Underwater polarization image enhancement method based on tensor low-rank decomposition

By acquiring multi-angle polarization images and performing tensor low-rank decomposition, the problems of color cast and low contrast in underwater polarization image enhancement are solved, achieving a more efficient image enhancement effect.

CN118918052APending Publication Date: 2024-11-08DALIAN MARITIME UNIVERSITY
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410974563.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-19
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

Existing underwater polarization image enhancement methods assume that the polarization degree of the target and the backscattered light is the same, and fail to effectively consider the non-uniformity of the polarization information of the backscattered light, resulting in unsatisfactory image enhancement effects, especially low contrast and color distortion.

Method used

By acquiring polarization images of the same scene at 0 degrees, 45 degrees, 90 degrees, and 135 degrees, the Stokes vector and degree of polarization of the polarized light are calculated. Histogram stretching and color correction are performed to construct a tensor low-rank model of the backscattered polarization degree. The backscattered polarization degree is separated using the alternating direction multiplier method. The intensity of the backscattered light is calculated, and the atmospheric light vector at infinity is obtained through the weighted atmospheric light function. Finally, the enhanced underwater image is obtained by inversion.

Benefits of technology

It overcomes the limitations of the backscattered light assumption in traditional polarization imaging methods, effectively corrects the problems of image color shift and low contrast, and improves the enhancement effect of underwater images.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118918052A_ABST
    Figure CN118918052A_ABST
Patent Text Reader

Abstract

The invention provides an underwater polarization image enhancement method based on tensor low-rank decomposition, and the method comprises the steps: obtaining and processing 0-degree, 45-degree, 90-degree and 135-degree polarization images of the same scene, and obtaining a stretched 90-degree polarization image, a stretched 135-degree polarization image, a corrected 0-degree polarization image and a corrected 45-degree polarization image, the calculation module is used for calculating corrected image polarization degree; a tensor low-rank model of the backscattering polarization degree is constructed according to the corrected image polarization degree to solve the backscattering polarization degree, and the backscattering light intensity is calculated according to the backscattering polarization degree; an atmospheric light calculation function based on weight guidance is constructed according to the backscattering polarization degree, an infinite atmospheric light vector is obtained, and an enhanced underwater image is obtained through atmospheric scattering model inversion in combination with the backscattering light intensity, so that the limitation of backscattering light assumption in a traditional polarization imaging method is broken through, and the imaging accuracy is improved. And the problems of image color cast and low contrast ratio are effectively corrected.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to an underwater polarization image enhancement method based on tensor low-rank decomposition. Background Art

[0002] Polarization imaging technology has become an important tool in the field of underwater image processing due to its powerful physical properties. However, due to the scattering and absorption of particles in the water, the underwater polarization images captured still face problems such as low contrast and color distortion, which makes underwater polarization image enhancement technology face severe challenges. Most of the existing underwater polarization image enhancement methods assume that the polarization degree of the target and the backscattering are consistent, and rarely consider the inhomogeneity of the polarization information of the backscattered light, resulting in unsatisfactory image enhancement effects. Improving the accuracy of backscattering polarization information estimation and effectively removing color polarization and other issues are still issues worthy of attention in the field of underwater polarization image enhancement.

[0003] In view of this, the present invention provides an underwater polarization image enhancement method based on tensor low-rank decomposition. Summary of the invention

[0004] According to the above-mentioned deficiencies, an underwater polarization image enhancement method based on tensor low-rank decomposition is provided. The present invention mainly obtains polarization images of 0 degrees, 45 degrees, 90 degrees, and 135 degrees in the same scene, and calculates the Stokes vector and polarization degree of polarized light; on the premise of maintaining the polarization relationship between orthogonal polarization images, the four polarization images are preprocessed by histogram stretching and color correction respectively; the corrected image polarization degree and polarization angle information of the image are calculated according to the Stokes formula; based on the low-rank characteristics of the backscattered polarization degree, a tensor low-rank model of the backscattered polarization degree is constructed; the backscattered polarization degree is separated by the alternating direction multiplier method (ADMM), so as to calculate the backscattered light of the image; the atmospheric light vector at infinity is obtained by the weight-guided atmospheric light function; finally, after model inversion, the enhanced underwater image is obtained, which breaks through the limitations of the backscattered light assumption in the traditional polarization imaging method and effectively corrects the problems of image color deviation and low contrast.

