An underwater polarization imaging method for restoring targets with non-uniform polarization characteristics

Abstract

The present invention discloses an underwater polarization imaging method for restoring a target with non-uniform polarization characteristics, comprising the following steps: step 1, obtaining multiple underwater images with different polarization directions, establishing an orthogonal decomposition model to perform orthogonal polarization decomposition on the target reflected light and backscattered light; step 2, establishing a polarization image intensity model, integrating it into the underwater imaging physical model, and establishing a restored image inversion model; step 3, performing inversion based on the underwater image contrast and the restored image inversion model to obtain the final imaging result. The present invention can directly calculate the fully polarized portion of the backscattered light, avoiding the unreasonable assumptions about the target reflected light in traditional underwater polarization imaging methods, and realizing automatic restoration of underwater images of targets with non-uniform polarization characteristics; and using low-pass filtering to perform a global estimation of the polarization degree of the backscattered light, without the need for human-computer interaction and without being restricted by the image background area, thereby expanding the application scope of underwater polarization imaging technology.

Description

Technical Field

[0001] The present invention relates to an underwater polarization imaging method, in particular to an underwater polarization imaging method for restoring a target with non-uniform polarization characteristics. Background Art

[0002] Underwater optical imaging technology has been widely used in many fields, including marine resource exploration, underwater archaeology, and marine rescue. However, due to the scattering and absorption effects of turbid media in water, enhancing underwater vision is an extremely challenging task. On the one hand, the absorption effect causes energy loss in the target's reflected light; on the other hand, the scattering effect changes the propagation direction of light and introduces backscattered light into the optical path. As a result, degraded underwater images often suffer from loss of detail, poor visibility, and low contrast, which seriously affects human perception. The main reason for the degradation of underwater image quality is the interference of backscattered light. Based on the partial polarization of backscattered light, underwater polarization imaging technology has been proven to be an effective method for restoring degraded underwater images.

[0003] However, existing underwater polarization imaging methods often assume that the polarization degree of the target's reflected light is constant. This makes them limited to single-polarization targets and incapable of recovering underwater images of targets with non-uniform polarization characteristics. Furthermore, most existing underwater polarization imaging methods are limited by background areas and human-computer interaction, making it impossible to fully automatically process underwater images of background-free areas. These issues significantly limit the practical application and scope of underwater polarization imaging technology. Summary of the Invention

[0004] Purpose of the invention: The purpose of the present invention is to propose an underwater polarization imaging method for restoring targets with non-uniform polarization characteristics, which can automatically realize underwater image restoration of targets with non-uniform polarization characteristics in the absence of background areas.

[0005] Technical solution: The present invention comprises the following steps:

[0006] Step 1: Acquire multiple underwater images with different polarization directions and establish an orthogonal decomposition model to perform orthogonal polarization decomposition on the target reflected light and backscattered light;

[0007] Step 2: Establish a polarization image intensity model and integrate it into the underwater imaging physical model to establish a restored image inversion model;

[0008] Step 3: Perform inversion based on the underwater image contrast and the restored image inversion model to obtain the final imaging result.

[0009] The step 1 specifically includes:

[0010] 1.1. Introduction of underwater imaging physical model:

[0011] I=D+B=L·t+A ∞ (1-t) (1)

[0012] Where I is the total light intensity of the underwater image; D is the target reflected light; B is the backscattered light; L is the restored image; t is the medium transmittance; A ∞ is the backscattered light intensity at infinity;

[0013] 1.2. Polarization decomposition of the target reflected light D and the backscattered light B in mutually orthogonal directions is performed, namely:

[0014] D=D ∥ +D ⊥ (2)

[0015] B=B ∥ +B ⊥ (3)

[0016] Among them, D ∥ The horizontal component of the target reflected light is decomposed into orthogonal polarization; D ⊥ The vertical component of the target reflected light is decomposed into orthogonal polarization; B ∥ B is the horizontal component of the orthogonal polarization decomposition of the backscattered light; ⊥ Decompose the vertical component into orthogonal polarizations of the backscattered light.

