Enhanced visualization method for removing turbidity of turbid underwater scene

Abstract

The invention discloses an enhanced visualization method for removing turbidity of a turbid underwater scene, which comprises the following steps of: firstly, acquiring a water body turbidity factor as a quantization basis for an environmental turbidity degree, then, executing a turbidity adaptive contrast limited histogram equalization TCLAHE algorithm on an original turbid image, and adjusting the enhancement amplitude and shear limitation of each sub-block according to the turbidity factor, therefore, noise amplification is effectively inhibited under the condition of high turbidity, and the local contrast ratio is fully improved under the condition of low turbidity. And then highlighting important targets and texture information in the image by adopting a saliency detail enhancement means. And finally, multi-scale fusion is performed on the enhanced image and other pre-processing information, and a turbidity factor is considered in fusion weight calculation, so that the turbidity removal degree and the visual reality sense can be obtained at the same time. Compared with an existing underwater defogging and enhancing technology, the underwater defogging and enhancing method has innovativeness, can obtain a stable and clear enhancing effect in various turbid scenes, and is widely applied to the fields of ocean observation, underwater robot vision, underwater cultural relic shooting and the like.

Description

Technical Field

[0001] The present invention relates to a method for enhancing visibility by removing turbidity from turbid underwater scenes, and belongs to the technical field of underwater image processing using computer vision. Background Art

[0002] With the growing demand for ocean exploration and river monitoring, the quality of underwater images is particularly important for research and application. Clear underwater images are the basis for observing the behavior of marine organisms, identifying species characteristics and seabed topography. Low-quality images may lead to errors in biological classification or misjudgment of geological structures. However, the quality of underwater images is easily affected by the characteristics of the water medium. In turbid waters, due to the scattering and absorption of light by suspended particles, images of underwater scenes often appear turbid like "haze", with low contrast, dull blue colors, blurred details, and a significant decrease in target recognizability. This degradation not only affects direct observation by the human eye, but also reduces the performance of computer vision algorithms (such as underwater target detection and robot visual navigation).

[0003] To improve the visual quality of turbid underwater images, industry and academia have proposed a variety of image dehazing and enhancement techniques. For example, physical-model-based dehazing methods utilize underwater imaging models and prior knowledge (such as dark channel priors and light propagation models) to estimate scene depth and medium transmittance, thereby restoring a clear image. However, such methods often require accurate estimation of multiple parameters (such as the water attenuation coefficient and background light). When water turbidity varies or field calibration is difficult, model calculations may be inaccurate, resulting in biased restoration results. Image enhancement methods directly improve image appearance through techniques such as histogram equalization, white balancing, and sharpening filters. For example, contrast-limited adaptive histogram equalization (CLAHE) is effective in enhancing local contrast and is widely used for underwater image enhancement. However, traditional CLAHE algorithms do not account for differences in ambient turbidity and may over-enhance noise or fail to fully enhance detail in highly turbid scenes with significant stray light. Fusion of the original image with a color-corrected and contrast-enhanced version achieves a balance between natural color and clarity. However, existing fusion algorithms usually use a fixed weight calculation method and lack the perception of water turbidity, so they cannot guarantee the best results under different turbidity conditions.

[0004] Therefore, there is an urgent need for a deturbidity enhancement method for turbid underwater scenes, which can adaptively adjust the deturbidity and enhancement strategies according to the turbidity conditions of the water body, fully remove the turbidity effect while avoiding over-processing. Summary of the Invention

[0005] The content of this application is used to briefly introduce concepts that will be described in detail in the detailed description section below. The content of this application is not intended to identify key features or essential features of the technical solution for which protection is sought, nor is it intended to limit the scope of the technical solution for which protection is sought.

[0006] In response to the problems and shortcomings of the prior art, the present invention aims to provide a method for enhancing visualization by de-turbidity in turbid underwater scenes. The method uses a turbidity factor, obtained through a turbidimeter or visual estimation based on image analysis, to process the original underwater image using the improved standard CLAHE algorithm in combination with the turbidity factor to produce a first intermediate image. The first intermediate image is then filtered for subtle details such as edges and textures to produce a second intermediate image. Finally, the second intermediate image is fused with the original underwater image to reconstruct and output an enhanced underwater image. This method achieves greater image gain in high turbidity conditions while retaining more original information in low turbidity conditions, resulting in a natural and clear final result. This method addresses the problems identified in the aforementioned background art.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] The present invention discloses a method for enhancing visibility by removing turbidity from turbid underwater scenes, comprising the following steps:

[0009] Step 1, in response to acquiring a turbid original underwater image;

[0010] Step 2, obtaining a turbidity factor using a turbidity meter or by image visual feature estimation based on the original underwater image;

[0011] Step 3, performing turbidity adaptive contrast limited histogram equalization processing on the original underwater image according to the turbidity factor to obtain a first intermediate image;

[0012] Step 4: extract high-frequency components and detect significant areas on the first intermediate image, and perform gain amplification on significant edges and textures based on the turbidity factor to obtain a second intermediate image;

[0013] Step 5: Decompose the original underwater image and the second intermediate image at multiple scales, assign fusion weights at each scale based on the turbidity factor and image features, and then fuse them accordingly.