[0005] The technical means adopted by the present invention are as follows:

[0006] The present invention provides an underwater polarization image enhancement method based on tensor low-rank decomposition, comprising:

[0007] Acquire a 0-degree polarization image, a 45-degree polarization image, a 90-degree polarization image, and a 135-degree polarization image of the same scene, and calculate the polarization degree of the polarized light according to the 0-degree polarization image, the 45-degree polarization image, the 90-degree polarization image, and the 135-degree polarization image;

[0008] Performing histogram stretching on the 90-degree polarization image and the 135-degree polarization image to obtain the stretched 90-degree polarization image and the stretched 135-degree polarization image; performing color correction processing on the stretched 90-degree polarization image and the stretched 135-degree polarization image and combining the polarization degree to obtain the corrected 0-degree polarization image and the corrected 45-degree polarization image;

[0009] Calculate the polarization degree of the corrected image using the Stokes formula according to the stretched 90-degree polarization image, the stretched 135-degree polarization image, the corrected 0-degree polarization image, and the corrected 45-degree polarization image;

[0010] Constructing a tensor low-rank model of backscattering polarization degree according to the corrected image polarization degree, solving the backscattering polarization degree using an alternating direction multiplier method, and calculating the backscattering light intensity according to the backscattering polarization degree;

[0011] Constructing a weight-guided atmospheric light calculation function according to the backscattering polarization degree to obtain an atmospheric light vector at infinity;

[0012] The enhanced underwater image is obtained by inverting the atmospheric light vector at infinity and the backscattered light intensity through an atmospheric scattering model.

[0013] Preferably, the calculating the polarization degree of the polarized light according to the 0-degree polarization image, the 45-degree polarization image, the 90-degree polarization image, and the 135-degree polarization image comprises:

[0014] Calculate the Stokes vector of the polarized light according to the 0-degree polarization image, the 45-degree polarization image, the 90-degree polarization image, and the 135-degree polarization image;

[0015] The polarization degree is calculated based on the Stokes vector.

[0016] Preferably, the Stokes vector of the polarized light is calculated according to the 0-degree polarization image, the 45-degree polarization image, the 90-degree polarization image, and the 135-degree polarization image, in the following manner:

[0017]

[0018] Wherein, S0 is the total light intensity, S1 is the intensity difference between the 0-degree polarization image and the 90-degree polarization image, S2 is the intensity difference between the 45-degree polarization image and the 135-degree polarization image, I0 is the 0-degree polarization image, I 45 is the 45 degree polarization image, I 90 is the 90 degree polarization image, I 135 is the 135 degree polarization image;

[0019] The polarization degree is calculated according to the Stokes vector in the following manner:

[0020]

[0021] Wherein, p is the degree of polarization of the polarized light.

[0022] Preferably, the histogram stretching of the 90-degree polarization image and the 135-degree polarization image to obtain the stretched 90-degree polarization image and the stretched 135-degree polarization image is obtained in the following manner:

[0023]

[0024] Among them, I α is the polarization image of α degree, α=90 or 135, min(I α ) is the minimum value of the elements in the matrix of the α-degree polarization image, max(I α ) is the maximum value of the elements in the matrix of the α degree polarization image, This is the α degree polarized image after stretching.

[0025] Preferably, the color correction processing is performed on the stretched 90-degree polarization image and the stretched 135-degree polarization image, and then combined with the polarization degree to obtain the corrected 0-degree polarization image and the corrected 45-degree polarization image, including:

[0026] Performing color correction processing on the stretched 90-degree polarization image and the stretched 135-degree polarization image to obtain a corrected 90-degree polarization image and a corrected 135-degree polarization image;

[0027] The orthogonal polarization maintaining coefficient is calculated based on the polarization degree, and is calculated in the following manner:

[0028]

[0029] Wherein, p is the polarization degree;

[0030] Multiplying the orthogonal polarization-maintaining coefficient with the corrected 90-degree polarization image to obtain the corrected 0-degree polarization image;

[0031] The orthogonal polarization-maintaining coefficient is multiplied by the corrected 135-degree polarization image to obtain the corrected 45-degree polarization image.

