Underwater image enhancement method based on frequency domain and spatial domain combined processing
The joint processing of frequency and spatial domains in underwater image enhancement addresses color distortion and low contrast by using quadtree search, channel compensation, and homomorphic filtering, resulting in improved image quality and detail enhancement.
Patent Information
- Application Number
- CN202411274576.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-12
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2044-09-12
AI Technical Summary
Underwater images suffer from color distortion, low contrast, and blurred details due to varying water environments, limiting the applicability of existing image enhancement methods.
An underwater image enhancement method based on joint processing of frequency domain and spatial domain, involving quadtree search for background area, channel difference calculation, color cast correction, homomorphic filtering for contrast enhancement, and fusion in Lab color space.
The method effectively corrects color casts, enhances contrast, and improves image details, achieving adaptive correction and restoration of underwater images.
Smart Images

Figure CN119090790A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of underwater image processing, and in particular to an underwater image enhancement method based on frequency domain and spatial domain joint processing. Background Art
[0002] Underwater optical imaging technology has the advantages of fast imaging speed and high spatial resolution, and is widely used in fields such as marine resource exploration, marine biological research and underwater combat. However, particles present in underwater environments with low visibility will scatter and absorb target signals, resulting in a decrease in image contrast. In addition, the attenuation of light underwater depends on the wavelength of light, resulting in color distortion in color images taken underwater.
[0003] There are many methods currently used to enhance the quality of underwater images, which can be roughly divided into computer vision-based methods and physical model-based methods. Visual enhancement methods do not consider the physical model of image degradation, and directly process the image pixel by pixel to improve image quality without relying on the physical model. Physical model-based methods consider the propagation characteristics and polarization characteristics of light underwater, use these characteristics to establish an imaging degradation model, and restore the image by inversely solving the model. Physical model-based methods require the use of some external devices or some prior knowledge, while visual enhancement methods are simpler to implement than image restoration methods, involving only the adjustment of image pixel values, without considering the degradation inversion process and assuming prior knowledge. However, underwater images will present different characteristics due to the different water environments in which they are shot, such as different color casts, so the applicability of a single image enhancement method is limited, and it is difficult to simultaneously deal with color distortion, low contrast, and blurred details. Summary of the invention
[0004] The present invention aims to solve the deficiencies of the prior art, realize the judgment of water color, realize adaptive correction of underwater image color, realize contrast enhancement of underwater images, and provide an underwater image enhancement method based on joint processing of frequency domain and spatial domain.
[0005] In order to achieve the above object, the present invention adopts the following technical solutions:
[0006] An underwater image enhancement method based on frequency domain and spatial domain joint processing comprises the following steps:
[0007] S1. Obtain a public underwater image dataset;
[0008] S2, find the best background area for the public dataset;
[0009] S3, according to the best background area obtained in step S2, respectively calculate the channel difference M between the maximum channel and the minimum channel;
[0010] S4, taking statistics of the channel difference M obtained in step S3, and comparing it with the channel difference M of the color-corrected image, to obtain a color cast factor threshold of the color cast image;
[0011] S5, based on the color cast factor threshold obtained in step S4, the color cast image is classified into green, blue and yellow;
[0012] S6, performing channel compensation on the different color cast images obtained in step S5 to correct the image color cast;
[0013] S7, using an improved homomorphic filtering method to improve the contrast of the original image;
[0014] S8. The color correction image obtained in step S6 and the contrast enhancement image obtained in step S7 are merged in Lab color space to obtain a color correction and contrast enhancement image.
[0015] In step S2, the method for finding the best background area is a quadtree search method, and the formula is as follows:
[0016]
[0017] in,
[0018] represents the variance of the c-channel of the ith sub-region;
[0019] i∈{1,2,3,4} represents the i-th block of the segmented image in each iteration;
[0020] m*n represents the number of pixels in the i-th region;
[0021] c∈{r,g,b} represents the red, green, and blue color channels of the image;
[0022] Represents the pixel value of the c channel at point x;
[0023] Represents the mean value of image pixels in the i-th region.
