Infrared Polarization Noise Suppression and Edge Enhancement Method Based on Gradient Vector Analysis

Abstract

The present invention belongs to the technical field of image processing, and discloses an infrared polarization noise suppression and edge enhancement method based on gradient vector analysis, including the following steps: According to the spatial distribution similarity in the polarization image, based on the double-polarization spatial neighborhood gradient weighting method, the polarization degree image is preliminarily used to suppress high-brightness noise; Utilizing the difference in polarization characteristics of small targets in the sky background, two-way feature coherence suppression is proposed to obtain significant targets; To prevent the problem of excessive suppression of target edge information during background suppression, based on the connected statistical analysis method of the gradient vector field, the target edge information is enhanced. The present invention adopts the above infrared polarization noise suppression and edge enhancement method based on gradient vector analysis, preliminarily suppresses the background high-brightness noise of the polarization degree image based on the double-polarization spatial neighborhood gradient weighting method; effectively obtains significant targets based on two-way feature coherence suppression; enhances the shape information of the target based on the connected statistical analysis method of the gradient vector field.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to an infrared polarization noise suppression and edge enhancement method based on gradient vector analysis. Background Art

[0002] Since the infrared intensity image is vulnerable to external heat source interference, which affects the image quality. And when the temperature difference between the target and the background is not obvious, the intensity contrast between the target and the background is low, resulting in difficult target detection. In addition, with the development of infrared camouflage and interference technologies, the difficulty of infrared detection has been further increased. With the development of polarization detection technology, polarization imaging detection technology has become an important detection means. Polarization describes the regularity of the vibration direction of light waves. When all objects with a temperature higher than absolute zero emit electromagnetic radiation, the asymmetry of the electromagnetic radiation will generate infrared polarization characteristics. Since the infrared polarization characteristics are not affected by external light changes and are only related to their own physical properties. Therefore, infrared polarization images can better distinguish targets and backgrounds by capturing the polarization information of light. The research on infrared polarization images has good application prospects in the fields of military target detection, environmental monitoring, and industrial detection. However, the basic parameters in polarization are calculated from the measured intensity through a non-linear operator, which easily amplifies the noise in the intensity measurement, meaning that the polarization information and the target are easily submerged in the noise, affecting the performance of polarization imaging. Especially for shaped targets, when using infrared polarization information to suppress background noise, it is easy to break the edge information of the target, resulting in a decline in the detection quality of small targets. Therefore, researching to suppress the background noise of shaped targets to improve the image quality is an urgent problem to be solved currently.

[0003] In order to enhance the edge information of small targets while suppressing background noise, many related algorithms have been proposed by researchers. These algorithms already include the step of denoising during the process of infrared polarization image enhancement. Since noise and image structures (including edges and details) belong to the same frequency component, the enhancement and denoising algorithms can be divided into two categories: traditional enhancement and denoising, and deep learning network enhancement and denoising. In traditional enhancement and denoising, mostly the input image is decomposed in the frequency domain and the spatial domain: the frequency domain is decomposed into low-frequency and high-frequency images; the spatial domain is decomposed into a detail layer (top layer) and a base layer (other layers). Then different strategies are adopted to process the decomposed images to achieve the purpose of suppressing noise and enhancing edge information. Since noise also belongs to high-frequency information, in the high-frequency information, in order to maintain the image structure while denoising, a neighborhood weighting method is used to design the contribution weights of adjacent pixels to the central pixel. In deep learning network enhancement, relevant features of the input image are deeply learned according to the convolutional neural network architecture, and then enhancement and denoising are performed. Commonly used network architectures include those based on autoencoders, Resnet, and generative adversarial network GAN, etc.

[0004] Although the above algorithm has certain effects in denoising and enhancement, the noise of small targets in infrared polarization images is still a challenge. Because the application premise of the above algorithm is that the target can be seen in the infrared intensity map, but when the target and the background have low contrast, the intensity information is not applicable. In addition, although the target is visible in the degree of polarization image, the target is small and easily submerged by noise, thus further increasing the difficulty of denoising. Summary of the Invention

[0005] The object of the present invention is to provide an infrared polarization noise suppression and edge enhancement method based on gradient vector analysis. When the target and the background have low contrast and the target information is submerged in the background noise, the infrared polarization characteristics difference of the target and the background in the gradient vector characteristics is used to suppress the noise and enhance the edge of the small target, so as to distinguish the target and the background.