[0032] Preferably, the color correction processing is performed on the stretched 90-degree polarization image and the stretched 135-degree polarization image to obtain the corrected 90-degree polarization image and the corrected 135-degree polarization image, and the calculation is performed in the following manner:

[0033]

[0034] in, is the α-degree polarization image after correction, α=90 or 135, is the average value of the R channel of the α-polarized image after stretching, is the average value of the G channel of the α-degree polarization image after stretching, is the average value of the B channel of the α-polarized image after stretching,

[0035] The orthogonal polarization-maintaining coefficient is multiplied by the corrected 90-degree polarization image to obtain the corrected 0-degree polarization image, which is calculated in the following manner:

[0036]

[0037] in, is the corrected 0-degree polarization image, k is the orthogonal polarization-maintaining coefficient, is the corrected 90-degree polarization image;

[0038] The orthogonal polarization-maintaining coefficient is multiplied by the corrected 135-degree polarization image to obtain the corrected 45-degree polarization image, which is calculated as follows:

[0039]

[0040] in, is the 45 degree polarization image after correction, is the 135 degree polarization image after correction.

[0041] Preferably, constructing a tensor low-rank model of backscattering polarization degree according to the corrected image polarization degree, solving the backscattering polarization degree, and calculating the backscattering light intensity according to the backscattering polarization degree includes:

[0042] The tensor low-rank model of the backscattered polarization degree is calculated as follows:

[0043]

[0044] in, is the tensor form of the polarization degree of the corrected image, is the tensor form of the backscattered polarization degree, is the tensor form of the target polarization degree.

[0045] Preferably, the backscattered light intensity is calculated according to the backscattered polarization degree in the following manner:

[0046]

[0047] Wherein, B is the backscattered light intensity, P B is the backscattered polarization degree, is the corrected 0 degree polarization image, is the corrected 90-degree polarization image, is the 45 degree polarization image after correction, is the 135 degree polarization image after correction.

[0048] Preferably, the atmospheric light vector at infinity is calculated in the following manner:

[0049]

[0050] Among them, i = {R, G, B} is the three channels of the image, is the atmospheric light intensity at infinity for each of the channels, is the atmospheric light intensity for each of the channels, W i is the weight vector of the atmospheric light function, Z i The atmospheric light intensity A for each channel i The intensity of each channel of the corrected image The proportion of , ||·||1 is the l1 norm of the matrix, and ||·||0 is the zero norm of the matrix.

[0051] Preferably, the enhanced underwater image is obtained by inverting the atmospheric light vector at infinity and the backscattered light intensity through an atmospheric scattering model, and is calculated in the following manner:

[0052]

[0053] Wherein, L is the enhanced underwater image, B is the backscattered light intensity, A ∞ is the atmospheric light vector at infinity, I c is the total light intensity of the corrected image.

[0054] Compared with the prior art, the present invention has the following advantages:

[0055] The present invention provides an underwater polarization image enhancement method based on tensor low-rank decomposition, which calculates the polarization degree of polarized light by acquiring a 0-degree polarization image, a 45-degree polarization image, a 90-degree polarization image and a 135-degree polarization image of the same scene; performs histogram stretching and color correction processing on the 90-degree polarization image and the 135-degree polarization image to obtain a stretched 90-degree polarization image, a stretched 135-degree polarization image, a corrected 0-degree polarization image and a corrected 45-degree polarization image, which are used to calculate the polarization degree of the corrected image; constructs a tensor low-rank model of the backscattering polarization degree according to the polarization degree of the corrected image, adopts an alternating direction multiplier method to solve the backscattering polarization degree, and calculates the backscattering light intensity according to the backscattering polarization degree; constructs an atmospheric light calculation function based on weight guidance according to the backscattering polarization degree to obtain an atmospheric light vector at infinity; obtains an enhanced underwater image according to the atmospheric light vector and the backscattering light intensity at infinity through inversion using an atmospheric scattering model, which breaks through the limitation of the backscattering light assumption in the traditional polarization imaging method and effectively corrects the problems of image color cast and low contrast. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0057] Figure 1 The present invention provides an underwater polarization image enhancement method based on tensor low-rank decomposition.

[0058] Figure 2 It is a schematic diagram of a 0-degree polarization image, a 45-degree polarization image, a 90-degree polarization image, and a 135-degree polarization image.

[0059] Figure 3 It is a schematic diagram of a stretched 90-degree polarization image, a stretched 135-degree polarization image, a corrected 0-degree polarization image, and a corrected 45-degree polarization image.

[0060] Figure 4 It is a polarization image corresponding to the polarization degree of the corrected image.

[0061] Figure 5 is a backscattered polarization image.

[0062] Figure 6 is the intensity of backscattered light.

