A method for color correction of farmed fish images with phase-constrained degradation perception

Abstract

The application provides a farmed fish image color correction method based on phase constraint degradation perception, aiming at improving the visibility of complex farmed fish underwater images, and comprises the following steps: constructing a color degradation weight map according to the attenuation intensity of different color channels of the farmed fish image; using the color degradation weight guide image to reserve the most complete channel to compensate the recession channel; constructing an average phase loss constraint compensation intensity to avoid over-correction; calculating the channel loss as an iterative threshold to iteratively optimize the correction parameters; the color degradation weight provided by the application aims to solve the non-uniform recession phenomenon caused by the selective absorption of water and the depth of field difference in intelligent aquaculture; the application can effectively restore the missing color and texture information in the underwater farming image, alleviate the problem of over-enhanced long-range view and insufficient near-range details, adapt to different water quality and water depth environments, and be used for fish identification and growth monitoring in intelligent aquaculture.

Description

Technical Field

[0001] This invention relates to the field of image color correction technology, specifically to a phase-constrained degradation-sensing method for color correction of farmed fish images. Background Technology

[0002] In aquaculture, the use of underwater vision systems for intelligent monitoring of fish behavior, growth status, and diseases has become a development trend. However, the aquatic environment of aquaculture ponds, cages, or recirculating aquaculture systems typically contains large amounts of feed residue, excrement, microalgae, and suspended sediment, resulting in extremely high turbidity, severe light attenuation, and highly non-uniform characteristics. The acquired underwater images often suffer from problems such as a blue / blue-green tint in the foreground, blurry foreground, and low local contrast. Key information such as fish body color and texture is severely damaged, directly affecting the accuracy of subsequent analysis tasks such as vision-based fish classification, counting, and health assessment.

[0003] Traditional visual monitoring of aquaculture fish relies primarily on manual inspections and experience-based judgment. However, limited by the vast aquaculture area, high labor costs, and complex underwater environment, this approach suffers from low monitoring efficiency, limited coverage, and insufficient real-time performance, failing to meet the demands of continuous fish status perception in large-scale aquaculture. With the development of underwater computer vision technology, image enhancement and intelligent perception methods are increasingly being applied to environmental monitoring of aquaculture fish, improving automated monitoring capabilities to some extent. However, due to suspended particles, light attenuation, water turbidity, and dynamic interference from fish movement in aquaculture water, underwater images typically exhibit significant color distortion, decreased contrast, and loss of detail, leading to problems such as blurred target boundaries, insufficient feature information, and reduced detection accuracy in fish target detection, vital sign recognition, and behavior analysis. Especially in deep-sea cages and high-density aquaculture environments, complex non-uniform degradation further limits the accurate perception of individual fish status by the visual system. Therefore, there is an urgent need for an underwater image enhancement method that can adapt to complex aquaculture environments, effectively restore the visual information of aquaculture fish, and improve the reliability of subsequent intelligent detection and status analysis. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to overcome the existing defects and provide a phase-constrained degradation perception color correction method for farmed fish images, which can effectively solve the problems in the background technology.

[0005] To achieve the above objectives, this invention discloses a color correction method for cultured fish images based on phase-constrained degradation sensing. The technical solution adopted includes the following steps: Step 1, Obtain the input image It is decomposed into red, green and blue color channels, and the average value is calculated to obtain the channel that is most fully preserved; Step 2: Calculate the blue dominant intensity, the red-green joint attenuation intensity, and the red-green intensity difference. The blue dominant intensity is used to evaluate the information advantage of the blue channel relative to the red and green channels. The red-green joint attenuation intensity is used to describe the joint attenuation characteristics of the red and green channels relative to the blue channel. The red-green intensity difference is used to measure the intensity advantage of the green channel relative to the red channel. Step 3: Calculate the color degradation weights of the red and green channels by combining the blue dominant intensity, the red-green joint attenuation intensity, and the red-green intensity difference; Step 4: Calculate the basic compensation weights based on the differences between the bright channel and the channel that retains the most complete image and other channels in the input image; Step 5: Use the color degradation weights obtained in Step 3 and the basic compensation weights obtained in Step 4 to perform channel compensation to correct color cast. Step 6: Calculate the average phase error of the compensated red, green, and blue channels compared to the input image; Step 7: Calculate the channel loss of the compensated image. If it is less than the preset threshold, end the iteration and output the final color-corrected image. Otherwise, use the compensated image as the new input and repeat steps 1 to 7 until the termination condition is met.