[0024] In step S3, the mean difference M between the maximum channel and the minimum channel is obtained by the following method, and the obtaining formula is:
[0025]
[0026] in,
[0027] Represents the maximum channel mean of the image;
[0028] Represents the minimum channel mean of the image.
[0029] In step S4, the method for obtaining the color cast factor threshold is:
[0030] The channel difference M obtained in step S3 is statistically distributed, and the channel difference M of the color-corrected image is calculated. The range of the channel difference M of the color-corrected image in the statistical distribution interval is compared to obtain the color cast factor threshold. When the image channel difference M> the color cast factor threshold, the image color is distorted. When the image channel difference M≤ the color cast factor threshold, the image does not need to be corrected.
[0031] In step S5, the color cast image is divided into green cast, blue cast and yellow cast. The judgment basis is that the formula of the color channel of the processed image is:
[0032]
[0033] in,
[0034] m c represents the mean value of each channel;
[0035] p i (x,y) represents the pixel value at the optimal background area (x,y);
[0036] n represents the number of pixels in the optimal background area;
[0037] c represents different color channels;
[0038] represents the mean of the image;
[0039] D c It represents the difference between the mean of each channel and the image mean.
[0040] In step S6, the color correction method includes:
[0041] For green-biased images, the green channel is used as the reference channel to compensate for the red and blue channels. The compensation formula is:
[0042]
[0043] For yellowish images, the red channel is used as the reference channel, and the green and blue channels are compensated. The compensation formula is:
[0044]
[0045] For bluish images, the blue channel is used as the reference channel to compensate the green channel, and the compensated green channel is used as the reference channel to compensate the red channel. The compensation formula is:
[0046]
[0047] in,
[0048] I rc (x), I gc (x), I bc (x) represents the red channel, green channel, and blue channel after color correction, respectively;
[0049] I r (x), I g (x), I b (x) represents the red channel, green channel, and blue channel of the original underwater image, respectively;
[0050] Respectively represent the pixel mean of the red channel, green channel, and blue channel in the best background area;
[0051] α and β represent constants.
[0052] In step S7, homomorphic filtering is performed on the image to enhance contrast, including:
[0053] S71. Use homomorphic filtering to enhance high-frequency information and suppress background low-frequency information. The original image is represented by the illumination-reflection model:
[0054] f(x,y)=i(x,y)·r(x,y);
[0055] in,
[0056] f(x,y) represents the grayscale image of the original underwater image;
[0057] i(x,y) represents the incident component;
[0058] r(x,y) represents the reflected component;
[0059] (x,y) represents the image coordinates;
[0060] Take the logarithm of both sides of the above equation:
[0061] lnf(x,y)=lni(x,y)+lnr(r,y);
[0062] Perform Fourier transform, which is:
[0063] F(u,v)=I(u,v)+R(u,v);
[0064] in,
[0065] F(u,v), I(u,v), and R(u,v) represent the Fourier transform of lnf(x,y), lni(x,y), and lnr(r,y), respectively;
[0066] (μ,υ) represents the image coordinates in the frequency domain;
[0067] Filtering is done on both sides:
[0068] S(μ,υ)=H(u,v)F(u,v)=H(u,v)I(u,v)+H(u,v)R(u,v);
[0069] in,
[0070] S(μ,υ) represents the Fourier transform of the result;
[0071] H(u,v) represents the transfer function of the homomorphic filter;
[0072] Step S72: Based on homomorphic filtering, a new transfer function S-curve is introduced. The expression of the S-curve is:
[0073]
[0074] in,
[0075] p(t) represents the ordinate value of the introduced S-curve;
[0076] x represents the horizontal coordinate value;
[0077] t represents the control curve parameter;
[0078] According to the S-curve having a structure similar to the homomorphic filter profile, a new transfer function is constructed as:
[0079]
[0080] in,
[0081] D0 represents the cut-off frequency;
[0082] k represents the parameter of the control function;
[0083] D(μ,υ) represents the distance from the frequency (μ,υ) to the filter center (μ0,υ0);
[0084] S73, performing inverse Fourier transform on the image after frequency domain filtering:
[0085] s(x,y)=IDFT[S(u,v)];
[0086] in,
[0087] s(x,y) represents the inverse Fourier transform result after the image is filtered;
[0088] IDFT stands for Inverse Fourier Transform;
[0089] Taking the antilogarithm and restoring it to the airspace, we get:
[0090] g(x,y)=exp s(x,y) =i0(x,y)·r0(x,y);
[0091] in,
[0092] g(x,y) is the image restored to the spatial domain;
[0093] i0(x,y) represents the incident component of the image after filtering;
[0094] r0(x,y) represents the illumination component of the image after filtering.