[0063] Figure 7 An enhanced underwater image. DETAILED DESCRIPTION

[0064] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0065] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0066] Reference Figure 1 , Figure 1 The present invention provides an underwater polarization image enhancement method based on tensor low-rank decomposition to illustrate a specific embodiment of the underwater polarization image enhancement method based on tensor low-rank decomposition provided by the present invention, including:

[0067] Obtain a 0-degree polarization image, a 45-degree polarization image, a 90-degree polarization image, and a 135-degree polarization image of the same scene, and calculate the polarization degree of the polarized light according to the 0-degree polarization image, the 45-degree polarization image, the 90-degree polarization image, and the 135-degree polarization image;

[0068] Performing histogram stretching on the 90-degree polarization image and the 135-degree polarization image to obtain a stretched 90-degree polarization image and a stretched 135-degree polarization image; performing color correction processing on the stretched 90-degree polarization image and the stretched 135-degree polarization image and combining the polarization degree to obtain a corrected 0-degree polarization image and a corrected 45-degree polarization image;

[0069] The polarization degree and polarization angle of the corrected image are calculated using the Stokes formula according to the stretched 90-degree polarization image, the stretched 135-degree polarization image, the corrected 0-degree polarization image, and the corrected 45-degree polarization image;

[0070] A tensor low-rank model of backscattered polarization is constructed based on the corrected image polarization, and the backscattered polarization is solved using the alternating direction multiplier method (ADMM). The backscattered light intensity is calculated based on the backscattered polarization.

[0071] According to the backscattering polarization degree, a weight-guided atmospheric light calculation function is constructed to obtain the atmospheric light vector at infinity.

[0072] The enhanced underwater image is obtained by inverting the atmospheric light vector and the backscattered light intensity at infinity through the atmospheric scattering model.

[0073] It should be noted that in this embodiment, a Triton Vision Phoenix TRI050S polarization camera is used to simultaneously capture polarization images of 0 degrees, 45 degrees, 90 degrees, and 135 degrees in the same scene, which are respectively denoted as I0, I1, and I2. 45 ,I 90 ,I 135 . Where I0 and I 90 , I 45 and I 135 It can be regarded as two sets of orthogonal polarization images. Of course, the method of obtaining polarization images is not limited to this. The 90-degree polarization image and the 135-degree polarization image are histogram stretched to obtain the stretched 90-degree polarization image and the stretched 135-degree polarization image; the stretched 90-degree polarization image and the stretched 135-degree polarization image are color corrected and combined with the polarization degree to obtain the corrected 0-degree polarization image and the corrected 45-degree polarization image. This process is carried out under the premise of maintaining the polarization relationship between the orthogonal polarization images. The tensor low-rank model of the backscattered polarization degree is built based on the low-rank characteristics of the backscattered polarization degree.

[0074] It can be understood that the underwater polarization image enhancement method based on tensor low-rank decomposition provided by the present invention obtains polarization images of 0 degrees, 45 degrees, 90 degrees, and 135 degrees in the same scene, and calculates the Stokes vector and polarization degree of the polarized light; while maintaining the polarization relationship between the orthogonal polarization images, the four polarization images are respectively preprocessed by histogram stretching and color correction; the corrected image polarization degree and polarization angle information of the image are calculated according to the Stokes formula; based on the low-rank characteristics of the backscattered polarization degree, a tensor low-rank model of the backscattered polarization degree is constructed; the backscattered polarization degree is separated by the alternating direction multiplier method (ADMM), so as to calculate the backscattered light of the image; the atmospheric light vector at infinity is obtained by the weight-guided atmospheric light function; finally, after model inversion, the enhanced underwater image is obtained, which breaks through the limitations of the backscattered light assumption in the traditional polarization imaging method and effectively corrects the problems of image color deviation and low contrast.

[0075] In some optional embodiments, combined with Figure 2 , Figure 2 It is a schematic diagram of a 0-degree polarization image, a 45-degree polarization image, a 90-degree polarization image, and a 135-degree polarization image. The polarization degree of polarized light is calculated according to the 0-degree polarization image, the 45-degree polarization image, the 90-degree polarization image, and the 135-degree polarization image, including:

[0076] Calculate the Stokes vector of polarized light according to the 0-degree polarization image, the 45-degree polarization image, the 90-degree polarization image, and the 135-degree polarization image;

[0077] Calculate the degree of polarization from the Stokes vector.