[0006] As a preferred embodiment of the present invention, the formula for calculating the average value of each color channel in step 1 is as follows: in, and Indicates the input image The height and width have the highest average value. .

[0007] As a preferred embodiment of the present invention, the formula for the blue dominant intensity in step 2 is: in, The dominant blue intensity is used to determine the severity of attenuation. , and These are the r, g, and b channels of the input image, respectively.

[0008] As a preferred embodiment of the present invention, the formula for calculating the combined red-green attenuation intensity in step 2 is as follows: in, The combined red-green attenuation intensity.

[0009] As a preferred embodiment of the present invention, the formula for the red-green intensity difference in step 2 is: in, The difference is due to the intensity of red and green.

[0010] As a preferred embodiment of the present invention, the specific steps of step 3 are as follows: Step 3.1, calculate the overall intensity of the degradation feature: in, Used to characterize the overall intensity of degradation features in the current image; mean(·) and std(·) are the mean function and standard deviation function of the matrix, respectively; Step 3.2, calculate the color degradation weight: in, The color degradation weight for the red channel, The color degradation weight for the green channel.

[0011] As a preferred embodiment of the present invention, the formula for the basic compensation weight in step 4 is: in, Basic compensation weight, This is the brightness channel of the original image. For the color channels of the input image.

[0012] As a preferred embodiment of the present invention, the formula for channel compensation in step 5 is: in, For the compensated image, ; , which represents the phase constraint weight; The phase difference threshold, For the average phase error, in hour, .

[0013] As a preferred embodiment of the present invention, the formula for the average phase error in step 6 is: in, For phase spectrum, The number of pixels in the image. The imaginary unit, .

[0014] As a preferred embodiment of the present invention, the formula for the channel loss in step 7 is: in, For channel loss, when When the value is less than 0.01, it is considered that each color channel has reached a stable equilibrium state, color distortion has been effectively corrected, and the iterative optimization process is terminated.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention constructs color degradation perception weights and adaptively adjusts the compensation intensity according to the degree of degradation in different regions, thereby achieving accurate restoration of color information in different depth regions and effectively avoiding the problems of over-enhancement of local regions and insufficient color restoration in traditional methods; furthermore, by introducing a phase consistency constraint mechanism, the structural changes in the color compensation process are dynamically monitored through frequency domain structural information, and the compensation amplitude is adaptively adjusted, which effectively maintains the fish body outline, texture and detail features while enhancing visual quality. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the process of the present invention; Figure 2 This is a schematic diagram of the functional architecture of the present invention; Figure 3 This is a diagram showing the color correction results of fish images in blue-distorted waters according to the present invention; Figure 4 This is a diagram showing the color correction results of fish images in blue-green distorted waters according to the present invention; Figure 5 This is a 3D color scatter plot of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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 are within the scope of protection of the present invention.