[0006] To achieve the above object, the present invention provides an infrared polarization noise suppression and edge enhancement method based on gradient vector analysis, including the following steps:

[0007] Step S1: First, based on the spatial distribution similarity in the polarization image and the double-polarization spatial neighborhood gradient weighting method, the high-brightness point noise of the degree of polarization image is preliminarily suppressed;

[0008] Step S2: Secondly, using the difference in polarization characteristics of small targets in the sky background, two-way feature coherence suppression is proposed to obtain significant targets;

[0009] Step S3: Finally, to prevent the problem of over-suppressing the target edge information during the background suppression process, based on the connected statistical analysis method of the gradient vector field, the target edge information is enhanced.

[0010] Preferably, in step S1, based on the spatial distribution similarity in the polarization image and the double-polarization spatial neighborhood gradient weighting method, the high-brightness point noise of the degree of polarization image is preliminarily suppressed, and the specific process is as follows:

[0011] Step S11: In the polarization imaging system, the Stokes vector is used to represent the polarization state of light, and the general definition of the Stokes vector is as follows:

[0012] S = ( I, Q, U, V ) T ;

[0013] Among them, S represents the Stokes vector matrix; I represents the total radiation intensity of light; Q represents the difference in the linear polarization intensity of light in the horizontal and vertical directions; U represents the difference in the linear polarization intensity of light in the +45° and -45° directions; V represents the circular polarization characteristic of light, that is, the intensity difference between left-handed circular polarization and right-handed circular polarization; T represents matrix transpose;

[0014] In the natural environment, the V component of circularly polarized light is very small and can be ignored. Therefore, the Stokes vector degenerates into a third-order matrix as follows:

[0015] S = ( I, Q, U ) T ;

[0016] Step S12: Select 0°, 60°, and 120° to obtain the polarization intensities of the object I0, I 60 and I 120 and represent the degree of polarization of light using Stokes parameters; among them, the Stokes vector is as follows:

[0017]

[0018] Step S13: In the Stokes parameters, Q and U describe the polarization information in different directions on the target surface. Due to their similar spatial distribution, perform spatial neighborhood gradient weighting on two different polarization images;

[0019] Step S14: According to the spatial neighborhood gradients of the obtained polarization images Q and U, perform fusion through linear weighting to obtain the Fimage image as follows:

[0020]

[0021] where Fimage represents the result of spatial neighborhood gradient weighted fusion; G Q_t represents the weighted gradient of image Q; G U_t represents the weighted gradient of image U;

[0022] Step S15: Use the Fimage image to replace the original polarization image to perform preliminary suppression on the degree of polarization image to obtain the FDoP image as follows:

[0023]

[0024] where FDoP represents the result of processing by the dual-polarization spatial neighborhood gradient weighting method.

[0025] Preferably, in step S12, the degree of polarization calculation formula is as follows:

[0026]

[0027] where I pol represents the intensity of the completely polarized component; I total represents the total intensity in the light vector.

[0028] Preferably, in step S13, the calculation formula for weighted spatial neighborhood gradient is as follows:

[0029] G t =G ν +G h ;

[0030] G h =|X(x,y) - X(x,y - 1)|+|X(x,y) - X(x,y + 1)|;

[0031] G ν =|X(x,y) - X(x - 1,y)|+|X(x,y) - X(x + 1,y)|;

[0032]

[0033] where, ( x,y ) are the coordinates of pixels in the image; G in the image t is the weighted gradient of pixel ( x,y ) and consists of two directions; G h is the gradient in the horizontal direction; G v is the gradient in the vertical direction; | W | is the number of pixels in the neighborhood window; g ( x 1) is the Gaussian weight function; f ( x + x1 and f(x) are the gray values of the pixel in the neighborhood and the central pixel respectively.

