An underwater image restoration method suitable for dark background under artificial light

Abstract

This invention provides a method for underwater image restoration under artificial lighting and dark backgrounds, belonging to the field of image processing technology. Based on the imaging characteristics of underwater images under dark backgrounds, the blur-based depth estimation method is improved to make it applicable to the estimation of depth d(x) in underwater images under dark backgrounds. Addressing the errors or even erroneous results of existing methods when applying them to underwater images with dark backgrounds, the method utilizes the depth difference between the foreground and background to segment the image, obtaining the water body region of the background. Within the water body region, the dark channel prior method is used to estimate the background light B of the image. λ Furthermore, based on the special imaging rules of light against a dark background, and utilizing the known depth of field d(x) and background light B... λ To solve for the underwater attenuation coefficient c λ This invention enables the restoration of underwater images. It can more accurately restore the clarity of underwater images with dark backgrounds under artificial lighting, and has greater adaptability to dark aquatic environments.

Description

Technical Field

[0001] This invention relates to the field of image processing, and in particular to a method for underwater image restoration under artificial lighting and dark backgrounds. Based on this invention, it is possible to restore underwater images acquired by unmanned, untethered underwater robots in dark underwater working environments, thereby restoring the clarity of degraded images. Background Technology

[0002] Unmanned, untethered underwater robots typically operate at depths of over 100 meters, in environments where natural light is absent and the scene is extremely dark. The artificial light used by these robots is severely absorbed by the water, and further scattered by suspended particles. Consequently, underwater images acquired in such conditions suffer significant degradation, resulting in blurry images, low contrast, color distortion, and excessive noise. To provide clear information support for vision-based underwater robots, enabling them to accurately and quickly identify targets, restoring degraded underwater images is a pressing issue that needs to be addressed.

[0003] The main idea behind underwater image restoration is to physically model the degradation principle of underwater images. By estimating and substituting the coefficients affecting attenuation, a clear underwater image can be obtained. Existing typical underwater image restoration methods improve and optimize fog dehazing methods based on dark channel priors and apply them to underwater environments. This method is called underwater dark channel prior. However, due to the lack of red channel information, the restored image exhibits color cast. Galdran et al. proposed a red channel prior method to address the problem of rapid red light attenuation in underwater environments. This method improves the dark channel prior method by using the inverse red channel, but its contrast enhancement of the restored image is not ideal. To address the shortcomings of applying dark channel priors to underwater environments, Chiang et al., assuming the normalized residual energy ratio in the underwater environment is known, proposed a method based on wavelength compensation and image dehazing, effectively enhancing the contrast of underwater images. However, these methods are all applied to relatively bright underwater images under natural or mixed lighting conditions. This differs from the working environment of underwater robots, such as underwater target identification in dark, uncharted waters. Existing underwater image restoration methods are not suitable for this environment. Therefore, it is significant to research an underwater image restoration method suitable for dark backgrounds under artificial lighting, particularly in dark and murky water environments. Summary of the Invention

[0004] The purpose of this invention is to provide an underwater image restoration method suitable for dark backgrounds under artificial lighting. This patent can solve the problem of unsatisfactory image processing results caused by inaccurate estimation of parameters such as background light and transmittance when processing images in turbid and dark underwater environments under artificial lighting, thereby improving the underwater image restoration effect.

[0005] The objective of this invention is achieved as follows:

[0006] Step (1): Input the original underwater image. For underwater images with a dark background, adopt the improved ambiguity depth estimation method based on the bright channel prior principle to obtain the depth d(x) corresponding to the underwater image scene.