[0078] Specifically, the Stokes vector of the polarized light is calculated according to the 0-degree polarization image, the 45-degree polarization image, the 90-degree polarization image, and the 135-degree polarization image, and is calculated in the following manner:

[0079]

[0080] Wherein, S0 is the total light intensity, S1 is the intensity difference between the 0-degree polarization image and the 90-degree polarization image, S2 is the intensity difference between the 45-degree polarization image and the 135-degree polarization image, I0 is the 0-degree polarization image, I 45 is a 45 degree polarization image, I 90 is a 90-degree polarization image, I 135 It is a 135 degree polarization image;

[0081] The degree of polarization is calculated from the Stokes vector as follows:

[0082]

[0083] Wherein, p is the degree of polarization of polarized light.

[0084] In some optional embodiments, combined with Figure 3 , Figure 3 A schematic diagram of a stretched 90-degree polarization image, a stretched 135-degree polarization image, a corrected 0-degree polarization image, and a corrected 45-degree polarization image is shown in FIG. The 90-degree polarization image and the 135-degree polarization image are histogram stretched to obtain a stretched 90-degree polarization image and a stretched 135-degree polarization image, which are obtained in the following manner:

[0085]

[0086] Among them, I α is the polarization image of α degree, α=90 or 135, min(I α ) is the minimum value of the elements in the matrix of the α-degree polarization image, max(I α ) is the maximum value of the elements in the matrix of the α degree polarization image, This is the α-degree polarization image after stretching.

[0087] In some optional embodiments, continue to refer to Figure 3 , after color correction processing is performed on the stretched 90-degree polarization image and the stretched 135-degree polarization image, combined with the polarization degree, a corrected 0-degree polarization image and a corrected 45-degree polarization image are obtained, including:

[0088] Performing color correction processing on the stretched 90-degree polarization image and the stretched 135-degree polarization image to obtain a corrected 90-degree polarization image and a corrected 135-degree polarization image;

[0089] Calculate orthogonal polarization-maintaining coefficients based on the degree of polarization;

[0090] Multiply the orthogonal polarization-maintaining coefficient with the corrected 90-degree polarization image to obtain a corrected 0-degree polarization image;

[0091] The orthogonal polarization-maintaining coefficient is multiplied by the corrected 135-degree polarization image to obtain the corrected 45-degree polarization image.

[0092] Specifically, the stretched 90-degree polarization image and the stretched 135-degree polarization image are subjected to color correction to obtain a corrected 90-degree polarization image and a corrected 135-degree polarization image, which are calculated in the following manner:

[0093]

[0094] in, is the α-degree polarization image after correction, α=90 or 135, is the average value of the R channel of the stretched α-polarized image, is the average value of the G channel of the stretched α-polarized image, is the average value of the B channel of the stretched α-polarized image,

[0095] It can be understood that the above formula is based on the gray world algorithm to perform color correction on the stretched α-degree polarization image.

[0096] The orthogonal polarization-maintaining coefficient is multiplied by the corrected 90-degree polarization image to obtain the corrected 0-degree polarization image, which is calculated as follows:

[0097]

[0098] in, is the corrected 0-degree polarization image, k is the coefficient for maintaining orthogonal polarization, is the corrected 90-degree polarization image;

[0099] The orthogonal polarization-maintaining coefficient is multiplied by the corrected 135-degree polarization image to obtain the corrected 45-degree polarization image, which is calculated as follows:

[0100]

[0101] in, is the corrected 45 degree polarization image, This is the corrected 135 degree polarization image.

[0102] Specifically, the orthogonal polarization coefficient is calculated as follows:

[0103]

[0104] Where p is the degree of polarization.

[0105] In some optional embodiments, referring to Figure 4 , Figure 4 is a polarization image corresponding to the polarization degree of the corrected image. The polarization degree of the corrected image is calculated using the Stokes formula according to the stretched 90-degree polarization image, the stretched 135-degree polarization image, the corrected 0-degree polarization image, and the corrected 45-degree polarization image. The calculation is as follows:

[0106]

[0107] Among them, P c is the polarization degree of the corrected image, is the corrected total light intensity, is the intensity difference between the corrected 0-degree polarization image and the corrected 90-degree polarization image, is the intensity difference between the corrected 45-degree polarization image and the corrected 135-degree polarization image, and The specific calculation method of can refer to the calculation method of the Stokes vector of polarized light, and no further details will be given here.