[0034] Preferably, in step S2, by utilizing the difference in polarization characteristics of small targets in the sky background, two-way feature coherence suppression is proposed to obtain significant targets. The specific process is as follows:

[0035] Step S21: First, remove the average direction of the columns of the Fimage image; if the obtained column direction value is less than 0, it indicates the background, and set its gray value to 0; otherwise, keep the original gray value;

[0036] Step S22: Remove the average direction of the rows of the Fimage image again; if the obtained row direction value is less than 0, it indicates the background, and set its gray value to 0; otherwise, keep the original gray value;

[0037] Step S23: Obtain the CR_image image according to the two-way average direction operations in step S21 and step S22;

[0038] Among them, the calculation formula for two-way polarization coherence suppression is as follows:

[0039]

[0040] Among them, C_Fimage represents the result of removing the average direction of its columns from the original image; CR_image represents the result of two-way feature coherence suppression; Fimage i,j represents the result of spatial neighborhood gradient weighted fusion at the position (i,j); N represents the total number of column pixels; k represents the number of pixels; M represents the total number of row pixels; C_Fimage k,j represents at the position ( k,j ) the result of removing the average direction of its columns from the original image at.

[0041] Preferably, in step S3, based on the connected statistical analysis method of the gradient vector field, target edge information enhancement is performed. The specific process is as follows:

[0042] Step S31: Calculate the gradient vector field and analyze the characteristics of small targets in the gradient vector field;

[0043] Step S32: Estimate the gradient in this area based on all pixel points within a local window to obtain the horizontal gradient and vertical gradient of the image;

[0044] Step S33: Use the least squares method to fit the gray values within the local window, fit to obtain the light intensity change at a point in any direction, and obtain the modulus and direction of the gradient;

[0045] Step S34: By calculating the gradient vector field, use the connected component to analyze the significant area of the gradient change in the image.

[0046] Preferably, in step S31, each point in the gradient vector field represents the gradient of the corresponding point in the original image. The gradient calculation formula is as follows:

[0047]

[0048] Among them, represents the gradient vector of image I at the pixel point ( i,j ) ; and respectively represent the horizontal gradient and vertical gradient of the image at the pixel ( i,j ) point.

[0049] Preferably, in step S32, based on all pixel points within a local window to estimate the gradient in this area, the specific process is as follows:

[0050] Within a local window, use a quadratic polynomial to represent the gray value of the image, as follows:

[0051] I ( x, y ) ≈ a0 + a1x + a2y + a3x 2 + a4xy + a5y 2 ;

[0052] Wherein, the horizontal gradient of the image at the point (x, y) and the vertical gradient are as follows:

[0053]

[0054] Wherein, a1, a2, a3, a4, a5 represent the estimated coefficients in the polynomial.

[0055] Preferably, in step S33, the least squares method is used to fit the gray values within the local window, and the light intensity change at any point in any direction is obtained by fitting. The modulus and direction of the gradient are as follows:

[0056]

[0057] Wherein, M ( x, y ) represents the modulus of the gradient; D ( x, y ) represents the direction of the gradient.

[0058] Preferably, in step S34, by calculating the gradient vector field, the significant regions of the gradient change in the image are analyzed using connected component analysis. The specific process is as follows:

[0059] Step S341: Use the binary image obtained by thresholding to mark the pixels with gradient values greater than the threshold as the significant change regions, and mark the other pixels as the background; wherein, the binary image formula is as follows:

[0060]

[0061] Wherein, Binary Map ( x, y ) represents the binary image;

[0062] Step S342: Perform connected component labeling on the binary image and count the number; the connected component labeling is to calculate the area and centroid position of each connected component. The area of each connected component is the number of pixels it contains; wherein, the calculation formulas for the area and centroid position are as follows:

[0063] Area ( k ) = ∑ x,y 1 ( label( x, y ) = k ) ;

[0064]

[0065] where Area ( k ) represents the area of the connected component; label ( x, y ) represents the label at the position (x, y); Centroid ( k ) represents the centroid position of the connected component; 1(·) is an indicator function that is 1 when the condition is true and 0 otherwise; k represents a substitution coefficient;

[0066] Step S343: Count the number of connected regions through connected component labeling; The expression for the number of connected components is as follows:

[0067]

[0068] where Num_D represents the number of connected components; num labels represents the number of connected component labels.

[0069] Therefore, the present invention adopts the above infrared polarization noise suppression and edge enhancement method based on gradient vector analysis, and the beneficial effects are as follows:

[0070] (1) The present invention proposes a double polarization space neighborhood gradient weighting method to preliminarily suppress the background high-brightness noise in the polarization degree image;

[0071] (2) The present invention proposes two-way feature coherence suppression to effectively obtain significant targets;

[0072] (3) The present invention proposes a connected component statistical analysis method based on the gradient vector field to enhance the shape information of the target.