[0017] The step 2 specifically includes the following steps:

[0018] 2.1. Based on Malus’ law and the directional angles of the horizontal polarization components of the target reflected light and backscattered light, the polarization image intensity model is established. The intensity models corresponding to the polarization images I(0), I(45), I(90), and I(135) are:

[0019]

[0020] Among them, D ∥ The angle with the 0° direction is denoted as α; D ⊥ The angle with the 0° direction is recorded as α+90; B ∥ The angle with the 0° direction is denoted as γ; B ⊥ The angle with the 0° direction is recorded as γ+90;

[0021] 2.2. Obtaining the restored image inversion model includes the following steps:

[0022] 2.2.1. Obtain the Stokes vector based on the polarization image intensity model, namely:

[0023]

[0024] Where I is the total light intensity; Q is the light intensity difference between 0° and 90°; U is the light intensity difference between 45° and 135°; ΔD is D∥ and D ⊥ The difference is the fully polarized part of the target reflected light; ΔB is B ∥ and B ⊥ The difference, i.e. the fully polarized portion of the backscattered light;

[0025] 2.2.2. Calculate the fully polarized portion ΔB of the backscattered light using the Stokes vector obtained from formula (5), namely:

[0026]

[0027] 2.2.3 According to the definition of polarization degree, the backscattered light B can be expressed as follows:

[0028]

[0029] Among them, P scat is the polarization degree of backscattered light;

[0030] 2.2.4. Combining formulas (1), (6) and (7), establish the restored image inversion model:

[0031]

[0032] Among them, L is the restored image; A ∞ is the backscattered light intensity at infinity, which can be set as the average value of the first 0.1% pixels with the largest grayscale value in the total light intensity image; P scat is the polarization degree of backscattered light.

[0033] The backscattered light polarization degree P scat The global estimation method includes the following steps:

[0034] 2.2.4.1. Perform low-pass filtering on the four polarization images I(0), I(45), I(90) and I(135), namely:

[0035]

[0036] Wherein, B(0), B(45), B(90) and B(135) are the estimated backscattered light in I(0), I(45), I(90) and I(135); LPF{·} is low-pass filtering;

[0037] 2.2.4.2. Obtain the Stokes vector of the backscattered light estimated from the polarization image above, i.e.:

[0038]

[0039] Among them, I B is the estimated backscattered light intensity; QB is the light intensity difference between 0° and 90°; U B The light intensity difference between 45° and 135° directions;

[0040] 2.2.4.3. Perform a global estimate of the polarization degree of the backscattered light, namely:

[0041]

[0042] in, is the globally estimated degree of polarization of the backscattered light.

[0043] The backscattered light estimated by the low-pass filtering is estimated based on the mutual information function, and the following expression is used to improve the accuracy of the backscattered light estimation, namely:

[0044]

[0045] Among them, B optimal is the backscattered light estimated through mutual information optimization; MI(D,B) is the mutual information, which is used to characterize the correlation between the target reflected light image and the backscattered light image.

[0046] The final imaging result is inverted using the direction angle of the horizontal polarization component of the target reflected light and the direction angle of the horizontal polarization component of the backscattered light as independent variables.

[0047] The horizontal polarization component angle of the target reflected light and the horizontal polarization component angle of the backscattered light are α and γ in step 2 respectively. The method for obtaining them is: taking the underwater image contrast as the objective function, obtaining the optimal solution that satisfies the maximum underwater image contrast, that is:

[0048] (α,γ) optimal =argmax{UIConM(L)} (13)

[0049] Among them, UIConM(L) is the underwater image contrast of the restored image.