[0095] In step S8, the color correction image in step S6 and the contrast enhancement image in step S7 are fused in the Lab color space to obtain a color correction and contrast enhancement image. The specific process is as follows:
[0096] I fused =ω·I dz +a cc +b cc ;
[0097] in,
[0098] I fused Represents the image after fusion in Lab color space;
[0099] I fused represents the fusion coefficient;
[0100] I dz represents the brightness channel of the contrast-enhanced image;
[0101] a cc Represents the a channel of the color-corrected image in the Lab color space;
[0102] b cc Represents the b channel of the color-corrected image in the Lab color space.
[0103] The beneficial effects of the present invention are as follows: the algorithm described in the present invention recognizes and corrects the color cast of underwater images, uses homomorphic filtering to achieve detail enhancement, and finally performs fusion processing in the Lab color space to achieve water color judgment, achieve adaptive correction of underwater image colors, and achieve contrast enhancement, detail enhancement and color restoration of underwater images. BRIEF DESCRIPTION OF THE DRAWINGS
[0104] Figure 1 It is a schematic diagram of the process of the present invention;
[0105] Figure 2 The difference distribution diagram of the maximum attenuation channel and the lowest attenuation channel of the underwater image statistically obtained by the present invention;
[0106] Figure 3 This is a comparison diagram of the color correction and contrast enhancement of underwater green color cast images by the present invention and other algorithms;
[0107] Figure 4 This is a comparison diagram of the color correction and contrast enhancement of underwater yellow color cast images by the present invention and other algorithms;
[0108] Figure 5 This is a comparison diagram of the color correction and contrast enhancement of underwater blue color cast images by the present invention and other algorithms;
[0109] The following is a detailed description of the embodiments of the present invention with reference to the accompanying drawings. DETAILED DESCRIPTION
[0110] The principles and features of the present invention are described below in conjunction with the accompanying drawings. The embodiments are only used to explain the present invention and are not used to limit the scope of the present invention. The present invention is described in more detail by way of example with reference to the accompanying drawings in the following paragraphs. The advantages and features of the present invention will become clearer according to the following description. It should be noted that the accompanying drawings are all in a very simplified form and are not in precise proportions, and are only used to facilitate and clearly assist in explaining the purpose of the embodiments of the present invention.
[0111] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items.
[0112] The present invention will be further described below in conjunction with the accompanying drawings and embodiments:
[0113] An underwater image enhancement method based on frequency domain and spatial domain joint processing, such as Figure 1 As shown, the following steps are included:
[0114] S1. Obtain a public underwater image dataset, which is the UIEB dataset.
[0115] S2, find the best background area for the public dataset;
[0116] Based on the quadtree method, the best background area of the above-mentioned public underwater image dataset is searched, that is, the area with the minimum image variance. The formula is as follows:
[0117]
[0118] in,
[0119] represents the variance of the c channel of the ith sub-region; i∈{1,2,3,4} represents the ith block of the segmented image in each iteration; m*n represents the number of pixels in the ith region; c∈{r,g,b} represents the red, green and blue color channels of the image; Represents the pixel value of the c channel at point x; Represents the mean value of image pixels in the i-th region.
[0120] S3, according to the best background area obtained in step S2, respectively calculate the channel difference M between the maximum channel and the minimum channel;
[0121] The mean difference M between the maximum channel and the minimum channel is obtained by the following method, and the formula is:
[0122]
[0123] in,
[0124] Represents the maximum channel mean of the image; Represents the minimum channel mean of the image.