[0014] Step 6: reconstruct and output the fused results to obtain the final enhanced underwater image.

[0015] Preferably, the calculation of the turbidity factor in step 2 further comprises the following steps:

[0016] Step 2.1, performing scaling on the acquired original underwater image;

[0017] Step 2.2, extract the image average gradient amplitude, image brightness standard deviation and image pixel saturation average in grayscale domain and HSV color space simultaneously;

[0018] Step 2.3, calculating the turbidity score based on the image average gradient amplitude, the standard deviation of the image brightness, and the average value of the image pixel saturation;

[0019] Step 2.4: normalize the turbidity score to obtain a turbidity factor corresponding to the actual physical turbidity.

[0020] Preferably, the step 3 further comprises the following steps:

[0021] Step 3.1, dividing the original underwater image into a plurality of grid sub-blocks;

[0022] Step 3.2: Use the clean water sample to calibrate the baseline clipping threshold clip0 and the baseline enhancement gain coefficient γ0;

[0023] Step 3.3, obtaining a modified clipping threshold clip and enhancement gain coefficient γ according to a modulation function monotonically related to the turbidity factor;

[0024] Step 3.4: redistribute the pixel histogram of the grid sub-block based on the modified clipping threshold clip and the enhancement gain coefficient γ.

[0025] Preferably, the step 4 further comprises the following steps:

[0026] Step 4.1, extracting high-frequency details from the first intermediate image;

[0027] Step 4.2, using a Laplace or Sobel operator to obtain an edge map of the first intermediate image;

[0028] Step 4.3, extracting fine texture structures of the first intermediate image through high-pass filtering;

[0029] Step 4.4: Introduce visual saliency detection or edge amplitude to filter out weak textures and edges in edge maps and fine texture structures;

[0030] Step 4.5: Apply detail gain to the filtered edge image and fine texture structure and superimpose them on the first intermediate image to obtain a second intermediate image.

[0031] Preferably, the step 5 further comprises the following steps:

[0032] Step 5.1, performing multi-scale decomposition on the original underwater image and the second intermediate image respectively to obtain sub-band representations from low frequency to high frequency;

[0033] Step 5.2, set the fusion weight factor based on the turbidity factor;

[0034] Step 5.3, fuse the corresponding sub-bands of the original underwater image and the second intermediate image at each scale according to the fusion weight factor.

[0035] Preferably, in Step 3.3, substitute the turbidity factor T into the monotonically related increasing function g(T) and decreasing function f(T), and correct the clipping threshold clip and enhancement gain coefficient γ based on clip = clip0·f(T) and γ = γ0·g(T);

[0036] When the turbidity factor T < 0.3, the clipping threshold clip and the enhancement gain coefficient γ are close to the reference value;

[0037] When the turbidity factor T is between 0.3 and 0.7, the clipping threshold clip and the enhancement gain coefficient γ linearly contract and expand;

[0038] When the turbidity factor T ≥ 0.7, the clipping threshold clip is close to 30% of the reference value, and the enhancement gain coefficient γ is increased to be close to 1.6 times.

[0039] Preferably, in Step 4.5, control the detail gain coefficient β according to the turbidity factor T. The detail gain coefficient β is limited within the interval [0.4, 1.2], expressed as β(T) = β_min + (β_max - β_min)·(1 - T), where β_max = 1.2 and β_min = 0.4;

[0040] When the turbidity factor T is close to 0, the detail gain coefficient β is close to 1.2, which can fully highlight the texture;

[0041] When the turbidity factor T is close to 1, the detail gain coefficient β drops to be close to 0.4 to suppress the pseudo-edges generated by suspended particles.