[0073] Next, through the drawings and embodiments, the technical solution of the present invention will be further described in detail. Description of the Drawings

[0074] Figure 1 is a flowchart of the infrared polarization noise suppression and edge enhancement method based on gradient vector analysis of the present invention;

[0075] Figure 2 is the overall framework diagram of the infrared polarization noise suppression and edge enhancement method based on gradient vector analysis of the present invention;

[0076] Figure 3 is the two-way feature coherence suppression framework diagram of the present invention;

[0077] Figure 4 It is a visualization graph of the connected component statistics of the present invention;

[0078] Figure 5 It is the final result graph after the processing of the present invention. Detailed implementation manners

[0079] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0080] As Figure 1 and Figure 2 shown, the infrared polarization noise suppression and edge enhancement method based on gradient vector analysis includes the following steps:

[0081] Step S1: First, based on the spatial distribution similarity in the polarization image and the double-polarization spatial neighborhood gradient weighting method, the polarization degree image is preliminarily used to suppress the high-brightness noise;

[0082] Step S2: Secondly, by using the difference in the polarization characteristics of small targets in the sky background, bidirectional feature coherence suppression is proposed to obtain significant targets;

[0083] Step S3: Finally, to prevent the problem of over-suppressing the target edge information during the background suppression process, based on the connected statistical analysis method of the gradient vector field, the target edge information is enhanced.

[0084] Embodiment

[0085] Step S1: First, based on the spatial distribution similarity in the polarization image and the double-polarization spatial neighborhood gradient weighting method, the polarization degree image is preliminarily used to suppress the high-brightness noise.

[0086] Step S11: In a polarization imaging system, the Stokes vector is usually used to represent the polarization state of light. The general definition of the Stokes vector is as follows:

[0087] S = (I, Q, U, V) T ;

[0088] where S represents the Stokes vector matrix; I represents the total radiation intensity of light; Q represents the difference in the linear polarization intensity of light in the horizontal and vertical directions; U represents the difference in the linear polarization intensity of light in the +45° and -45° directions; V represents the circular polarization characteristic of light, that is, the intensity difference between left-handed circular polarization and right-handed circular polarization; T represents matrix transpose.

[0089] In the natural environment, the V component of circularly polarized light is very small and can be ignored. Therefore, the Stokes vector degenerates into a third-order matrix, as follows:

[0090] S = (I, Q, U) T ;

[0091] Step S12: Select 0°, 60°, and 120° to obtain the polarization intensities of object I0, I 60 and I 120 , that is, the Stokes vector, as follows:

[0092]

[0093] The degree of polarization of light is represented by Stokes parameters. The physical meaning of the degree of polarization is the ratio of the intensity of the completely polarized component in the light vector to the total intensity in the light vector. The calculation formula for the degree of polarization is as follows:

[0094]

[0095] where I pol represents the intensity of the completely polarized component; I total represents the total intensity in the light vector.

[0096] Step S13: Among the Stokes parameters, Q and U describe the polarization information in different directions on the target surface. Due to their spatial distribution similarity, spatial neighborhood gradient weighting (BNGW) is performed on two different polarization images. The calculation formula for spatial neighborhood gradient weighting is as follows:

[0097] G t = G ν + G h ;

[0098] G h = |X ( x,y ) - X ( x,y - 1 ) | + |X ( x,y ) - X ( x,y + 1 ) |;

[0099] G ν = |X ( x,y ) - X ( x - 1,y ) | + |X ( x,y ) - X ( x + 1,y ) |;

[0100]

[0101] where ( x,y ) are the coordinates of the pixels in the image; G tis a pixel ( x,y ) is the weighted gradient, consisting of two directions; G h is the gradient in the horizontal direction; G v is the gradient in the vertical direction; | W | is the number of pixels in the neighborhood window; g ( x 1) is the Gaussian weight function; f ( x + x1 and f(x) are the gray values of the pixel in the neighborhood and the central pixel respectively.

[0102] Step S14: Obtain the spatial neighborhood gradients of the polarization images Q and U respectively through the above formula, and fuse them through linear weighting to obtain the Fimage image, as follows:

[0103]

[0104] Among them, Fimage represents the result of weighted fusion of spatial neighborhood gradients; G Q_t represents the weighted gradient of image Q; G U_t represents the weighted gradient of image U.