[0108] By processing the Stokes vector, the polarization angle can also be calculated as follows:

[0109]

[0110] Among them, θ c is the polarization angle.

[0111] In some optional embodiments, referring to Figure 5 and Figure 6 , Figure 5 is a backscattered polarization image, Figure 6is a backscattered light intensity. A tensor low-rank model of the backscattered polarization degree is constructed according to the corrected image polarization degree. The backscattered polarization degree is solved and the backscattered light intensity is calculated according to the backscattered polarization degree, including:

[0112] The tensor low-rank model of the backscattered polarization degree is calculated as follows:

[0113]

[0114] in, is the tensor form of the polarization degree of the corrected image, is the tensor form of the backscattered polarization degree, is the tensor form of the target polarization degree.

[0115] It can be understood that, according to the inconsistency between the target and backscattered polarization degrees, the polarization degree P of the corrected image of polarized light is c Expressed as the backscattered polarization degree P B The linear combination of the target polarization degree G is calculated as follows:

[0116] P c =P B +G

[0117] According to the low-rank characteristics of the backscattered polarization degree, a tensor low-rank model of the initial backscattered polarization degree is constructed and calculated as follows:

[0118]

[0119] Among them, rank t (·) is the tube rank function of the tensor of backscattered polarization degree, It is the sum function of the absolute values ​​of all elements of the tensor.

[0120] Since the tensor low-rank model of the initial backscattered polarization degree is an NP-hard problem, the model replacement method is used to solve it.

[0121] Letting the nuclear norm of the tensor be a non-convex relaxation of the rank function, the initial low-rank model of the tensor of backscattered polarization is converted to a low-rank model of the tensor of backscattered polarization, calculated as follows:

[0122]

[0123] in, is the nuclear norm of the tensor of the backscattered polarization degree, is the frontal slice matrix of the backscattered polarization degree tensor, and n is the number of frontal slices of the backscattered polarization degree tensor.

[0124] In some optional embodiments, continue to refer to Figure 5 , the backscattered polarization degree is solved by the alternating direction multiplier method, including:

[0125] Step 5.1, using the alternating direction multiplier method to solve the tensor low-rank model of the backscattered polarization degree, decompose the backscattered polarization degree, and its augmented Lagrangian function is calculated as follows:

[0126]

[0127] in, is the Lagrange multiplier, β is the penalty factor, β>0, is the inner product of the tensor, is the Frobenius norm of the tensor.

[0128] Step 5.2: Initialize variables β 0 =10 -4 , β max =10 10 , ε=10 -8 , the number of iterations k = 0;

[0129] Step 5.3: Fixed variables and Then the variable The (k+1)th iteration formula is:

[0130]

[0131] Step 5.4: Fixed variables and Then the variable The (k+1)th iteration formula is:

[0132]

[0133] Step 5.5: Fix variables and Then the variable The (k+1)th iteration formula is:

[0134]

[0135] Step 5.6, let ρ = 1.1, the (k+1)th iteration formula of parameter μ is:

[0136] μ k+1 =min(ρμ k ,μ max )

[0137] Step 5.7, update the number of iterations, that is, k = k + 1; alternately perform steps 5.3 to 5.6 until the iteration stops when the iteration stop condition is met, and the tensor form of the backscattered polarization degree is obtained, that is, the backscattered polarization degree is obtained. In order to avoid the introduction of noise in subsequent calculations, the backscattered polarization degree can be Gaussian blurred. Specifically, the iteration stop condition is:

[0138]

[0139] Of course, it is not limited to this and can be set according to actual conditions.

[0140] In some optional embodiments, continue to refer to Figure 6 , the backscattered light intensity is calculated according to the backscattered polarization degree, as follows:

[0141]

[0142] Where B is the backscattered light intensity, P B is the backscattered polarization degree, is the corrected 0 degree polarization image, is the corrected 90 degree polarization image, is the corrected 45 degree polarization image, This is the corrected 135 degree polarization image.

[0143] In some optional embodiments, a weight-guided atmospheric light calculation function is constructed according to the backscattered polarization degree to obtain an atmospheric light vector at infinity, including:

[0144] From the polarization angle θ c The value with the highest frequency in each channel is selected as the polarization angle of the corresponding channel to obtain the polarization angle vector Where i = {R, G, B} represents the channel of the image.