[0007] Step (2): Combining the depth map d(x) obtained in step (1), according to the definition of background light, find the background water area in the depth map, and find the point with the largest dark channel value in the background water area. The intensity corresponding to this point is the background light intensity B. λ Where λ is the wavelength, λ∈{R,G,B};

[0008] Step (3): Based on the imaging law of light in a dark water environment, taking advantage of the characteristic that the scene depth of the water area is "infinite", the light emitted from the artificial light source is attenuated in the background area after passing through the "infinite" light path, and the intensity of the direct component tends to "0". Finally, only the backscattered component returns to the camera. Combining the depth of field d(x) and background light B obtained in steps (1) and (2), λ The underwater attenuation coefficient c can then be calculated. λ Then, the transmittance t at each point in the scene can be calculated. λ (x).

[0009] Step (4): Combining Step (2) and Step (3), the underwater image background light estimation result and transmittance estimation result are given. The underwater image is restored according to the underwater image restoration formula, and the final clear underwater image is output.

[0010] Furthermore, step (1) specifically includes:

[0011] This invention proposes an improved method for estimating depth of field. In an underwater image, the pixels at any point gradually approach the surrounding pixels as the image blurs. The brightness channel prior principle states that the brightness channel better represents the intensity of light received by the image, thus more accurately reflecting the distance of objects from the light source in an artificial light scene. The difference between the intensity value of each pixel in the brightness channel image and the intensity value of the corresponding pixel in the image after multi-scale Gaussian filtering is calculated; this difference represents the blur feature of each pixel.

[0012]

[0013] Where x represents the coordinates of a point in the scene, x = (x′, y′), G k,σ This indicates that the kernel size is k×k and the variance is σ. 2 A Gaussian filter, where both k and σ are 2. i n+1. The smaller the Blur(x), the smaller the difference between the intensity value of the pixel at that location and the intensity values ​​of the surrounding pixels, the more blurred that location is in the underwater image, and the greater the imaging distance between the object at that location and the camera, and vice versa. Therefore, the relative depth of each point in the scene can be represented by the magnitude of the blur value Blur(x). Finally, the relative depth map is stretched to obtain the true depth estimation map d(x).

[0014] Furthermore, step (2) specifically includes:

[0015] Background light B λ The solution is based on the pixels of the background water region. Therefore, accurately segmenting the background water region in the image is the key to determining the background light B. λ The key to solving this problem lies in the severe backscattering caused by artificial light sources in underwater images. Besides the foreground objects being brighter than the background, suspended impurities in the water, located at distances less than the foreground depth, also appear as bright white dots under artificial light. To segment the background water region and eliminate the influence of impurities on subsequent parameter calculations, the Ostu thresholding method is employed. This method leverages the proximity of the foreground and impurities to the camera, resulting in a shallow scene depth, while the background water scene depth approaches infinity. This leads to the segmentation of the background water region from the foreground and impurities, yielding the background water region A.

[0016] In the segmented background water region A, the dark channel prior method is used to estimate the background light B. λ Select the brightest pixels in the top 0.1% of the dark channel values ​​in the background water area A, and map these pixels back to the original image. These areas represent the regions with the highest fog density. Then, select the value of the pixel with the highest value as the background light estimate B. λ .

[0017] Furthermore, step (3) specifically includes:

[0018] This invention proposes a method for measuring the transmittance t of underwater images against a dark background. λ (x) Solution method. When light propagates in water, its intensity decreases exponentially with increasing distance d(x). Generally, the transmittance t is used as the solution. λ Transmittance (x) reflects the degree of light intensity attenuation and is an important parameter for compensating for the energy attenuation of light at different wavelengths in underwater image restoration. The expression for transmittance is: In the formula cλ Let d(x) be the total attenuation coefficient, and d(x) be the distance between a point x on the target and the underwater camera. Given that the scene depth map d(x) has been estimated, in order to obtain an accurate transmittance map t for an underwater image... λ (x), the underwater attenuation coefficient c must be required. λ .

[0019] For underwater images acquired under artificial lighting with a dark background, the underwater attenuation coefficient c is calculated using the backscattering intensity of the background water region A, based on the unique imaging characteristics of the dark underwater environment. λ .