[0050] Beneficial effects: The present invention can directly calculate the fully polarized portion of backscattered light, avoiding the unreasonable assumptions about target reflected light in traditional underwater polarization imaging methods, and realizing automatic restoration of underwater images of targets with non-uniform polarization characteristics; and uses low-pass filtering to perform a global estimation of the polarization degree of backscattered light, without the need for human-computer interaction and without being restricted by the image background area, thus expanding the application scope of underwater polarization imaging technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 is a flow chart of the present invention;

[0052] Figure 2 This is a model diagram of the polarization imaging system in the present invention that simulates an underwater environment;

[0053] Figure 3 Schematic diagram of polarization decomposition direction of the present invention;

[0054] Figure 4 (a) is the total light intensity image of the background-free area; Figure 4 (b) Schematic diagram of the effect of practical application of the present invention. DETAILED DESCRIPTION

[0055] The present invention will be further described below with reference to the accompanying drawings.

[0056] like Figure 1 As shown, the imaging method of the present invention comprises the following steps:

[0057] Step 1: Acquire multiple underwater images with different polarization directions and establish an orthogonal decomposition model to perform orthogonal polarization decomposition on the target reflected light and backscattered light.

[0058] like Figure 2 The figure shows a polarization imaging system of the present invention in a simulated underwater environment. An initial beam is emitted by an LED white light source 1. The initial beam passes through a first polarizer 2 with a horizontal polarization direction, a transparent water tank 5, and turbid water 6, irradiating a target 4 and suspended particles 3. The turbid water 6 is made by adding 15ml of milk to clear water. The target 4 is composed of plastic and metal materials. In this embodiment, a plastic Rubik's Cube with a metal coin attached to it is used. The metal coin with high polarization properties and the plastic Rubik's Cube with low polarization properties together form a non-uniform polarization target. The target 4 reflects the initial beam, generating target reflected light. The suspended particles 3 scatter the initial beam, generating backscattered light. The target reflected light and backscattered light pass through a second polarizer 7 and are irradiated by a CCD detector 8.

[0059] The total light intensity received by the CCD detector has the following relationship with the target reflected light and backscattered light, which is the classic underwater imaging physical model:

[0060] I=D+B=L·t+A ∞ (1-t) (1)

[0061] Where I is the total light intensity of the underwater image; D is the target reflected light; B is the backscattered light; L is the restored image; t is the medium transmittance; A ∞ is the backscattered light intensity at infinity.

[0062] The polarization direction of the second polarizer 7 is rotated to 0° to obtain a polarization image I(0); the polarization direction of the second polarizer 7 is rotated to 45° to obtain a polarization image I(45); the polarization direction of the second polarizer 7 is rotated to 90° to obtain a polarization image I(90); the polarization direction of the second polarizer 7 is rotated to 135° to obtain a polarization image I(135).

[0063] like Figure 3 The figure shows the polarization decomposition direction of the present invention. The target reflected light D and the backscattered light B can be polarized in mutually orthogonal directions, that is:

[0064] D=D ∥ +D ⊥ (2)

[0065] B=B ∥ +B ⊥ (3)

[0066] Among them, D ∥ The horizontal component of the target reflected light is decomposed into orthogonal polarization; D ⊥ The vertical component of the target reflected light is decomposed into orthogonal polarization; B ∥ B is the horizontal component of the orthogonal polarization decomposition of the backscattered light; ⊥ Decompose the vertical component into orthogonal polarizations of the backscattered light.

[0067] Step 2: Based on Malus's law and the azimuth angles of the horizontal polarization components of the target's reflected and backscattered light, a polarization image intensity model is established, which is then integrated into the classical underwater imaging physics model to establish an image restoration inversion model. This includes the following steps:

[0068] 2.1. According to Malus’ law and the directional angles of the horizontal polarization components of the target reflected light and backscattered light, the intensity models corresponding to the four polarization images I(0), I(45), I(90), and I(135) are:

[0069]

[0070] Among them, D ∥ The angle with the 0° direction is denoted as α; D ⊥ The angle with the 0° direction is recorded as α+90; B ∥ The angle with the 0° direction is denoted as γ; B ⊥ The angle with the 0° direction is recorded as γ+90.