[0018] Example 1 like Figures 1 to 2 As shown, this invention discloses a phase-constrained degradation-sensing color correction method for farmed fish images. The technical solution adopted includes the following steps: Step 1: Obtain the input image of farmed fish. It is decomposed into red, green and blue color channels, and the average value is calculated to obtain the channel that is most fully preserved; The formula for calculating the average value of each color channel is: in, and Indicates the input image The height and width have the highest average value. ; Step 2, calculate the blue dominance intensity to assess the information advantage of the blue channel relative to the red and green channels, specifically: The formula for calculating the blue dominant intensity based on three-channel information is: in, The dominant blue intensity is used to determine the severity of attenuation. , and These are the r, g, and b channels of the input image, respectively; Step 3: Calculate the combined red-green attenuation intensity, which describes the combined attenuation characteristics of the red and green channels relative to the blue channel; the formula for calculating the combined red-green attenuation intensity is: in, The combined red-green attenuation intensity; Step 4: Calculate the red-green intensity difference, which measures the intensity advantage of the green channel relative to the red channel; the formula for the red-green intensity difference is: in, The difference in red and green intensity; in underwater fading images, green is generally stronger than red, calculated... The purpose is to determine whether there is a special case where green is not stronger than red. If the result is non-zero, it is assumed that the color before local decay should be red. If it is zero, it means that green is dominant, which is a general case. This will directly affect the color degradation weight in step 5. Step 5: Calculate the color degradation weights for the red and green channels by combining the dominant blue intensity, the combined red-green attenuation intensity, and the red-green intensity difference; the specific color degradation weights are as follows: in, Used to characterize the overall intensity of degradation features in the current image; mean(·) is the matrix averaging function, and std(·) is the matrix standard deviation function; in, The color degradation weight for the red channel, Assign color degradation weights to the green channel; Step 6: Construct basic compensation weights based on the differences between the bright channel and the channel that retains the most complete image and other channels in the input image: in, Basic compensation weight; For the input image The bright channel, by input image Obtained by converting to the L channel of CIELab space; These are the color channels of the input image, with values ​​ranging from 0 to 1. Step 7: Perform channel compensation using base compensation weights and color degradation weights to correct color cast; specifically: in, For the compensated image, ; , which represents the phase constraint weight; The phase difference threshold, For the average phase error, in hour, ; Step 8: Calculate the average phase error of the compensated red, green, and blue channels relative to the source image: in, The phase spectrum is obtained by performing a Fourier transform on the compensated image. The number of pixels in the image. The imaginary unit, ; Step 9: Calculate the channel loss of the compensated image: in, For channel loss, To compensate for the average value of each color channel in the resulting image, To compensate for the maximum value among the average values ​​of each color channel in the resulting image. and Indicates the compensated image Height and width; in all passageways If all values ​​are less than 0.01, the iteration ends and the image (the final compensated image, i.e., the color-corrected image) is output. Otherwise, the compensated image is used as the new input image, and steps 1 to 9 are repeated until the termination condition is met.

[0019] Additional notes: During the first run, The value is 1, and in subsequent iterations, the value from the previous round is used. Value for this round Determining the value.

[0020] Comparison of fish image color correction between Example 1 (i.e., "this method" in the figure) and existing technologies (GE, MAXRGB, SOG): like Figure 3 The image shows a comparison of color correction results for fish images under a blue-distorted environment. This environment is dominated by blue light from the water, with uneven light attenuation and color channel shifts. The color features of the target object and the background are severely distorted, and boundary features are blurred. Furthermore, underwater suspended objects and light refraction further exacerbate the difficulty of color correction and increase recognition errors. The correction results show that traditional methods fail in environments with severe color degradation, making it difficult to correct severe color distortion. In contrast, Example 1 can adaptively correct the channel attenuation features in a blue-distorted environment, effectively restoring the true color information of the fish and background areas, while avoiding overcorrection and color distortion in local areas. It exhibits superior color recovery capability and robustness in complex underwater blue-degradation scenarios.

[0021] like Figure 4 The image shown is the color correction result for a fish image under a blue-green distorted environment. This environment is mainly affected by the strong absorption of red light by the water and the dominance of the blue-green spectrum, resulting in a significant blue-green shift in the overall image, severe loss of the original color information of the fish, and color mixing in some areas. The correction results show that traditional methods fail in environments with severe color degradation and are unable to correct severe color distortion. In contrast, this invention can effectively sense the color attenuation characteristics of different areas under blue-green distorted environments, specifically compensate for damaged color channels, restore the true color distribution of the fish and background areas, and avoid local over-enhancement and color drift problems. It exhibits stronger color recovery capabilities and environmental adaptability in complex blue-green degradation scenarios.

[0022] like Figure 5The figure shows a comparison of the 3D color space scatter distribution of Example 1, the input image, and existing technologies to visualize the color restoration capabilities of different methods. As shown, the pixels in the original degraded image are mainly clustered in the blue area, and the overall distribution range is relatively narrow. This indicates that due to wavelength-selective attenuation during underwater light propagation, the red and green channel information is severely lost, leading to image color shift and color distribution imbalance. After processing with traditional color correction methods, although the point cloud distribution range expands somewhat, it is still concentrated near the blue area, showing a significant blue shift. This indicates that these methods mainly rely on global statistical information for compensation and are difficult to effectively restore the damaged weak color channel information. In contrast, the image processed by Example 1 shows that its 3D color scatter distribution gradually migrates from the original single blue cluster area to the central area of ​​the RGB space, exhibiting a more balanced and continuous distribution trend. This shows that Example 1 can effectively compensate for the loss of red and green colors due to underwater attenuation, achieving a more reasonable balance between the color channels. Simultaneously, the expansion of the color point cloud range also reflects that the enhanced image has a richer color expression capability and can restore more potential scene information. Therefore, the 3D color space distribution results further validate the effectiveness of Example 1 in underwater color distortion correction and color fidelity restoration.