[0020] It is known that the main ways light attenuates in water are absorption and scattering. Scattering is divided into forward scattering and backscattering. Forward scattering occurs in a relatively small proportion and is therefore ignored in the calculation of physical models for underwater imaging. Backscattering, on the other hand, refers to the change in the light path caused by reflection from the water medium and impurities in the water, and is the main reason for the "fog" effect in scenes.

[0021] Therefore, in images captured in a dark underwater environment, the depth of field of the background water area approaches "infinity," and light emitted from an artificial light source enters the background area A. Part of this light undergoes backscattering, entering the camera through the water medium and impurities, creating a "fog" effect in the image. The remaining unscattered light travels along the "infinity" light path and is completely absorbed, thus no light returns to the camera. Ultimately, the background area appears as a white fog shrouding a black background. Its physical imaging model is as follows:

[0022]

[0023] The imaging model after the light attenuates in the background water region A through an "infinite" optical path can be simplified as follows:

[0024]

[0025] Therefore, using the already solved background light B λ Given the depth of field d(x), the underwater attenuation coefficient can be solved by the following formula:

[0026]

[0027] Therefore, by combining the already estimated depth map d(x), the transmittance map t of the underwater image can be solved. λ (x):

[0028]

[0029] Compared with existing technologies, the advantages of this invention are as follows: Current underwater image restoration methods are mainly designed for well-lit, bright underwater environments. Research on underwater image restoration under artificial lighting and dark backgrounds is scarce. Due to the influence of artificial lighting, scene points closer to the camera and light source receive higher intensity illumination, meaning the foreground brightness is higher than the background brightness, which is the opposite of the imaging pattern in bright underwater scenes. Furthermore, most existing literature relies on assumptions such as "the normalized residual energy ratio is known" and "the attenuation factor can be empirically obtained" when processing underwater images. These assumptions are unsuitable for the complex and variable working environment of underwater robots and cannot be met. This invention, however, addresses practical needs and the working environment of most deep-sea underwater robots. First, it improves the ambiguity-based depth estimation method to make it applicable to underwater images with dark backgrounds under artificial lighting. Second, it proposes a background light B-type method based on the imaging patterns in dark underwater environments. λ and underwater attenuation coefficient c λ The solution method utilizes scene depth to segment the water region in the background of the image, and estimates the background light B of the image in the dark background water region. λ and underwater attenuation coefficient c λ The underwater attenuation coefficient c is then calculated using the method described above. λ And the depth of field d(x), that is, the transmittance t of the corresponding channel of the underwater image. λ (x). This invention can more accurately restore the clarity of images in dark water environments under artificial lighting, and the method has greater adaptability to water bodies against dark backgrounds. Attached Figure Description

[0030] Figure 1 This is an underwater imaging model under artificial lighting in this invention patent.

[0031] Figure 2 Flowchart of the improved ambiguity-based depth estimation method

[0032] Figure 3 These are underwater degradation images used in the experiments of this invention patent.

[0033] Figure 4 (a)-(f) are comparison diagrams of the restoration effects of the present invention and existing underwater image restoration algorithms in this patent.

[0034] Figure 5 This is the overall flowchart of the present invention. Detailed Implementation

[0035] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0036] Combination Figures 1 to 5 The specific implementation steps of the underwater image restoration method based on multiple frames under artificial lighting according to the present invention are as follows:

[0037] Step (1): Input the original underwater image. For underwater images with a dark background, adopt the improved ambiguity depth estimation method based on the bright channel prior principle to obtain the depth d(x) corresponding to the underwater image scene.

[0038] Depth of field, or scene depth, represents the distance between a point x in the scene and the camera. In underwater image restoration methods based on model parameter estimation, scene depth is crucial to the calculation of transmittance in the model. Furthermore, in dark and turbid underwater environments, the attenuation and scattering of light are far more severe than under natural lighting conditions. Light of all wavelengths attenuates to almost the same degree in water, color shift no longer occurs, and severe backscattering attempts to create a dense fog effect.