[0042] Preferably, in Step 5.2, determine the fusion weight factor α according to the turbidity factor T, and use linear modulation to express as α(T) = α_min + (α_max - α_min)·T, where α_min = 0.3 and α_max = 1.0;

[0043] When the turbidity factor T ≤ 0.3, the fusion weight factor α is close to 0.3, and only a small amount of enhanced components are taken to maintain the authenticity of the original image;

[0044] When the turbidity factor T is 0.2 < T < 0.8, the fusion weight factor α continuously changes between 0.3 and 0.8 with the turbidity factor T, realizing gradual enhancement;

[0045] When the turbidity factor T≥0.7, the fusion weight factor α≥0.79, and a large number of detail enhancement images are used to suppress the turbidity residue.

[0046] As a second aspect of the present application, the present invention further discloses an enhanced visualization system for deturbidity of turbid underwater scenes, comprising:

[0047] Turbidity analysis module, used to obtain the original underwater image, and obtain the turbidity factor using a turbidity meter or by estimating the image visual features;

[0048] a turbidity adaptive contrast enhancement module, configured to perform turbidity adaptive contrast limited histogram equalization processing on the original underwater image according to the turbidity factor to obtain a first intermediate image;

[0049] a salient detail enhancement module, configured to extract high-frequency components and detect salient regions from the first intermediate image, and perform gain amplification on salient edges and textures based on the turbidity factor to obtain a second intermediate image;

[0050] The multi-scale fusion module is used to perform multi-scale decomposition on the original underwater image and the second intermediate image, and to reconstruct the corresponding fused output after allocating fusion weights at each scale according to the turbidity factor and image features.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] The present invention provides a method for enhancing visualization by deturbidity in turbid underwater scenes. The improved turbidity adaptive contrast-limited histogram equalization algorithm TCLAHE adjusts the amplitude and limiting threshold of the histogram equalization in real time according to the turbidity of the water body, overcoming the problem of unbalanced effects of traditional fixed enhancement processing in different turbidity scenes, while ensuring the improvement of contrast and avoiding the introduction of excessive noise. Combined with significant detail enhancement and multi-scale fusion, the two-stage processing not only highlights the details of important targets but also maintains the global naturalness of the image. The fusion strategy dynamically configures weights based on turbidity perception, avoiding the drawbacks of simple enhancement methods such as oversharpening details or color distortion. The present invention also incorporates environmental turbidity information throughout the image enhancement process, and in particular couples the image deturbidity level with the fusion weight in the fusion stage, which is an innovation that distinguishes it from existing fusion algorithms. This mechanism ensures that the output image can achieve the best balance between clarity and authenticity regardless of whether the water is clear or turbid. Due to the introduction of the turbidity factor, the method of the present invention can self-adjust parameters according to the on-site water quality conditions. Therefore, it is applicable to different types of waters (clear seawater, turbid rivers, muddy ponds, etc.) and has good robustness and practical value. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The drawings constituting a part of this application are used to provide a further understanding of this application and make other features, purposes and advantages of this application more apparent. The drawings and descriptions of the exemplary embodiments of this application are used to explain this application and do not constitute an improper limitation on this application.

[0054] In the attached figure:

[0055] Figure 1 A connection diagram of the main steps of the method for enhancing visual dehazing in an embodiment of the present invention;

[0056] Figure 2 FIG. 1 is an overall flow chart of the method for enhancing visual dehazing according to an embodiment of the present invention;

[0057] Figure 3 (a)-(d) are comparison diagrams of the effects of NTU13-16 before and after enhancement under medium turbidity conditions in the embodiment of the present invention;

[0058] Figure 4 In the embodiment of the present invention, (e)-(g) are comparison diagrams of the effects before and after enhancement under medium turbidity conditions of NTU17-19, and (h) is a comparison diagram of the effects before and after enhancement under extremely high turbidity conditions of NTU40;

[0059] Figure 5 In the embodiment of the present invention, (a) is the input original underwater image, (b)-(k) are comparison diagrams of the visual effects before and after enhancement using different methods under different turbidity conditions, and (l) is the output effect diagram of the method of the present invention. DETAILED DESCRIPTION

[0060] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments described herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are for illustrative purposes only and are not intended to limit the scope of protection of the present disclosure.

[0061] It should also be noted that, for ease of description, only the parts related to the invention are shown in the drawings. In the absence of conflict, the embodiments and features in the embodiments of the present disclosure may be combined with each other.

[0062] The present invention discloses a method for enhancing visibility by removing turbidity from turbid underwater scenes. The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. Figure 1 and Figure 2 As shown, it mainly includes the following steps:

[0063] Step 1, in response to acquiring a turbid original underwater image;

[0064] Step 2: Obtain a turbidity factor using a turbidity meter or by estimating the visual features of the original underwater image;

[0065] Step 3, performing turbidity adaptive contrast limited histogram equalization processing on the original underwater image according to the turbidity factor to obtain a first intermediate image;

[0066] Step 4: extract high-frequency components and detect salient regions from the first intermediate image, and perform gain amplification on salient edges and textures based on a turbidity factor to obtain a second intermediate image.