[0145] The atmospheric light intensity is calculated by combining the polarization angle and the backscattered polarization degree as follows:

[0146]

[0147] Among them, P B is the backscattered polarization degree, is the corrected 0-degree polarization image, I c is the total light intensity of the corrected image.

[0148] Based on the atmospheric light intensity, the atmospheric light function guided by the weight is used to obtain the atmospheric light vector at infinity, which is calculated as follows:

[0149]

[0150] Among them, i = {R, G, B} is the three channels of the image, is the atmospheric light intensity at infinity for each channel, A i is the atmospheric light intensity for each channel, W i is the weight vector of the atmospheric light function, Z i The atmospheric light intensity A for each channel i The intensity of each channel of the corrected image The proportion of , ||·||1 is the l1 norm of the matrix, and ||·||0 is the zero norm of the matrix.

[0151] In some optional embodiments, referring to Figure 7 , Figure 7 is an enhanced underwater image. According to the inversion of the atmospheric light vector and the backscattered light intensity at infinity, the enhanced underwater image is obtained and calculated as follows:

[0152]

[0153] Among them, L is the enhanced underwater image, B is the backscattered light intensity, A is ∞ is the atmospheric light vector at infinity, I c is the total light intensity of the corrected image.

[0154] It can be understood that the method of this method takes into account the inhomogeneity of the polarization information of the backscattered light, utilizes the low-rank characteristics of the backscattered polarization degree, and constructs a tensor low-rank model of the backscattered polarization degree, thereby improving the accuracy of the estimation of the backscattered polarization information and effectively solving the problems of color cast and low contrast of underwater images.

[0155] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0156] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0157] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An underwater polarization image enhancement method based on tensor low-rank decomposition, characterized in that: include: Acquire a 0-degree polarization image, a 45-degree polarization image, a 90-degree polarization image, and a 135-degree polarization image of the same scene, and calculate the polarization degree of the polarized light according to the 0-degree polarization image, the 45-degree polarization image, the 90-degree polarization image, and the 135-degree polarization image; Performing histogram stretching on the 90-degree polarization image and the 135-degree polarization image to obtain the stretched 90-degree polarization image and the stretched 135-degree polarization image; performing color correction processing on the stretched 90-degree polarization image and the stretched 135-degree polarization image and combining the polarization degree to obtain the corrected 0-degree polarization image and the corrected 45-degree polarization image; Calculate the polarization degree of the corrected image using the Stokes formula according to the stretched 90-degree polarization image, the stretched 135-degree polarization image, the corrected 0-degree polarization image, and the corrected 45-degree polarization image; Constructing a tensor low-rank model of backscattering polarization degree according to the corrected image polarization degree, solving the backscattering polarization degree using an alternating direction multiplier method, and calculating the backscattering light intensity according to the backscattering polarization degree; Constructing a weight-guided atmospheric light calculation function according to the backscattering polarization degree to obtain an atmospheric light vector at infinity; The enhanced underwater image is obtained by inverting the atmospheric light vector at infinity and the backscattered light intensity through an atmospheric scattering model.

2. The underwater polarization image enhancement method based on tensor low-rank decomposition according to claim 1, characterized in that: The calculating the polarization degree of polarized light according to the 0-degree polarization image, the 45-degree polarization image, the 90-degree polarization image, and the 135-degree polarization image includes: Calculate the Stokes vector of the polarized light according to the 0-degree polarization image, the 45-degree polarization image, the 90-degree polarization image, and the 135-degree polarization image; The polarization degree is calculated based on the Stokes vector.

3. The underwater polarization image enhancement method based on tensor low-rank decomposition according to claim 2, characterized in that: The Stokes vector of the polarized light is calculated according to the 0-degree polarization image, the 45-degree polarization image, the 90-degree polarization image, and the 135-degree polarization image, and is calculated in the following manner: Wherein, S0 is the total light intensity, S1 is the intensity difference between the 0-degree polarization image and the 90-degree polarization image, S2 is the intensity difference between the 45-degree polarization image and the 135-degree polarization image, I0 is the 0-degree polarization image, I 45 is the 45 degree polarization image, I 90 is the 90 degree polarization image, I 135 is the 135 degree polarization image; The polarization degree is calculated according to the Stokes vector in the following manner: Wherein, p is the degree of polarization of the polarized light.