[0039] Therefore, in order to improve the accuracy of depth estimation in underwater images with dark backgrounds, this patent improves the ambiguity-based depth estimation method, making it applicable to underwater images with turbid and dark backgrounds under artificial lighting.

[0040] First, the basic idea of ​​the brightness channel prior is that in most blurred images, some pixels will always have at least one color channel with a relatively high intensity. For any image I, its brightness channel expression is:

[0041]

[0042] Where x represents the coordinates of a point in the scene, x = (x', y'); λ (x) represents a specific R, G, B color channel of a point in a color image I, and Ω represents a neighborhood window centered on pixel x. The brightness channel prior principle states that the brightness channel better represents the intensity of light received by an image. This allows for a more accurate reflection of the distance between objects and the light source in a scene under artificial lighting conditions. Therefore, this patent utilizes brightness channel images to improve the blur-based depth estimation method, eliminating scene depth estimation errors caused by the color of objects themselves.

[0043] Secondly, based on the obtained brightness channel image, a depth-of-field estimation method based on blurriness is used to estimate the depth of field of underwater images with dark backgrounds under artificial lighting. Generally, the blurriness of objects in a scene is affected by two factors: underwater turbidity and the imaging distance between the object and the camera. Although water turbidity is the main cause of underwater object blurriness, for the same underwater scene, the imaging distance between the object and the camera is the direct cause of the blurriness. Generally, the greater the imaging distance between the target object and the camera, the more likely blurriness will occur, and vice versa.

[0044] Utilizing the principle that pixels at any point in an underwater image gradually approach each other as the image blurs, we first extract the bright channel image corresponding to the original image. Then, we calculate the difference between the intensity value of each pixel in the bright channel image and the intensity value of the corresponding pixel in the image after applying multi-scale Gaussian filtering. This difference represents the blur feature for each pixel.

[0045]

[0046] Where x represents the coordinates of a point in the scene, x = (x′, y′); G k,σ The kernel size is k×k; the variance is σ. 2 Gaussian filter; k and σ are both 2 i n+1. The smaller Blur(x) is, the smaller the difference between the intensity value of the pixel at this position and the intensity value of the surrounding pixels. The more blurred the corresponding position is in the underwater image, the greater the imaging distance between the object at this position and the camera.

[0047] Therefore, the blurriness of a pixel can represent the relative scene depth of that pixel, but not the actual scene depth. To convert the relative distance between points in an underwater image into an absolute distance, it is necessary to first determine the actual distance of the nearest point in the image. In actual underwater robot operations, the distance d0 between the underwater robot and the target object can be measured by its own equipped Doppler log (DVL).

[0048] Therefore, the actual depth of field is:

[0049] d(x)=D×Blur(x)+d0

[0050] In the formula: D is the conversion coefficient, which, through extensive experimental verification, is 8 in this invention patent. A detailed flowchart for estimating scene depth is attached. Figure 2 As shown.

[0051] Step (2): Combining the depth map d(x) obtained in step (1), according to the definition of background light, find the background water area A in the depth map, and find the point with the largest dark channel value in the background water area. The intensity corresponding to this point is the background light intensity B. λ Where λ is the wavelength, λ∈{R,G,B}.

[0052] Background light refers to the pixel value corresponding to the brightest point at infinity in a non-foreground region of an image. Therefore, for an underwater image, the background light value is a constant. In dark and murky underwater environments, the effect of artificial light sources will cause the brightness of pixels in the foreground closer to the camera and light source, as well as impurities suspended in the water, to be higher than the brightness of pixels in the background region. Therefore, classic methods are very prone to misclassifying bright spots on impurities or objects as pixels that can represent the background light value.

[0053] Therefore, based on the above analysis, in order to accurately solve the background light value in the image scene, we first consider segmenting the background water area, and then use the dark channel prior method to estimate the background light in the background water area.