[0105] Step S15: Use the Fimage image to replace the original polarization image to preliminarily suppress the degree of polarization image and obtain the FDoP image, as follows:

[0106]

[0107] Among them, FDoP represents the result of processing by the dual-polarization spatial neighborhood gradient weighting method.

[0108] Step S2: Secondly, utilize the difference in the polarization characteristics of small targets in the sky background to propose two-way feature coherent suppression to obtain significant targets.

[0109] Since the target and the background have different polarization characteristics, the background has low polarization characteristics for the target, the target has high polarization characteristics, and moreover, the Fimage obtained according to the dual-polarization spatial gradient weighting has polarization information in different directions.

[0110] Therefore, the present invention proposes two-way feature coherent suppression to effectively obtain significant targets, and the specific process is as follows:

[0111] Step S21: First, since the target only accounts for a small part of the image with respect to the background, the average direction of the columns of the Fimage image is removed. If the obtained column direction value is less than 0, it indicates the background, and its gray value is set to 0; otherwise, the original gray value is maintained.

[0112] Step S22. After that, remove the average direction of the rows of the Fimage image again. If the obtained row direction value is less than 0, it indicates the background, and its grayscale value is set to 0; otherwise, the original grayscale value is maintained.

[0113] Step S23. Finally, obtain the CR_image image according to the operations of the two-way average direction in Step S21 and Step S22.

[0114] Among them, the framework diagram of two-way polarization coherence suppression is as Figure 3 shown, and its calculation formula is as follows:

[0115]

[0116] Among them, C_Fimage represents the result of removing the average direction of the columns of the original image; CR_image represents the result of two-way feature coherence suppression; Fimage i,j represents the result of spatial neighborhood gradient weighted fusion at the position (i, j); N represents the total number of column pixels; k represents the number of pixels; M represents the total number of row pixels; C_Fimage k,j represents the result of removing the average direction of the columns of the original image at the position ( k, j ) on the original image.

[0117] Step S3. Since the shape of the target is damaged after two-way polarization coherence suppression and there is still some noise not removed completely. Therefore, the present invention proposes a connected statistical analysis method based on the gradient vector field to enhance the target edge information and denoise while maintaining the shape of the target.

[0118] Step S31. Calculate the gradient vector field and analyze the characteristics of small targets in the gradient vector field.

[0119] In the gradient vector field, each point represents the gradient of the corresponding point in the original image, and the gradient calculation formula is as follows:

[0120]

[0121] Among them, represents the gradient vector of the image I at the pixel point ( i, j ) ; and respectively represent the horizontal gradient and vertical gradient of the image at the pixel ( i, j ) point.

[0122] Step S32. Generally, the method of calculating the gradient of an image is to calculate using the information of adjacent two points. However, in order to consider the correlation of the gray values of other pixel points, it is proposed to estimate the gradient within a region based on all pixel points within a local window.

[0123] Assume that within a local window (usually a w×w window), the gray value of the image is approximately represented by a quadratic polynomial:

[0124] I ( x,y ) ≈a0+a1x+a2y+a3x 2 +a4xy+a5y 2 ;

[0125] Under this model, the horizontal gradient and the vertical gradient at the point (x, y) of the image are as follows:

[0126]

[0127] where a1, a2, a3, a4, and a5 represent the estimated coefficients in the polynomial.

[0128] Step S33. To estimate the coefficients, the least squares method is used to fit the gray values within the local window. The light intensity change at any point in any direction can be obtained by fitting through the above formula.

[0129] As Figure 4 shown, two gradient vector blocks of small target shapes are given. Each point in the block is represented by an arrow, which is composed of a direction and a length. The direction of the arrow represents the gradient direction of this point, and the length of the arrow represents the magnitude of the gradient. The magnitude of the gradient M ( x,y ) and the direction D ( x,y ) are as follows:

[0130]

[0131] Step S34. After calculating the gradient vector field, the connected component analysis is used to analyze the significant regions of the gradient change in the image.

[0132] Step S341. First, for the binary image obtained by thresholding processing, the pixels with gradient values greater than a certain threshold (threshold T = 10) are marked as significant change regions, and other pixels are marked as the background. The binary image formula is as follows:

[0133]

[0134] where Binary Map( x, y ) represents a binary image.