[0067] Step 5: Select the original underwater image and the second intermediate image for multi-scale decomposition, and assign fusion weights at each scale based on the turbidity factor and image features, and then fuse them accordingly;

[0068] Step 6: reconstruct and output the fused results to obtain the final enhanced underwater image.

[0069] The present invention discloses a method for obtaining a turbid original underwater image based on the acquired water turbidity information, and then clarifying the subsequent color correction steps through the information. First, the turbidity factor T is obtained by turbidity measurement or image analysis of the environmental water body. In this embodiment, the turbidity factor can be calculated by a turbidity meter and an image visual estimation method. When conditions permit, a turbidity meter can be used to directly measure the physical turbidity of the water body, obtain a turbidity value expressed in NTU (Nitride Unit), and normalize it to a turbidity factor T that can be used by the algorithm. However, the cost of using a turbidity meter is relatively high, mainly due to its high-precision sensor and real-time monitoring function. Therefore, in the absence of a dedicated turbidity sensor, we can use the image visual estimation-based method disclosed in the present invention to calculate an equivalent turbidity factor T.

[0070] The image visual estimation method of the present invention determines water turbidity by analyzing the clarity and color distribution of the original image. Statistical features such as edge gradient intensity, brightness standard deviation, and color saturation are calculated, and these features are combined to construct a score reflecting the turbidity level. The method also includes the following steps:

[0071] Step 2.1, scaling the acquired original underwater image;

[0072] Step 2.2, extract the image average gradient amplitude, image brightness standard deviation and image pixel saturation average in grayscale domain and HSV color space simultaneously;

[0073] Step 2.3, calculate the turbidity score based on the image average gradient amplitude, the standard deviation of image brightness, and the average image pixel saturation;

[0074] In step 2.4, the turbidity score is normalized to obtain the turbidity factor corresponding to the actual physical turbidity.

[0075] Specifically, the original underwater image obtained in step 1 is scaled to a uniform size and then three statistical features are extracted in the grayscale domain and HSV color space, namely, the image average gradient amplitude, the standard deviation of the image brightness, and the average image pixel saturation. Then the turbidity score is expressed as,

[0076]

[0077] Among them, G represents the average gradient amplitude of the image, which reflects the clarity of the image edge. L It represents the standard deviation of image brightness, which reflects the contrast of the image. The image's average pixel saturation value reflects the vividness of the image's colors. a, b, and c are weight coefficients, and ∈ is a minimum constant to avoid division by zero. The image's average gradient magnitude G is obtained by applying the Sobel operator to the grayscale image to obtain the horizontal and vertical gradients, and then taking the average of the gradient magnitudes of each pixel. The image's brightness standard deviation σ is L The global standard deviation of the brightness channel in Lab or HSV space is taken, the RGB image is converted to Lab or HSV space, the channel data is extracted and the global standard deviation is calculated. Take the average value of the S channel in the HSV space, convert the RGB image to the HSV color space, extract the S channel data to calculate the matrix and find the average value. Channel S represents the color purity.

[0078] Turbidity score T v The maximum-minimum range is normalized to the turbidity factor T corresponding to the actual physical turbidity, and the turbidity score T is v The linear transformation is expressed as [0, 1]. v Convert it into a turbidity factor in the interval [0, 1], eliminate the dimension difference, and make its value correspond to the turbidity factor T in terms of dimension. When the turbidity of the water body is higher, the image tends to have lower gradient and brightness contrast and weaker color saturation, so the turbidity score T v will increase, whereas in clear water the turbidity score is T v The turbidity factor T obtained in this way can be understood as a measure of the degree of image degradation caused by the water environment and can be used as a reference in subsequent steps.