[0054] This patent considers using image segmentation technology to separate the background water from the image. However, traditional grayscale segmentation methods suffer from blurred edges due to water turbidity, numerous suspended impurities, and severe underwater scattering. Furthermore, objects of different colors exhibit significant differences in grayscale values ​​after conversion, making it impossible to completely segment the water using grayscale thresholding. To address this issue, this invention proposes an underwater dark background image segmentation method based on scene depth. This method enables complete segmentation of water regions against a dark background, eliminating interference from foreground colors and bright spots formed by suspended impurities in the water on background light estimation.

[0055] First, the underwater image with a dark background is captured simulating the process of an underwater robot hovering and moving forward in deep water. Therefore, the background area is a dark water environment with "infinity," while the foreground consists of objects in the water within the range of artificial light. Taking advantage of the fact that the foreground and impurities are close to the camera and have a shallow scene depth, while the background water scene depth tends towards "infinity," the Ostu thresholding method is used to segment the background water area from the foreground and impurities, preparing for the next step of background light estimation.

[0056] Then, the background light is estimated using the dark channel prior method in the segmented background water region. The brightest pixels in the top 0.1% of the dark channel values ​​in background water region A are selected, and these pixels are mapped to the original image. These regions represent the areas with the highest fog density. The value of the pixel with the highest dark channel value is then selected as the estimated background light value B. λ .

[0057] Step (3): Based on the imaging law of light in a dark water environment, taking into account that the scene depth of the water area is "infinity", the artificial light is attenuated after passing through the "infinity" light path after being emitted from the light source, and the ideal intensity of the background water area is "0", the depth of field d(x) and the background light B obtained in steps (1) and (2) are combined. λ The underwater attenuation coefficient c can then be calculated. λ Then, the transmittance t at each point in the scene is calculated by combining the depth of field d(x). λ (x).

[0058] When light travels through water, its intensity decreases exponentially with increasing distance *d*, and this is generally expressed by the transmittance *t*. λ (x) reflects the degree of attenuation and is an important parameter for compensating for the energy attenuation of light at different wavelengths in underwater image restoration. Transmittance t λ The expression for (x) is:

[0059]

[0060] In the formula: x represents the coordinates of a point in the scene, x = (x′, y′); d(x) represents the distance between a point x in the scene and the underwater camera, i.e., the depth of field. λ This represents the total attenuation coefficient of light with wavelength λ.

[0061] Therefore, in order to obtain an accurate transmittance t of an underwater image λ Given that the depth of field has been estimated, a set of estimation methods for the underwater attenuation coefficient c needs to be designed. λ method.

[0062] Total attenuation coefficient c λ Besides being affected by the wavelength of light, it is also influenced by seawater salinity and phytoplankton concentration. Currently, most studies provide an attenuation coefficient c based on the type of marine environment. λ Based on empirical values, and considering the variable underwater environment in which underwater robots operate, it is impossible to determine a known underwater attenuation coefficient for image restoration. Therefore, this invention proposes an attenuation coefficient c for underwater images with a dark background. λ The solution method.

[0063] In underwater environments, the main ways light attenuates are absorption and scattering. Scattering is divided into forward scattering and backscattering. Forward scattering occurs in a relatively small proportion and is therefore ignored in the calculation of physical models for underwater imaging. Backscattering, on the other hand, refers to the change in the optical path caused by light being reflected by the water medium and impurities in the water, and is the main reason for the "fog" effect in the scene.

[0064] Images captured in a dark underwater environment have a depth of field approaching infinity in the background water area. Light emitted from an artificial light source enters the background area. Some of this light is backscattered, passing through the water and impurities before reaching the camera, creating a "fog" effect. The remaining unscattered light travels along the "infinite" optical path and is almost completely absorbed, resulting in no light returning to the camera and making the direct component in the physical imaging model approach zero. Ultimately, the background area appears as a black background shrouded in a layer of white fog.