[0125] S4. Statistically distribute the channel difference M obtained in step S3, calculate the channel difference M of the color-corrected image, compare the range of the channel difference M of the color-corrected image in the statistical distribution interval, and obtain the color cast factor threshold. When the image channel difference M> the color cast factor threshold, the image color is distorted. When the image channel difference M≤ the color cast factor threshold, the image does not need to be corrected.
[0126] The specific operations are as follows:
[0127] Here, the channel difference M is a concept. First, solve the channel difference M of the underwater image dataset obtained in step S1 (the public underwater image dataset obtained in step S1 includes the real color and distorted ones), and make a statistical analysis of the distribution of M values. For example, suppose the channel difference M obtained in this underwater image dataset (100 images) is (0.5, 0.6, 0.9, 0.2, 0.1, ..., ..., ...), which is 100 values. Make a statistical distribution of these values, such as Figure 2 shown.
[0128] Select the images with real colors in this dataset, and then extract the channel difference M of these images. For example, find images with real colors in this dataset of 100 underwater images. Assume that 10 images are found, and the channel difference M of these 10 images are (0.01, 0.1, 0.2, 0.05, ..., ...) respectively.
[0129] The channel difference M of the color-true image is compared in the area where the statistical distribution is located, so as to obtain the image color cast factor threshold; in this example, the channel difference M of the color-true image is less than 0.2, and the image color cast threshold is determined to be 0.2. When M>0.2, the image color is distorted, and when M≤0.2, the image does not need to be corrected.
[0130] S5. Based on the color cast factor threshold obtained in step S4, the color cast image is divided into green, blue and yellow. The judgment basis is that the formula of the color channel of the processed image is:
[0131]
[0132]
[0133] in,
[0134] m c represents the mean value of each channel; p i (x, y) represents the pixel value at the optimal background area (x, y); n represents the number of pixels in the optimal background area; c represents different color channels; represents the mean of the image; D c It represents the difference between the mean of each channel and the image mean.
[0135] S6, performing channel compensation on the different color cast images obtained in step S5 to correct the image color cast;
[0136] Color correction methods include:
[0137] For green-biased images, the green channel has the lowest attenuation, so the green channel is used as the reference channel to compensate the red and blue channels. The compensation formula is:
[0138]
[0139] For yellowish images, the red channel has the lowest attenuation, so the red channel is used as the reference channel to compensate the green and blue channels. The compensation formula is:
[0140]
[0141] For bluish images, the blue channel is used as the channel with the least attenuation. However, unlike the other two tonal compensations, the image will appear reddish after the blue channel is used to compensate the red channel. This is caused by overcompensation of the red channel. Therefore, the blue channel is used as the reference channel to compensate the green channel, and the compensated green channel is used as the reference channel to compensate the red channel. The compensation formula is:
[0142]
[0143] in,
[0144] I rc (x), I gc (x), I bc (x) represents the red channel, green channel, and blue channel after color correction, respectively; I r (x), I g (x), I b (x) represents the red channel, green channel, and blue channel of the original underwater image, respectively; They represent the pixel means of the red channel, green channel, and blue channel in the optimal background area respectively; α and β represent constants.
[0145] In order to further improve the image quality, after the attenuation channel compensation, the local histogram stretching method is used to adjust the dynamic range of each channel, and the color deviation is removed by linear stretching. The specific formula is:
[0146]
[0147] where c∈{R,G,B}; L c , U c They represent the lower and upper limits of the tail removed from the histogram of each channel of the image, which are determined by removing 0.1% of the pixels to select the interception position of the tail; I c (x) is the image after channel compensation; I c (x) is the image after the final linear stretching to remove the color cast.