[0079] The method of the present invention is based on the improvement of the contrast limited histogram equalization algorithm CLAHE, and proposes a turbidity-adaptive contrast limited histogram equalization algorithm (Turbidity-adaptive CLAHE) to perform local contrast enhancement processing on the original underwater image. The CLAHE histogram equalization method is to divide the image into several small grid blocks, perform histogram equalization on each sub-block separately, and at the same time limit the maximum threshold of its histogram pixel frequency to avoid excessive enhancement of noise. The improvement of the turbidity adaptive contrast limited histogram equalization algorithm TCLAHE of the present invention compared to the standard CLAHE algorithm is that the histogram stretching amplitude and contrast limitation parameters of each sub-block are dynamically adjusted according to the global turbidity factor T. Specifically, it also includes the following steps:

[0080] Step 3.1, divide the original underwater image into several grid sub-blocks;

[0081] Step 3.2: Use the clean water sample to calibrate the baseline clipping threshold clip0 and the baseline enhancement gain coefficient γ0;

[0082] Step 3.3, obtain the modified clipping threshold clip and enhancement gain coefficient γ according to the modulation function monotonically related to the turbidity factor;

[0083] In step 3.4, the pixel histogram of the grid sub-block is redistributed based on the modified clipping threshold clip and the enhancement gain coefficient γ.

[0084] Specifically, a baseline clipping threshold clip0 and a baseline enhancement gain coefficient γ0 are pre-set. We use a clear water sample calibration experiment to determine the baseline clipping threshold clip0 and the baseline enhancement gain coefficient γ0, which are usually defaulted to 1, corresponding to the standard contrast-limited histogram equalization method CLAHE. For each grid sub-block of the current original underwater image, the normalized turbidity factor T is used to redistribute the pixel histogram of the sub-block using the modified clipping threshold clip and enhancement gain coefficient γ. The clipping threshold clip is expressed as clip = clip0·f(T), and the enhancement gain coefficient γ is expressed as γ = γ0·g(T). Among them, f(T) and g(T) are modulation functions monotonically related to the turbidity factor. f(T) is selected as a decreasing function and g(T) is selected as an increasing function, so that the clipping threshold is reduced when the turbidity factor is larger.

[0085] Furthermore, by substituting the normalized turbidity factor T∈[0,1] into the increasing and decreasing linear modulation curves, expressed as f(T)=1–0.7T and g(T)=1+0.6T, we can obtain the modified clipping threshold clip based on clip=clip0·f(T) and γ=γ0·g(T). clip decreases monotonically with the turbidity factor T in the range [0.3clip0, clip0], while the enhancement gain coefficient γ increases monotonically in the range [γ0, 1.6γ0]. When the turbidity factor T is less than 0.3 (low turbidity), the clipping threshold clip and the enhancement gain coefficient γ approach the baseline values, i.e., clip≈clip0 and γ≈γ0. When the turbidity factor T is between 0.3–0.7 (medium turbidity), the clipping threshold clip and the enhancement gain coefficient γ linearly shrink and expand, respectively. When the turbidity factor T ≥ 0.7 (high turbidity), the clipping threshold clip is only close to 30% of the baseline value, and the enhancement gain coefficient γ is increased to nearly 1.6 times to compensate for the dim local contrast caused by haze.

[0086] The clipping threshold clip is used to truncate the pixel count of the sub-block histogram to prevent the noise from being over-equalized in the high turbidity factor T scene. The enhancement gain coefficient γ is used as a power-law gain to amplify or suppress the equalization result as a whole. The turbidity factor T only scales the clipping threshold clip and the enhancement gain coefficient γ through the above-mentioned modulation function, without participating in other operations. When implementing, these two parameters can be updated once before calling TCLAHE. After processing by the turbidity adaptive contrast limited histogram equalization method TCLAHE in step 3, the first intermediate image can be obtained. It can significantly improve the local contrast of the original underwater image, brighten the dark details, enrich the bright layer, and preliminarily reduce the gray effect caused by turbidity.

[0087] Then, based on improving the contrast, we further enhance the details and texture of the image to highlight the salient objects in the underwater scene. The specific steps include:

[0088] Step 4.1, extracting high-frequency details from the first intermediate image;

[0089] Step 4.2, using the Laplace or Sobel operator to obtain an edge map of the first intermediate image;

[0090] Step 4.3, extracting fine texture structures of the first intermediate image through high-pass filtering;

[0091] Step 4.4: Introduce visual saliency detection or edge amplitude to filter out weak textures and edges in edge maps and fine texture structures;

[0092] Step 4.5: Apply detail gain to the filtered edge image and fine texture structure and superimpose them on the first intermediate image to obtain a second intermediate image.

