derive the hidden variables of the reconstruction That is ; Step 35: The decoder will resample the obtained data. As input, the background light is reconstructed inversely through convolutional layers, and the pixel values ​​are constrained to the range [0,1] by the Sigmoid function; Step 36: Introduce a hierarchical search strategy based on quadtree subdivision to obtain... Further refine the initial background light estimate.

5. The self-supervised multi-task underwater image restoration method based on an underwater imaging model according to claim 4, characterized in that, Step 36 includes the following steps: Step 361: Initial division: Divide the color-corrected image into four sub-regions of equal size; Step 362, Region Scoring: Assign a score to each sub-region; the score is obtained by subtracting the standard deviation of the pixel values ​​within the region from the average pixel value; Step 363, Iterative subdivision: Select the region with the highest score from all sub-regions in the current layer as a candidate region, continue to divide it into four sub-regions, and repeat the scoring and selection process; Step 364: Repeat steps 361-363 until the pixel value in the candidate region is less than [a certain value]. This yields the final background light.

6. The self-supervised multi-task underwater image restoration method based on an underwater imaging model according to claim 1, characterized in that, In step 4, the clear image reconstruction network includes: a downsampling module and an upsampling module; the downsampling module consists of several sets of Conv-ReLU; the upsampling module amplifies the upsampled high-level feature map through bilinear interpolation and spatially aligns it to the corresponding low-level features; The working process of the sharp image reconstruction network is represented as follows: ； ； in, and These represent the upsampling module and the downsampling module, respectively. Represents a join operation; in addition, , and These represent the input, upsampled, and feature maps after feature enhancement, respectively.

7. The self-supervised multi-task underwater image restoration method based on an underwater imaging model according to claim 1, characterized in that, Step 5 includes the following steps: Step 51: During underwater optical imaging, the absorption and scattering of light by suspended particles leads to a significant attenuation of light intensity. Therefore, the underwater image is represented as follows: ； in, This refers to the light that reaches the camera directly after physical reflection, i.e., the direct reflection component; This represents the light scattered back toward the camera by suspended particles in the water, i.e., the backscattering component. The forward scattering component refers to the blurry halo formed by forward scattering of light during its propagation. Step 52: Since forward scattering is usually much smaller than direct reflection and backscattering, it can be ignored, thus simplifying the underwater imaging model to: ； in, Indicates a clear image. This represents a transmission diagram, reflecting the degree of attenuation of light as it propagates through water. This represents background light, which originates from ambient light components scattered in all directions from the water body. Step 53: Establish a self-supervised multi-task underwater image restoration method based on an underwater imaging model, and train it in a self-supervised training manner; by comparing the input underwater image with the image reconstructed from the clear image and the background light based on the network estimation, calculate the reconstruction error, and use this error to update the network parameters of the projection map estimation through backpropagation; Step 54: Introduce the loss function Underwater images after reconstructing the unwrapped layers : in, The calculation of the value involves merging the outputs of the three sub-networks, and the joint training objective of the entire network can be expressed as the sum of the losses of each sub-network. Step 55: The total loss during network training is summarized as follows: 。 8. The self-supervised multi-task underwater image restoration method based on an underwater imaging model according to claim 1, characterized in that, The total loss of the background light estimation network includes: reconstruction loss, KL divergence loss, and regularization loss; the total loss is: ； ； ； ； in, Indicates reconstruction loss and measures reconstruction background light. and initial background light Differences between them: Describes the KL third-order loss, which makes the posterior distribution... Close to the prior distribution : The neighborhood regularization loss is represented by a smooth estimation using pixel neighborhood consistency constraints: Represents pixels The set of pixels in the neighborhood, This indicates the number of pixels in the neighborhood.

9. The self-supervised multi-task underwater image restoration method based on an underwater imaging model according to claim 1, characterized in that, The loss function used by the sharp image reconstruction network is: ； in, Denotes the p-norm of a matrix. express brightness, Indicates its saturation; loss For an ideal, distortion-free image, its brightness and saturation should be highly consistent in value. It should approach 0, that is 。

Images