[0148] S7: Based on the image after color compensation, the image contrast is still low, and the contrast is further improved, including:
[0149] S71. Use homomorphic filtering to enhance high-frequency information and suppress background low-frequency information. The original image is represented by the illumination-reflection model:
[0150] f(x,y)=i(x,y)·r(x,y);
[0151] in,
[0152] f(x,y) represents the grayscale image of the original underwater image; i(x,y) represents the incident component; r(x,y) represents the reflected component; (x,y) represents the image coordinates;
[0153] Take the logarithm of both sides of the above equation:
[0154] lnf(x,y)=lni(x,y)+lnr(r,y);
[0155] Perform Fourier transform, which is:
[0156] F(u,v)=I(u,v)+R(u,v);
[0157] in,
[0158] F(u,v), I(u,v), R(u,v) represent the Fourier transform of lnf(x,y), lni(x,y), lnr(r,y), respectively; (μ,υ) represents the image coordinates in the frequency domain;
[0159] Filtering is performed on both sides:
[0160] S(μ,υ)=H(u,v)F(u,v)=H(u,v)I(u,v)+H(u,v)R(u,v);
[0161] in,
[0162] S(μ,υ) represents the Fourier transform of the result; H(u,v) represents the transfer function of the homomorphic filter;
[0163] H(u,v) can be expressed as follows:
[0164]
[0165] Among them, r H represents the high frequency gain, r L represents the low-frequency gain, D0 represents the cutoff frequency, c is a constant that controls the filter shape, and D(μ,υ) represents the distance from the frequency (μ,υ) to the filter center (μ0,υ0).
[0166] The transfer function can be changed by setting different parameters. However, since there are many control parameters, it is not conducive to parameter setting.
[0167] Step S72: In order to simplify the setting of the transfer function parameters of the homomorphic filter, a new transfer function S-curve is introduced, which has the characteristics of a faster growth rate in the initial stage and a slower growth rate in the middle stage, and is very close to the transfer function curve of the homomorphic filter. The S-curve expression is:
[0168]
[0169] in,
[0170] p(t) represents the ordinate value of the introduced S-curve; x represents the abscissa value; t represents the control curve parameter;
[0171] According to the S-curve having a structure similar to the homomorphic filter profile, a new transfer function is constructed as:
[0172]
[0173] in,
[0174] D0 represents the cutoff frequency; k represents the parameter of the control function; D(μ,υ) represents the distance from the frequency (μ,υ) to the filter center (μ0,υ0);
[0175] S73, performing inverse Fourier transform on the image after frequency domain filtering:
[0176] s(x,y)=IDFT[S(u,v)];
[0177] in,
[0178] s(x,y) represents the inverse Fourier transform result after the image is filtered; IDFT represents the inverse Fourier transform;
[0179] Taking the antilogarithm and restoring it to the airspace, we get:
[0180] g(x,y)=exp s(x,y) =i0(x,y)·r0(x,y);
[0181] in,
[0182] g(x,y) is the restored spatial domain image; i0(x,y) represents the incident component of the image after filtering; r0(x,y) represents the irradiation component of the image after filtering.
[0183] S8, the color correction image obtained in step S6 and the contrast enhancement image obtained in step S7 are merged in the Lab color space to obtain a color correction and contrast enhancement image. The specific process is as follows:
[0184] I fused =ω·I dz +a cc +b cc ;
[0185] in,
[0186] I fused Represents the image after fusion in Lab color space; I fused Represents the fusion coefficient; I dz represents the brightness channel of the contrast-enhanced image; a cc a channel of the color-corrected image in the Lab color space; b cc Represents the b channel of the color-corrected image in the Lab color space. Specific embodiment:
[0188] Figure 3 , Figure 4 , Figure 5The figures are experimental effect diagrams of underwater images with different color casts processed by the present invention and other algorithms, wherein UICONM represents underwater image contrast measurement; UICM represents underwater image color measurement; UISM represents underwater image clarity measurement; UIQM (Universal Image Quality Metric) loss function is a commonly used image quality measurement method, which can accurately evaluate the effect of image processing algorithm and the quality of image restoration; UCIQE (Uncertainty in Color, Intensity, and Saturation of an Image) is an indicator for evaluating underwater image quality, which is a linear combination of color concentration, saturation and contrast; AG (Autogynephilia) represents average gradient, the larger the average gradient, the better, the larger the image details and the higher the clarity; PCQI (patch-based contrast quality index) represents patch-based contrast quality index.