[0071] 2.2. Obtaining the restored image inversion model includes the following steps:

[0072] 2.2.1. Obtain the Stokes vector based on the polarization image intensity model, namely:

[0073]

[0074] Where I is the total light intensity; Q is the light intensity difference between 0° and 90°; U is the light intensity difference between 45° and 135°; ΔD is D ∥ and D ⊥ The difference is the fully polarized part of the target reflected light; ΔB is B∥ and B ⊥ The difference is the fully polarized portion of the backscattered light.

[0075] 2.2.2. Calculate the fully polarized portion ΔB of the backscattered light using the Stokes vector obtained from formula (5), namely:

[0076]

[0077] 2.2.3 According to the definition of polarization degree, the backscattered light B can be expressed as follows:

[0078]

[0079] Among them, P scat is the polarization degree of backscattered light, which can be globally estimated through low-pass filtering technology and Stokes vector calculation.

[0080] 2.2.4. Combining formulas (1), (6) and (7), establish the restored image inversion model:

[0081]

[0082] Among them, L is the restored image; A ∞ is the backscattered light intensity at infinity, which can be set as the average value of the first 0.1% pixels with the largest grayscale value in the total light intensity image. In this embodiment, A ∞ is 0.9875; P scat is the polarization degree of backscattered light.

[0083] Backscattered light polarization degree P scat The global estimation method includes the following steps:

[0084] 2.2.4.1. Perform low-pass filtering on the four polarization images I(0), I(45), I(90) and I(135), namely:

[0085]

[0086] Wherein, B(0), B(45), B(90), and B(135) are the backscattered light estimated in I(0), I(45), I(90), and I(135); LPF{·} is a low-pass filtering process. Specifically, this embodiment adopts Gaussian low-pass filtering.

[0087] 2.2.4.2. Calculate the Stokes vector of the backscattered light estimated from the four polarization images above, that is:

[0088]

[0089] Among them, I Bis the estimated backscattered light intensity; Q B is the light intensity difference between 0° and 90°; U B It is the light intensity difference between 45° and 135°.

[0090] 2.2.4.3. Perform a global estimate of the polarization degree of the backscattered light, namely:

[0091]

[0092] in, is the globally estimated degree of polarization of the backscattered light.

[0093] In order to make the backscattered light estimated by low-pass filtering more accurate, the following formula can be used to improve the accuracy of the backscattered light estimated by low-pass filtering based on the function of mutual information, namely:

[0094]

[0095] Among them, B optimal is the backscattered light estimated through mutual information optimization; MI(D,B) is the mutual information, which is used to characterize the correlation between the target reflected light image and the backscattered light image. It is greater than 0. Therefore, when the mutual information is smaller, that is, closer to 0, the better the separation effect of the target reflected light and backscattered light. The mutual information is obtained by combining the probability distribution function and the marginal distribution function, and its expression is:

[0096]

[0097] Among them, MI(D,B) is the mutual information; prob(d,b) represents the joint probability distribution function; prob(b) is the second edge probability distribution function; prob(d) is the first edge probability distribution function. The first edge probability distribution function is obtained according to the grayscale value of the target reflected light image, and the second edge probability distribution function is obtained according to the grayscale value corresponding to the backscattered light image; the joint probability distribution function is obtained according to the grayscale value corresponding to the target reflected light image and the grayscale value corresponding to the backscattered light image. Specifically, the first edge probability distribution function and the second edge probability distribution function can be obtained through the grayscale histogram. For example, the horizontal axis represents the grayscale value, which ranges from 0 to 255, and the vertical axis represents the number of pixels in the image with a fixed grayscale value. When the grayscale value is 120, the number of pixels is 300, indicating that the number of pixels with a grayscale value of 120 in the image is 300. Specifically,

[0098] The first marginal probability distribution function can be expressed as:

[0099]

[0100] Wherein, d(i) represents the data obtained by dividing the number of fixed-value pixels of the target reflected light image by the total number of pixels.