[0093] First, high-frequency detail extraction is performed on the first intermediate image, obtained after enhancement using the improved TCLAHE algorithm. The Laplace or Sobel operator is used to obtain an edge map. The Sobel operator synthesizes the edge map by calculating the xy gradient, while the Laplace operator detects edges through second-order differentials. The Laplace operator is more sensitive to fine edges. Fine texture structures are then extracted using high-pass filtering. By allowing high-frequency signals to pass and suppressing low-frequency components, the resulting high-frequency components represent the image's detail information. The degree of detail enhancement is then determined based on saliency. A visual saliency detection model is introduced, or simply based on edge amplitude, to filter out important edges and textures, thereby filtering out weak textures and edges caused by noise. Next, a detail gain is applied to the filtered edge map and fine texture structures. This means that the high-frequency components are amplified by a certain ratio and then superimposed back onto the first intermediate image, resulting in a second intermediate image. Let D(x, y) represent the filtered edge map and fine texture structures, and β be the detail gain coefficient. The pixel value update for the first intermediate image is then expressed as I′(x, y) = I(x, y) + β·D(x, y).

[0094] Furthermore, the detail gain coefficient β is controlled according to the turbidity factor T∈[0,1], and the detail gain coefficient β is always limited to [0.4,1.2]. Specifically, β_max=1.2, β_min=0.4, that is, β(T)=β-min+(β_max-β_min)·(1-T). When the ultra-low turbidity factor T is close to 0, that is, T≈0, the detail gain coefficient β is close to 1.2, that is, β≈1.2, to fully highlight the texture. When the ultra-high turbidity factor T is close to 1, that is, T≈1, the detail gain coefficient β is reduced to about 0.4 to suppress the pseudo edges generated by suspended particles. This significant detail enhancement processing can make the key target contours and surface details in the image clearer, such as the contours of underwater creatures and the texture of basement rocks, laying the foundation for subsequent fusion.

[0095] To adaptively consider the impact of water turbidity on deturbidity requirements during fusion, this paper introduces a turbidity-aware fusion strategy into the weight calculation. Its core is to couple the image deturbidity level with the fusion weights to ensure that the fusion result minimizes the effects of turbidity while maintaining visual naturalness. Specifically, the following steps are included:

[0096] Step 5.1, perform multi-scale decomposition on the original underwater image and the second intermediate image respectively to obtain sub-band representations from low frequency to high frequency;

[0097] Step 5.2, setting the fusion weight factor based on the turbidity factor;

[0098] Step 5.3, at each scale, fuse the corresponding sub-bands of the original underwater image and the second intermediate image according to the fusion weight factor.

[0099] Specifically, perform multi-scale fusion on the original underwater image and the detail-enhanced image obtained through the above enhancement processing, and output the final result by integrating the advantages of both. The fusion uses multi-scale image fusion methods such as Laplacian pyramid or wavelet decomposition, that is, perform multi-scale decomposition on the original underwater image and the second intermediate image respectively to obtain a series of sub-band representations from low frequency to high frequency. The high-frequency sub-bands contain detail information such as edges and textures. Set a fusion weight factor α, and this fusion weight factor α is determined based on the turbidity factor T. Then, at each scale, synthesize the corresponding sub-bands of the original underwater image and the second intermediate image according to a certain fusion weight, and finally reconstruct to obtain the fused image.

[0100] Furthermore, the fusion weight factor α is determined based on the turbidity factor T. Use linear modulation α(T)=α_min+(α_max - α_min)·T, where α_min = 0.3 and α_max = 1.0. Therefore, when the turbidity factor T ≤ 0.3 (low turbidity), the fusion weight factor α is close to 0.3, that is, α ≈ 0.3, and only take a small amount of enhanced components to maintain the authenticity of the original image. When the turbidity factor 0.2 < T < 0.8 (medium turbidity), the fusion weight factor α continuously changes between 0.3 - 0.8 with the turbidity factor T, realizing gradual enhancement. When the turbidity factor T ≥ 0.7 (high turbidity), the fusion weight factor α ≥ 0.79, and a large amount of detail-enhanced images are used to suppress the remaining turbidity. The turbidity factor T, as a global adjustment factor, affects the proportion configuration of the weights. During fusion, construct Laplacian pyramids for the original underwater image and the second intermediate image respectively. At each scale, select the enhanced image sub-bands according to the fusion weight factor α and select the original image sub-bands according to 1 - α. After synthesis, reconstruct and output the fused image, that is, the enhanced underwater image.

[0101] The turbidity-aware multi-scale fusion proposed in Step 5 enables the enhanced processing results to fully play their roles in the scenarios where turbidity removal is most needed. In relatively clear scenarios, it protects the authenticity of the original image. The fused image not only greatly eliminates the foggy feeling, but also has rich and clear details such as object edges and textures, and the overall tone and brightness are coordinated, without oversharpening or artifacts.