[0065] In summary, the physical principle imaging model for the background water area is as follows:

[0066]

[0067] Therefore, the imaging model of the background water area can be simplified as follows:

[0068]

[0069] Using the already estimated background light B in the background water area λ With depth of field d(x), underwater attenuation coefficient c λ It can then be solved by the following formula:

[0070]

[0071] Therefore, the attenuation coefficients c of the three channels R, G, and B are... r c g c b The solution can then be obtained using the above formula. Combined with the already estimated depth map d(x), the transmittance map t of the underwater image can then be calculated. λ (x).

[0072] Step (4): Combining Step (2) and Step (3), the underwater image background light estimation result and transmittance estimation result are given. The underwater image is restored according to the underwater image restoration formula, and the final clear underwater image is output.

[0073] Combination Figure 1 According to the Jaffe-McGlamery underwater optical model, the underwater imaging model under artificial illumination is as follows:

[0074] I λ (x)=J λ (x)t λ (x)+B λ (1-t λ (x))

[0075] In the formula: I λ (x) is a blurred underwater image captured by an underwater camera; J λ(x) represents the ideal, clear image obtained after restoration; t λ (x) represents the transmittance of the underwater image.

[0076] Using the background light B obtained in steps (2) and (3) λ and transmittance t λ (x), for the input original image I λ (x), then output the restored clear underwater image J. λ (x):

[0077]

[0078] (VI) Application Cases

[0079] To verify the effectiveness of the underwater image restoration method for dark backgrounds under artificial lighting proposed in this invention, the algorithm proposed in this invention was experimentally compared with the UDCP algorithm by Drews et al., the red channel prior algorithm by Galdran et al., the MIP algorithm by Bianco et al., the IBLA algorithm by Peng et al., and the WCID algorithm by Chiang et al. The original degraded underwater images used in the experiments are attached. Figure 3 As shown in the attached figure, the underwater image restoration effect comparison is as follows. Figure 4 As shown. Wherein:

[0080] Figure 4 (a) Using the algorithm proposed in this patent to... Figure 3 The result of the processing;

[0081] Figure 4 (b) Correspondingly, the UDCP algorithm proposed by Drews et al. is used for... Figure 3 The result of the processing;

[0082] Figure 4 (c) Correspondingly using the red channel prior algorithm proposed by Galdran et al. Figure 3 The result of the processing;

[0083] Figure 4 (d) Corresponding to the MIP algorithm proposed by Bianco et al. Figure 3 The result of the processing;

[0084] Figure 4 (e) Correspondingly, the algorithm proposed by Peng et al. is used to... Figure 3 The result of the processing;

[0085] Figure 4 (f) Corresponding to the WCID algorithm proposed by Chiang et al. Figure 3 The result of the processing.

[0086] Attachment Figure 3 The experimental results show that the UDCP algorithm by Drews et al. resulted in a reddish color cast in the image, caused by incorrect estimation of transmittance in each channel. Simultaneously, foreground objects exhibited distortion and overcompensation, and the halo effect on the outer ring exacerbated the image quality. The red channel prior algorithm by Galdran et al. resulted in an overall reddish tint due to incorrect estimates of transmittance and background light. The yellow and red spheres in the foreground also showed reddish color casts and were darkened, while the blue sphere was almost indistinguishable from the background. The MIP algorithm by Bianco et al. resulted in severe color distortion. The IBLA algorithm by Peng et al., due to incorrect estimation of background light, incorrectly enhanced the darker background of the original underwater image, turning it blue, and darkened the overall color of the processed image. The WCID algorithm by Chiang et al. resulted in numerous white spots in the background area, covering the target area and causing a loss of target detail. In contrast, the algorithm proposed in this invention effectively improved image clarity, made target details more apparent, and resulted in a more natural color in the restored image compared to other algorithms.

[0087] To more objectively evaluate the image quality of the algorithm's experimental results, this invention patent selects two commonly used underwater image quality evaluation metrics, UCIQE (Underwater Color Image Quality Evaluation) and UIQM (Underwater Image Quality Measurement), as the quality evaluation metrics for underwater image restoration. UCIQE reflects the linear quantitative evaluation result of the relationship between color cast, blur, and contrast after underwater image restoration; a higher value indicates a better image processing effect. UIQM mainly uses a linear combination of the color measurement metric (UICM), sharpness measurement metric (UISM), and contrast measurement metric (UIConM) as the evaluation basis; a higher value indicates better color balance, sharpness, and contrast in the processed image.