[0189] From the analysis of the experimental effect diagrams, (a) original image, (b) CBAF algorithm, (c) DAC algorithm, (d) GDCP algorithm, (e) ACDC algorithm, (f) TWO algorithm, (g) PCDE algorithm and (h) the algorithm of the present invention, it can be seen that other algorithms cannot solve the color cast and contrast enhancement at the same time, while after the image is processed by the algorithm of the present invention, the color cast of the image is improved, the contrast enhancement is moderate, the details and texture of the image are clear, and the subjective visual quality is better than that of several other image processing algorithms.
[0190] Table 1 Average objective evaluation indicators of different algorithms for different images
[0191]
[0192]
[0193] Table 1 shows the objective evaluation values. The larger the value of the evaluation index is, the clearer the image is and the better the image enhancement effect is. It can be seen from the table that the method of the present invention shows a higher evaluation value as a whole, indicating that the method of the present invention has a more significant effect in enhancing image quality than other methods.
[0194] The present invention is described above by way of example in conjunction with the accompanying drawings. It is obvious that the specific implementation of the present invention is not limited to the above-mentioned method. As long as various improvements are made using the method concept and technical solution of the present invention, or are directly applied to other occasions without improvement, they are all within the protection scope of the present invention.
Claims
1. An underwater image enhancement method based on frequency domain and spatial domain joint processing, characterized in that: The following steps are involved: S1. Obtain a public underwater image dataset; S2, find the best background area for the public dataset; S3, according to the best background area obtained in step S2, respectively calculate the channel difference M between the maximum channel and the minimum channel; S4, taking statistics of the channel difference M obtained in step S3, and comparing it with the channel difference M of the color-corrected image, to obtain a color cast factor threshold of the color cast image; S5, based on the color cast factor threshold obtained in step S4, the color cast image is classified into green, blue and yellow; S6, performing channel compensation on the different color cast images obtained in step S5 to correct the image color cast; S7, using an improved homomorphic filtering method to improve the contrast of the original image; S8. The color correction image obtained in step S6 and the contrast enhancement image obtained in step S7 are merged in Lab color space to obtain a color correction and contrast enhancement image.
2. An underwater image enhancement method based on frequency domain and spatial domain joint processing according to claim 1, characterized in that: In step S2, the method for finding the best background area is a quadtree search method, and the formula is as follows: in, represents the variance of the c-channel of the ith sub-region; i∈{1,2,3,4} represents the i-th block of the segmented image in each iteration; m*n represents the number of pixels in the i-th region; c∈{r,g,b} represents the red, green, and blue color channels of the image; Represents the pixel value of the c channel at point x; Represents the mean value of image pixels in the i-th region.
3. The underwater image enhancement method based on frequency domain and spatial domain joint processing according to claim 1, characterized in that: In step S3, the mean difference M between the maximum channel and the minimum channel is obtained by the following method, and the obtaining formula is: in, Represents the maximum channel mean of the image; Represents the minimum channel mean of the image.
4. The underwater image enhancement method based on frequency domain and spatial domain joint processing according to claim 1, characterized in that: In step S4, the method for obtaining the color cast factor threshold is: The channel difference M obtained in step S3 is statistically distributed, and the channel difference M of the color-corrected image is calculated. The range of the channel difference M of the color-corrected image in the statistical distribution interval is compared to obtain the color cast factor threshold. When the image channel difference M> the color cast factor threshold, the image color is distorted. When the image channel difference M≤ the color cast factor threshold, the image does not need to be corrected.
5. An underwater image enhancement method based on frequency domain and spatial domain joint processing according to claim 4, characterized in that: In step S5, the color cast image is divided into green cast, blue cast and yellow cast. The judgment basis is that the formula of the color channel of the processed image is: in, m c represents the mean value of each channel; p i (x,y) represents the pixel value at the optimal background area (x,y); n represents the number of pixels in the optimal background area; c represents different color channels; Represents the mean of the image; D c It represents the difference between the mean of each channel and the image mean.