[0102] Through the above steps, the present invention achieves efficient image turbidity removal and enhancement for turbid underwater scenarios. Compared with the original turbid underwater image, the finally output enhanced underwater image has higher contrast, more vivid colors, and clearer details, realizing a significant improvement in the visibility of underwater scenarios. The method of the present invention can be applied to multiple fields such as underwater robot vision, submersible cameras, and ocean exploration, so as to help obtain more reliable image information in complex turbid water environments. Refer to Figures 3 to 5 As shown, Figure 3 (a)-(d) are comparisons of the effects before and after enhancement of NTU13-16 under medium turbidity conditions. Figure 4 (e)-(g) are comparisons of the effects before and after enhancement under medium turbidity conditions of NTU17-19. Figure 4 (h) is a comparison of the effects of NTU40 before and after enhancement under extremely high turbidity conditions. Figure 5 (a) is the input original underwater image, Figure 5 (b)-(k) are comparisons of the visual effects before and after enhancement using different methods at different turbidities. Figure 5 (1) is the output effect diagram of the method of the present invention.

[0103] To implement the above-mentioned embodiment, the present invention proposes an enhanced visual system for deturbidity of turbid underwater scenes, comprising: a turbidity analysis module for obtaining a turbidity factor from the original underwater image using a turbidity meter or by estimating the visual features of the image; a turbidity adaptive contrast enhancement module for performing turbidity adaptive contrast limited histogram equalization on the original underwater image based on the turbidity factor to obtain a first intermediate image; a salient detail enhancement module for extracting high-frequency components and detecting salient areas from the first intermediate image, and performing gain amplification on salient edges and textures based on the turbidity factor to obtain a second intermediate image; and a multi-scale fusion module for performing multi-scale decomposition of the original underwater image and the second intermediate image, and reconstructing the corresponding fused output after assigning fusion weights at each scale based on the turbidity factor and image features.

[0104] In order to implement the above embodiment, the present invention also proposes an electronic device. The electronic device may include a processing device (such as a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) or a program loaded from a storage device into a random access memory (RAM). In the RAM, various programs and data required for the operation of the electronic device are also stored. The processing device, ROM, and RAM are connected to each other via a bus. An input / output (I / O) interface is also connected to the bus. Typically, the following devices can be connected to the I / O interface: input devices including, for example, a touch screen, a touchpad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; output devices including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices including, for example, a magnetic tape, a hard disk, and a communication device. The communication device can allow the electronic device to communicate with other devices wirelessly or wired to exchange data.

[0105] In addition, the present invention also provides a computer storage medium. The computer storage medium can be a computer-readable signal medium or a computer-readable storage medium or any combination of the above. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or component, or any combination of the above. More specific examples of computer-readable storage media can include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0106] In the present invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In some embodiments of the present disclosure, a computer-readable signal medium can include a data signal transmitted in baseband or as part of a carrier wave, carrying computer-readable program code. This transmitted data signal can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can also be any computer storage medium other than a computer-readable storage medium that can transmit, propagate, or transfer a program for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the computer storage medium can be transmitted using any suitable medium, including but not limited to wires, optical cables, RF (radio frequency), or any suitable combination thereof. The computer storage medium can be included in the electronic device or exist independently, not incorporated into the electronic device. The computer storage medium carries one or more programs that, when executed by the electronic device, enable the electronic device to implement a method for enhancing visualization of turbid underwater scenes.

[0107] The above description is only an illustration of some preferred embodiments of the present disclosure and the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by the specific combination of the above-mentioned technical features, but should also cover other technical solutions formed by any combination of the above-mentioned technical features or their equivalent features without departing from the above-mentioned inventive concept. For example, the above-mentioned features are replaced with (but not limited to) technical features with similar functions disclosed in the embodiments of the present disclosure.

Claims

1. A method for enhancing visibility by removing turbidity from turbid underwater scenes, characterized in that: The steps include: Step 1, in response to acquiring a turbid original underwater image; Step 2, obtaining a turbidity factor using a turbidity meter or by image visual feature estimation based on the original underwater image; Step 3, performing turbidity adaptive contrast limited histogram equalization processing on the original underwater image according to the turbidity factor to obtain a first intermediate image; Step 4: extract high-frequency components and detect significant areas on the first intermediate image, and perform gain amplification on significant edges and textures based on the turbidity factor to obtain a second intermediate image; Step 5: Decompose the original underwater image and the second intermediate image at multiple scales, assign fusion weights at each scale based on the turbidity factor and image features, and then fuse them accordingly. Step 6: reconstruct and output the fused results to obtain the final enhanced underwater image.