[0135] Step S342: Statistically calculate the number of connected components by performing connected component labeling on the binary image. Connected component labeling calculates the area and centroid position of each connected component. The area of each connected component is the number of pixels it contains. The calculation formulas for the area and centroid position are as follows:

[0136] Area ( k ) = ∑ x,y 1 ( label ( x, y ) = k ) ;

[0137]

[0138] Among them, Area ( k ) represents the connected component area; label ( x, y ) represents the label at the position (x, y); Centroid ( k ) represents the centroid position of the connected component; 1(·) is an indicator function that is 1 when the condition is true and 0 otherwise; k represents a substitution coefficient.

[0139] Step S343: Since the number of connected components represents the number of independent regions with the same pixel value in the image. Each pixel in these regions is directly or indirectly connected to other pixels in the region. The number of connected regions can be statistically obtained through connected component labeling. The expression for the number of connected components is as follows:

[0140]

[0141] Among them, Num_D represents the number of connected components; num labels represents the number of connected component labels.

[0142] According to the analysis of the gradient vector field, it is found that different image patches have different gradient characteristics. As Figure 4 shown, the arrows in the gradient blocks of small shapeless targets all point to the same position, and the length of the arrows is very long, which means that the modulus vector of the gradient is very large and the gradient blocks are concentrated. There are multiple blocks in the gradient blocks of small shaped targets where the arrows point to the same position, and the gradient blocks are more dispersed. Based on the uniqueness of small targets in the gradient vector field, the connected statistical analysis method is used to statistically analyze the distribution of gradient blocks and then distinguish different shaped small targets and use different methods to preserve edge information. As Figure 4As shown, the third column is a visualization graph of connected component statistics.

[0143] Therefore, the present invention adopts the above infrared polarization noise suppression and edge enhancement method based on gradient vector analysis. By utilizing the polarization image information in different directions and according to the concept of spatial distribution similarity, the obtained weighted image is used to replace the original polarization image to achieve the purpose of initially suppressing high-brightness noise in the polarization degree image. Since small targets occupy few pixels in the sky background, the average gradients in the horizontal and vertical directions are used to obtain significant targets based on two-way feature coherence suppression. A connected statistical analysis method based on the gradient vector field is proposed for small targets in different situations to suppress the background while maintaining the edge information of the targets. When there are small targets with shapes, it is easy to damage the edge information of the targets during background suppression. Therefore, the double-polarization weighted image is used as a guiding image to filter the image after initial suppression, which can effectively protect the edge information of the targets. The final processing results are as Figure 5 shown.

[0144] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that they can still modify or equivalently replace the technical solutions of the present invention, and these modifications or equivalent replacements do not make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. An infrared polarization noise suppression and edge enhancement method based on gradient vector analysis, characterized in that It includes the following steps: Step S1: First, based on the spatial distribution similarity in the polarization image and the double-polarization spatial neighborhood gradient weighting method, the polarization degree image is preliminarily suppressed for high-brightness noise. The specific process is as follows: Step S11: In the polarization imaging system, the Stokes vector is used to represent the polarization state of light. The general definition of the Stokes vector is as follows: ； Among them, represents the Stokes vector matrix; represents the total radiation intensity of light; represents the difference in the linear polarization intensity of light in the horizontal and vertical directions; represents the difference in the linear polarization intensity of light in the +45° and -45° directions; represents the circular polarization characteristic of light, that is, the intensity difference between left-handed circular polarization and right-handed circular polarization; represents matrix transpose; In the natural environment, the component of circularly polarized light is very small and can be ignored. Therefore, the Stokes vector degenerates into a third-order matrix as follows: ； Step S12: Select 0°, 60°, and 120° to obtain the object , and of the polarization intensity, and use the Stokes parameters to represent the degree of polarization of light; among them, the Stokes vector is as follows: ； ； ； Step S13. Among the Stokes parameters, and describe the polarization information in different directions of the target surface. Due to their spatial distribution similarity, spatial neighborhood gradient weighting is performed on two different polarization images. Step S14: According to the obtained polarization image and the spatial neighborhood gradient of, fuse through linear weighting to obtain the Fimage image, as follows: ； Among them, represents the spatially neighboring gradient weighted fusion result; represents the weighted gradient of image Q; represents the weighted gradient of image U; Step S15: Use the Fimage image to replace the original polarization image, and preliminarily suppress the polarization degree image to obtain the FDoP image, as follows: ； Among them, represents the processing result of the double-biased spatial neighborhood gradient weighting method; Step S2: Second, utilize the difference in polarization characteristics of small targets in the sky background, and propose two-way feature coherence suppression to obtain significant targets. Step S3: Finally, to prevent the problem of over-suppressing the target edge information during the background suppression process, based on the connected statistical analysis method of the gradient vector field, the target edge information is enhanced.