4. The underwater polarization image enhancement method based on tensor low-rank decomposition according to claim 1 or 2, characterized in that: The histogram stretching of the 90-degree polarization image and the 135-degree polarization image is performed to obtain the stretched 90-degree polarization image and the stretched 135-degree polarization image, which is obtained in the following manner: Among them, I α is the polarization image of α degree, α=90 or 135, min(I α ) is the minimum value of the elements in the matrix of the α-degree polarization image, max(I α ) is the maximum value of the elements in the matrix of the α degree polarization image, This is the α degree polarized image after stretching.

5. The underwater polarization image enhancement method based on tensor low-rank decomposition according to claim 1, characterized in that: The color correction processing is performed on the stretched 90-degree polarization image and the stretched 135-degree polarization image, and then combined with the polarization degree to obtain the corrected 0-degree polarization image and the corrected 45-degree polarization image, including: Performing color correction processing on the stretched 90-degree polarization image and the stretched 135-degree polarization image to obtain a corrected 90-degree polarization image and a corrected 135-degree polarization image; The orthogonal polarization maintaining coefficient is calculated based on the polarization degree, and is calculated in the following manner: Wherein, p is the polarization degree; Multiplying the orthogonal polarization-maintaining coefficient with the corrected 90-degree polarization image to obtain the corrected 0-degree polarization image; The orthogonal polarization-maintaining coefficient is multiplied by the corrected 135-degree polarization image to obtain the corrected 45-degree polarization image.

6. The underwater polarization image enhancement method based on tensor low-rank decomposition according to claim 5, characterized in that: The color correction process is performed on the stretched 90-degree polarization image and the stretched 135-degree polarization image to obtain the corrected 90-degree polarization image and the corrected 135-degree polarization image, and the calculation is performed in the following manner: in, is the α-degree polarization image after correction, α=90 or 135, is the average value of the R channel of the α-polarized image after stretching, is the average value of the G channel of the α-degree polarization image after stretching, is the average value of the B channel of the α-polarized image after stretching, The orthogonal polarization-maintaining coefficient is multiplied by the corrected 90-degree polarization image to obtain the corrected 0-degree polarization image, which is calculated in the following manner: in, is the corrected 0-degree polarization image, k is the orthogonal polarization-maintaining coefficient, is the corrected 90-degree polarization image; The orthogonal polarization-maintaining coefficient is multiplied by the corrected 135-degree polarization image to obtain the corrected 45-degree polarization image, which is calculated as follows: in, is the 45 degree polarization image after correction, is the 135 degree polarization image after correction.

7. The underwater polarization image enhancement method based on tensor low-rank decomposition according to claim 1, characterized in that: The step of constructing a tensor low-rank model of backscattered polarization degree according to the corrected image polarization degree, solving the backscattered polarization degree, and calculating the backscattered light intensity according to the backscattered polarization degree includes: The tensor low-rank model of the backscattered polarization degree is calculated as follows: in, is the tensor form of the polarization degree of the corrected image, is the tensor form of the backscattered polarization degree, is the tensor form of the target polarization degree.

8. The underwater polarization image enhancement method based on tensor low-rank decomposition according to claim 1, characterized in that: The backscattered light intensity is calculated according to the backscattered polarization degree, and is calculated in the following manner: Wherein, B is the backscattered light intensity, P B is the backscattered polarization degree, is the corrected 0 degree polarization image, is the corrected 90-degree polarization image, is the 45 degree polarization image after correction, is the 135 degree polarization image after correction.

9. The underwater polarization image enhancement method based on tensor low-rank decomposition according to claim 1, characterized in that: The atmospheric light vector at infinity is calculated as follows: Among them, i = {R, G, B} is the three channels of the image, is the atmospheric light intensity at infinity for each of the channels, is the atmospheric light intensity for each of the channels, W i is the weight vector of the atmospheric light function, Z i The atmospheric light intensity A for each channel i The light intensity of each channel of the corrected image is I i c The proportion of , ||·||1 is the l1 norm of the matrix, and ||·||0 is the zero norm of the matrix.

10. The underwater polarization image enhancement method based on tensor low-rank decomposition according to claim 1, characterized in that: The enhanced underwater image is obtained by inverting the atmospheric light vector at infinity and the backscattered light intensity through the atmospheric scattering model, and is calculated in the following manner: Wherein, L is the enhanced underwater image, B is the backscattered light intensity, A ∞ is the atmospheric light vector at infinity, I c is the total light intensity of the corrected image.

Citation Information

Cited By

  • Method and system for enhancing low-illumination color polarization image

    CN120298220A