[0101] The second marginal probability distribution function can be expressed as:

[0102]

[0103] Wherein, b(i) represents the data obtained by dividing the number of fixed-value pixels of the backscattered light image by the total number of pixels.

[0104] Similarly, the expression of the joint probability distribution function is:

[0105]

[0106] Where prob(d,b) represents the joint probability distribution function.

[0107] Step 3: Based on the underwater image contrast and the restored image inversion model, the horizontal polarization component angle of the target reflected light and the horizontal polarization component angle of the backscattered light are used as independent variables to perform inversion to obtain the final imaging result.

[0108] The independent variables are the α and γ in step 2, respectively, and the method for obtaining them is to use the underwater image contrast as the objective function and obtain the optimal solution that satisfies the maximum underwater image contrast, that is:

[0109] (α,γ) optimal =argmax{UIConM(L)} (17)

[0110] Among them, UIConM(L) is the underwater image contrast of the restored image.

[0111] Specifically, the underwater image contrast is obtained by the following formula:

[0112]

[0113] Among them, the restored image L is divided into k1×k2 blocks numbered (k, l); Θ, is the parameterized logarithmic image processing symbol; L max,k,l and L min,k,l are the maximum and minimum values ​​in each block numbered (k, l).

[0114] Specifically, this embodiment uses a genetic algorithm to obtain the optimal solution, sets the population size to 50, the maximum genetic generation to 200, the crossover probability to 0.6, and the mutation probability to 0.01, and finally gives a solution value of (-0.3720, 0.3567) (rad).

[0115] The polarization degree of the backscattered light obtained in step 2 Backscattered light intensity A at infinity ∞ And the above optimal solution (α,γ) optimal Substitute into formula (8) to restore the image inversion model and obtain the final imaging result.

[0116] In order to verify the effectiveness of the present invention, the target image without background area is obtained by using image segmentation technology, such as Figure 4 (a) is the total light intensity image in turbid water. The image is not very clear to the naked eye and is covered with a layer of "mist". After being processed by the method of the present invention, the image clarity and visibility are significantly improved. Figure 4 (b), and the method of the present invention is not limited by the background area and human-computer interaction, and can achieve underwater image restoration of targets with non-uniform polarization characteristics.

[0117] To quantitatively evaluate image quality, in addition to the previously cited underwater image contrast, we also used standard deviation, information entropy, and mean gradient to evaluate the quality of the restored image. Larger values ​​indicate higher image quality. The results are shown in the following table:

[0118] Table 1 Quality evaluation of restored images using underwater image contrast, standard deviation, information entropy, and average gradient

[0119] contrast Underwater image contrast Standard deviation Information entropy Average gradient Total light intensity diagram 0.1719 0.1034 5.4066 0.0083 The present invention 0.3313 0.1348 6.3075 0.0281

[0120] As can be seen from the table, compared with the original total light intensity image, the restored image of the present invention has significantly improved various objective evaluation indicators, verifying the superior descattering effect of the present invention.

Claims

1. An underwater polarization imaging method for restoring a target with non-uniform polarization characteristics, characterized in that: The following steps are involved: Step 1: Acquire multiple underwater images with different polarization directions and establish an orthogonal decomposition model to perform orthogonal polarization decomposition of the target reflected light and backscattered light. Specifically, the model includes: 1.

1. Introduction of underwater imaging physical model: I=D+B=L·t+A ∞ (1-t) (1) Where I is the total light intensity of the underwater image; D is the target reflected light; B is the backscattered light; L is the restored image; t is the medium transmittance; A ∞ is the backscattered light intensity at infinity; 1.

2. Polarization decomposition of the target reflected light D and the backscattered light B is performed in mutually orthogonal directions, that is: D=D ∥ +D ⊥ (2) B=B ∥ +B ⊥ (3) Among them, D ∥ The horizontal component of the target reflected light is decomposed into orthogonal polarization; D ⊥ The vertical component of the target reflected light is decomposed into orthogonal polarization; B ∥ B is the horizontal component of the orthogonal polarization decomposition of the backscattered light; ⊥ Decompose the vertical component into orthogonal polarizations of the backscattered light; Step 2: Establish a polarization image intensity model and integrate it into the underwater imaging physical model to establish a restored image inversion model. The specific steps include: 2.