2. The method for enhancing visibility by removing turbidity from turbid underwater scenes according to claim 1, characterized in that: The calculation of the turbidity factor in step 2 further includes the following steps: Step 2.1, performing scaling on the acquired original underwater image; Step 2.2, extract the image average gradient amplitude, image brightness standard deviation and image pixel saturation average in grayscale domain and HSV color space simultaneously; Step 2.3, calculating the turbidity score based on the image average gradient amplitude, the standard deviation of the image brightness, and the average value of the image pixel saturation; Step 2.4: normalize the turbidity score to obtain a turbidity factor corresponding to the actual physical turbidity.

3. The method for enhancing visibility by removing turbidity from turbid underwater scenes according to claim 2, characterized in that: The step 3 further comprises the following steps: Step 3.1, dividing the original underwater image into a plurality of grid sub-blocks; Step 3.2: Use the clean water sample to calibrate the baseline clipping threshold clip0 and the baseline enhancement gain coefficient γ0; Step 3.3, obtaining a modified clipping threshold clip and enhancement gain coefficient γ according to a modulation function monotonically related to the turbidity factor; Step 3.4: redistribute the pixel histogram of the grid sub-block based on the modified clipping threshold clip and the enhancement gain coefficient γ.

4. The method for enhancing visibility by removing turbidity from turbid underwater scenes according to claim 3, characterized in that: The step 4 further comprises the following steps: Step 4.1, extracting high-frequency details from the first intermediate image; Step 4.2, using a Laplace or Sobel operator to obtain an edge map of the first intermediate image; Step 4.3, extracting fine texture structures of the first intermediate image through high-pass filtering; Step 4.4: Introduce visual saliency detection or edge amplitude to filter out weak textures and edges in edge maps and fine texture structures; Step 4.5: Apply detail gain to the filtered edge image and fine texture structure and superimpose them on the first intermediate image to obtain a second intermediate image.

5. The method for enhancing visibility by removing turbidity from turbid underwater scenes according to claim 4, characterized in that: The step 5 further comprises the following steps: Step 5.1, performing multi-scale decomposition on the original underwater image and the second intermediate image respectively to obtain sub-band representations from low frequency to high frequency; Step 5.2, setting a fusion weight factor based on the turbidity factor; Step 5.3: Fusing the corresponding sub-bands of the original underwater image and the second intermediate image at each scale according to the fusion weight factor.

6. The method for enhancing visibility by removing turbidity from turbid underwater scenes according to claim 3, characterized in that: In step 3.3, the turbidity factor T is substituted into the monotonically related increasing function g(T) and decreasing function f(T), and the clipping threshold clip and enhancement gain coefficient γ are corrected based on clip = clip0·f(T) and γ = γ0·g(T); When the turbidity factor T < 0.3, the clipping threshold clip and the enhancement gain coefficient γ approach the reference value; When the turbidity factor T is between 0.3 and 0.7, the clipping threshold clip and the enhancement gain coefficient γ linearly contract and expand; When the turbidity factor T ≥ 0.7, the clipping threshold clip approaches 30% of the reference value, and the enhancement gain coefficient γ is increased to nearly 1.6 times.

7. The method for enhancing visibility by removing turbidity from turbid underwater scenes according to claim 4, characterized in that: In step 4.5, the detail gain coefficient β is controlled according to the turbidity factor T. The detail gain coefficient β is limited within the range of [0.4, 1.2] and is expressed as β(T) = β_min + (β_max - β_min)·(1 - T), where β_max = 1.2 and β_min = 0.4; When the turbidity factor T approaches 0, the detail gain coefficient β approaches 1.2, which can fully highlight the texture; When the turbidity factor T approaches 1, the detail gain coefficient β drops to nearly 0.4 to suppress the pseudo-edges generated by suspended particles.

8. The method for enhancing visibility by removing turbidity from turbid underwater scenes according to claim 5, characterized in that: In step 5.2, the fusion weight factor α is determined according to the turbidity factor T and is expressed by linear modulation as α(T) = α_min + (α_max - α_min)·T, where α_min = 0.3 and α_max = 1.0; When the turbidity factor T ≤ 0.3, the fusion weight factor α approaches 0.3, and only a small amount of enhanced components are taken to maintain the authenticity of the original image; When the turbidity factor T is 0.2 < T < 0.8, the fusion weight factor α continuously changes between 0.3 and 0.8 with the turbidity factor T to achieve gradual enhancement; When the turbidity factor T ≥ 0.7, the fusion weight factor α ≥ 0.79, and a large amount of detail-enhanced images are used to suppress the turbidity residue.

9. An enhanced visual system for de-turbidity of turbid underwater scenes, characterized in that: [[ID= ​ ​ ​ ​

Images