[0023] The performance of Example 1 was objectively evaluated using underwater image objective evaluation metrics, including UIQM (Underwater Image Quality Measurement) and UCIQE (Underwater Color Image Quality Evaluation). A higher UIQM value indicates colors closer to human vision, higher sharpness, and better contrast; a higher UCIQE value indicates higher image quality. The objective metric results for color restoration in blue distortion and blue-green distortion environments for Example 1 and existing technologies are as follows: It can be seen that in both types of image distortion environments, the UIQM and UCIQE values ​​of Example 1 are the highest, indicating that Example 1 can correct the input image color to a high-definition, high-contrast, high-quality image that is close to human vision in both color distortion environments.

[0024] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for color correction of cultured fish images using phase-constrained degradation sensing, characterized in that, Includes the following steps: Step 1, Obtain the input image It is decomposed into red, green, and blue color channels; Step 2: Calculate the dominant blue intensity, the combined red-green attenuation intensity, and the red-green intensity difference; Step 3: Calculate the color degradation weights of the red and green channels by combining the blue dominant intensity, the red-green joint attenuation intensity, and the red-green intensity difference; Step 4: Calculate the basic compensation weights based on the differences between the bright channel and the channel that retains the most complete image and other channels in the input image; Step 5: Use the color degradation weights obtained in Step 3 and the basic compensation weights obtained in Step 4 to perform channel compensation to correct color cast. Step 6: Calculate the average phase error of the compensated red, green, and blue channels compared to the input image; Step 7: Calculate the channel loss of the compensated image. If it is less than the preset threshold, end the iteration and output the final color-corrected image. Otherwise, use the compensated image as the new input and repeat steps 1 to 7 until the termination condition is met.

2. The color correction method for cultured fish images based on phase-constrained degradation sensing according to claim 1, characterized in that: The formula for calculating the average value of each color channel in step 1 is as follows: in, and Indicates the input image The height and width have the highest average value. .

3. The color correction method for cultured fish images based on phase-constrained degradation sensing according to claim 2, characterized in that: The formula for the blue dominant intensity in step 2 is: in, The dominant intensity is blue. , and These are the r, g, and b channels of the input image, respectively.

4. The color correction method for cultured fish images based on phase-constrained degradation sensing according to claim 3, characterized in that: The formula for calculating the combined red-green attenuation intensity in step 2 is as follows: in, The combined red-green attenuation intensity.

5. The color correction method for cultured fish images based on phase-constrained degradation sensing according to claim 4, characterized in that: The formula for the red-green intensity difference in step 2 is: in, The difference is due to the intensity of red and green.

6. The color correction method for cultured fish images based on phase-constrained degradation sensing according to claim 5, characterized in that, The specific steps of step 3 are as follows: Step 3.1, calculate the overall intensity of the degradation feature: in, Used to characterize the overall intensity of degradation features in the current image; mean(·) and std(·) are the mean function and standard deviation function of the matrix, respectively; Step 3.2, calculate the color degradation weight: in, The color degradation weight for the red channel, The color degradation weight for the green channel.

7. The color correction method for cultured fish images based on phase-constrained degradation sensing according to claim 6, characterized in that: The formula for the basic compensation weight in step 4 is: in, Basic compensation weight, This is the brightness channel of the original image. For the color channels of the input image.

8. The color correction method for cultured fish images based on phase-constrained degradation sensing according to claim 7, characterized in that: The formula for channel compensation in step 5 is: in, For the compensated image, ; , which represents the phase constraint weight; The phase difference threshold, For the average phase error, in hour, .

9. The color correction method for cultured fish images based on phase-constrained degradation sensing according to claim 8, characterized in that, The formula for the average phase error in step 6 is: in, For phase spectrum, The number of pixels in the image. The imaginary unit, .

10. The color correction method for cultured fish images based on phase-constrained degradation sensing according to claim 2, characterized in that: The formula for the channel loss in step 7 is: in, This is a channel loss.

Images