6. The underwater image enhancement method based on frequency domain and spatial domain joint processing according to claim 1, characterized in that: In step S6, the color correction method includes: For green-biased images, the green channel is used as the reference channel to compensate for the red and blue channels. The compensation formula is: For yellowish images, the red channel is used as the reference channel, and the green and blue channels are compensated. The compensation formula is: For bluish images, the blue channel is used as the reference channel to compensate the green channel, and the compensated green channel is used as the reference channel to compensate the red channel. The compensation formula is: in, I rc (x), I gc (x), I bc (x) represents the red channel, green channel, and blue channel after color correction, respectively; I r (x), I g (x), I b (x) represents the red channel, green channel, and blue channel of the original underwater image, respectively; Respectively represent the pixel mean of the red channel, green channel, and blue channel in the best background area; α and β represent constants.
7. The underwater image enhancement method based on frequency domain and spatial domain joint processing according to claim 1, characterized in that: In step S7, homomorphic filtering is performed on the image to enhance contrast, including: S71. Use homomorphic filtering to enhance high-frequency information and suppress background low-frequency information. The original image is represented by the illumination-reflection model: f(x,y)=i(x,y)·r(x,y); in, f(x,y) represents the grayscale image of the original underwater image; i(x,y) represents the incident component; r(x,y) represents the reflected component; (x,y) represents the image coordinates; Take the logarithm of both sides of the above equation: lnf(x,y)=lni(x,y)+lnr(r,y); Perform Fourier transform, which is: F(u,v)=I(u,v)+R(u,v); in, F(u,v), I(u,v), and R(u,v) represent the Fourier transform of lnf(x,y), lni(x,y), and lnr(r,y), respectively; (μ,υ) represents the image coordinates in the frequency domain; Filtering is performed on both sides: S(μ,υ)=H(u,v)F(u,v)=H(u,v)I(u,v)+H(u,v)R(u,v); in, S(μ,υ) represents the Fourier transform of the result; H(u,v) represents the transfer function of the homomorphic filter; Step S72: Based on homomorphic filtering, a new transfer function S-curve is introduced. The expression of the S-curve is: in, p(t) represents the ordinate value of the introduced S-curve; x represents the horizontal coordinate value; t represents the control curve parameter; According to the S-curve having a structure similar to the homomorphic filter profile, a new transfer function is constructed as: in, D0 represents the cut-off frequency; k represents the parameter of the control function; D(μ,υ) represents the distance from the frequency (μ,υ) to the filter center (μ0,υ0); S73, performing inverse Fourier transform on the image after frequency domain filtering: s(x,y)=IDFT[S(u,v)]; in, s(x,y) represents the inverse Fourier transform result after the image is filtered; IDFT stands for Inverse Fourier Transform; Taking the antilogarithm and restoring it to the airspace, we get: g(x,y)=exp s(x,y) =i0(x,y)·r0(x,y); in, g(x,y) is the image restored to the spatial domain; i0(x,y) represents the incident component of the image after filtering; r0(x,y) represents the illumination component of the image after filtering.
8. The underwater image enhancement method based on frequency domain and spatial domain joint processing according to claim 1, characterized in that: In step S8, the color correction image in step S6 and the contrast enhancement image in step S7 are fused in the Lab color space to obtain a color correction and contrast enhancement image. The specific process is as follows: I fused =ω·I dz +a cc +b cc ; in, I fused Represents the image after fusion in Lab color space; I fused represents the fusion coefficient; I dz represents the brightness channel of the contrast-enhanced image; a cc Represents the a channel of the color-corrected image in the Lab color space; b cc Represents the b channel of the color-corrected image in the Lab color space.
Citation Information
Patent Citations
Underwater image enhancement method combining frequency domain and spatial domain
CN110533614A
Underwater image enhancement method based on selective color compensation and three-interval equalization
CN111127359A
Underwater image enhancement method
CN115034985A
Underwater image processing method and system
CN116320344A
Underwater image enhancement method and enhancement device
GB2585754A
Cited By
Deepwater image enhancement method based on multi-color space coupling
CN120410953A
A deep-water image enhancement method based on multi-color space coupling
CN120410953B
Underwater image enhancement method and system based on double-domain collaboration
CN121053048A