1. Based on Malus’ law and the directional angles of the horizontal polarization components of the target reflected light and backscattered light, the polarization image intensity model is established. The intensity models corresponding to the polarization images I(0), I(45), I(90), and I(135) are: Among them, D ∥ The angle with the 0° direction is denoted as α; D ⊥ The angle with the 0° direction is recorded as α+90; B ∥ The angle with the 0° direction is denoted as γ; B ⊥ The angle with the 0° direction is recorded as γ+90; 2.

2. Obtaining the restored image inversion model includes the following steps: 2.2.

1. Obtain the Stokes vector based on the polarization image intensity model, namely: Where I is the total light intensity; Q is the light intensity difference between 0° and 90°; U is the light intensity difference between 45° and 135°; ΔD is D ∥ and D ⊥ The difference is the fully polarized part of the target reflected light; ΔB is B ∥ and B ⊥ The difference, i.e. the fully polarized portion of the backscattered light; 2.2.

2. Calculate the fully polarized portion ΔB of the backscattered light using the Stokes vector obtained from formula (5), namely: 2.2.3 According to the definition of polarization degree, the backscattered light B can be expressed as follows: Among them, P scat is the polarization degree of backscattered light; 2.2.

4. Combining formulas (1), (6) and (7), establish the restored image inversion model: Among them, L is the restored image; A ∞ is the backscattered light intensity at infinity; P scat is the polarization degree of backscattered light; Step 3: Perform inversion based on the underwater image contrast and the restored image inversion model to obtain the final imaging result.

2. The underwater polarization imaging method for restoring a target with non-uniform polarization characteristics according to claim 1, characterized in that: The backscattered light polarization degree P scat The global estimation method includes the following steps: 2.2.4.

1. Perform low-pass filtering on the four polarization images I(0), I(45), I(90) and I(135), namely: Wherein, B(0), B(45), B(90) and B(135) are the estimated backscattered light in I(0), I(45), I(90) and I(135); LPF{·} is low-pass filtering; 2.2.4.

2. Obtain the Stokes vector of the backscattered light estimated by the low-pass filter, i.e.: Among them, I B is the estimated backscattered light intensity; Q B is the light intensity difference between 0° and 90°; U B The light intensity difference between 45° and 135° directions; 2.2.4.

3. Perform a global estimate of the polarization degree of the backscattered light, namely: in, is the globally estimated degree of polarization of the backscattered light.

3. The underwater polarization imaging method for restoring a target with non-uniform polarization characteristics according to claim 2, characterized in that: The backscattered light estimated by the low-pass filtering is estimated based on the mutual information function, and the following expression is used to improve the accuracy of the backscattered light estimation, namely: Among them, B optimal is the backscattered light estimated through mutual information optimization; MI(D,B) is the mutual information, which is used to characterize the correlation between the target reflected light image and the backscattered light image.

4. The underwater polarization imaging method for restoring a target with non-uniform polarization characteristics according to claim 1, characterized in that: The final imaging result is inverted using the direction angle of the horizontal polarization component of the target reflected light and the direction angle of the horizontal polarization component of the backscattered light as independent variables.

5. The underwater polarization imaging method for restoring a target with non-uniform polarization characteristics according to claim 4, characterized in that: The horizontal polarization component angle of the target reflected light and the horizontal polarization component angle of the backscattered light are α and γ in step 2 respectively. The method for obtaining them is: taking the underwater image contrast as the objective function, obtaining the optimal solution that satisfies the maximum underwater image contrast, that is: (α,γ) optimal =argmax{UIConM(L)} (13) Wherein, UIConM(L) is the underwater image contrast of the restored image.

Images