[0088] The quantitative analysis results of image restoration by different algorithms are shown in Table 1. By comparing and analyzing the evaluation indicators of each algorithm, it can be seen that the algorithm proposed in this invention is superior to other underwater image restoration algorithms in both the objective evaluation indicators of UCIQE and UIQM. Therefore, for turbid water environments with artificial light sources, the algorithm proposed in this invention can better restore underwater images with dark backgrounds under artificial lighting compared to other underwater image restoration algorithms.

[0089] Table 1 Comparison of objective evaluation metrics among different algorithms

[0090]

[0091] In summary, this invention provides a method for underwater image restoration under artificial lighting and dark backgrounds. It belongs to the field of image processing technology. This method improves upon blur-based depth estimation methods based on the imaging characteristics of underwater images under dark backgrounds, making it suitable for estimating the depth d(x) of underwater images under dark backgrounds. Furthermore, addressing the errors or even erroneous results of existing methods when applying background light estimation to underwater images under dark backgrounds, it utilizes the depth difference between the foreground and background to segment the image. After obtaining the water body region of the background, the dark channel prior method is used to estimate the background light B of the image within the water body region. λ Finally, based on the special imaging rules of light against a dark background, and using the known depth of field d(x) and background light B... λ To solve for the underwater attenuation coefficient c λ This invention enables the restoration of underwater images. It can more accurately restore the clarity of underwater images with dark backgrounds under artificial lighting, and the method is more adaptable to dark aquatic environments.

Claims

1. A method for underwater image restoration under dark backgrounds under artificial lighting, characterized in that, The steps are as follows: Step 1: Input the original underwater image. For underwater images against a dark background, extract the bright channel of the image. Then, use a blur-based depth estimation method on the bright channel image to obtain the depth of field corresponding to the underwater scene. ; Step 2: Combine with depth map Based on the definition of background light, locate the background water area in the depth map. Find the point with the highest dark channel value within the background water area; its corresponding intensity is the background light intensity. ;in, Wavelength; Step 3: After light is emitted from the artificial light source, it attenuates along an "infinite" optical path in the background area, causing the direct component intensity to approach "0," with only the backscattered component returning to the camera. This, combined with the depth of field... and background light The attenuation coefficient under water outlet was obtained. This allows us to obtain the transmittance at various points in the scene. ; in, ; Step 4: Estimating background light by combining underwater images Results and transmittance The underwater image is reconstructed based on the underwater image imaging formula, and the final clear underwater image is output.

2. The underwater image restoration method suitable for dark backgrounds under artificial lighting according to claim 1, characterized in that, Step one specifically includes: representing the blur feature of each pixel by the difference between the intensity value of each pixel in the bright channel image and the intensity value of the corresponding pixel in the image after multi-scale Gaussian filtering. in, Represents the coordinates of a point in the scene. , Indicates the kernel size as The variance is Gaussian filter, and All for ; Utilizing the fuzziness values ​​of each scene point The size represents the relative depth of each point in the scene; stretching the relative depth map yields the true depth estimation map. .

3. The underwater image restoration method suitable for dark backgrounds under artificial lighting according to claim 1, characterized in that, Step two specifically includes: selecting the background water area. The brightest pixels in the top 0.1% of the dark channel values ​​are selected, and these pixels are mapped to the original image. These areas represent the regions with the highest fog density. The value of the pixel with the highest value is selected as the background light estimate. .

4. The underwater image restoration method suitable for dark backgrounds under artificial lighting according to claim 1, characterized in that, Step three specifically includes: transmittance is: In the formula: The total attenuation coefficient is... For a point on the target object The distance between the location and the underwater camera.

Images