2. The infrared polarization noise suppression and edge enhancement method based on gradient vector analysis according to claim 1, characterized in that In step S12, the polarization degree calculation formula is as follows: ； Among them, represents the intensity of the completely polarized component; represents the total intensity in the optical vector.

3. The method for infrared polarization noise suppression and edge enhancement based on gradient vector analysis according to claim 1, wherein In step S13, the calculation formula of the spatial neighborhood gradient weighting is as follows: ； ； ； ； Among them, are the coordinates of pixels in the image; in the image is the pixel weighted gradient, which consists of two directions; is the gradient in the horizontal direction; is the gradient in the vertical direction; is the number of pixels in the neighborhood window; is the Gaussian weight function; and are the gray values of the pixels in the neighborhood and the central pixel respectively.

4. The infrared polarization noise suppression and edge enhancement method based on gradient vector analysis according to claim 1, characterized in that In step S2, utilize the difference in polarization characteristics of small targets in the sky background, and propose two-way feature coherence suppression to obtain significant targets. The specific process is as follows: Step S21: First, remove the average direction of the columns of the Fimage image; if the obtained column direction value is less than 0, it represents the background, and set its gray value to 0; otherwise, keep the original gray value. Step S22: Remove the average direction of the rows of the Fimage image again; if the obtained row direction value is less than 0, it represents the background, and set its gray value to 0; otherwise, keep the original gray value. Step S23: According to the two-way average direction operations in steps S21 and S22, obtain the CR_image image. Among them, the calculation formula of the two-way polarization coherence suppression is as follows: ； ； Among them, represents the result of removing the average direction of its columns from the original image; represents the result of two-way feature coherence suppression; represents at position the result of spatial neighborhood gradient weighted fusion; represents the total number of column pixels; represents the number of pixels; represents the total number of row pixels; represents at position the result of removing the average direction of its columns from the original image.

5. The infrared polarization noise suppression and edge enhancement method based on gradient vector analysis according to claim 1, characterized in that In step S3, based on the connected statistical analysis method of the gradient vector field, the target edge information is enhanced. The specific process is as follows: Step S31: Calculate the gradient vector field and analyze the characteristics of small targets in the gradient vector field. Step S32: Estimate the gradient within the region based on all pixel points within a local window to obtain the horizontal gradient and vertical gradient of the image. Step S33: Use the least squares method to fit the gray values within the local window, and fit the light intensity change at any point in any direction to obtain the modulus and direction of the gradient. Step S34: By calculating the gradient vector field, use the connected component analysis to analyze the significant region of the gradient change in the image.

6. The method for infrared polarization noise suppression and edge enhancement based on gradient vector analysis according to claim 5, wherein In step S31, each point in the gradient vector field represents the gradient of the corresponding point in the original image. The gradient calculation formula is as follows: ； Among them, represents the gradient vector of image I at pixel point ; and respectively represent the horizontal gradient and vertical gradient of the image at pixel point.

7. The method for infrared polarization noise suppression and edge enhancement based on gradient vector analysis according to claim 5, characterized in that, In step S32, estimate the gradient within the region based on all pixel points within a local window. The specific process is as follows: Within a local window, use a quadratic polynomial to represent the gray value of the image, as follows: ； Among them, the horizontal gradient of the image at point and the vertical gradient are as follows: ； ； Among them, , , , , represent the estimated coefficients in the polynomial.

8. The infrared polarization noise suppression and edge enhancement method based on gradient vector analysis according to claim 5, characterized in that In step S33, use the least squares method to fit the gray values within the local window, and fit the light intensity change at any point in any direction. The modulus and direction of the gradient are as follows: ； ； Among them, represents the magnitude of the gradient; represents the direction of the gradient.

9. The infrared polarization noise suppression and edge enhancement method based on gradient vector analysis according to claim 5, characterized in that In step S34, by calculating the gradient vector field, the significant regions of gradient changes in the image are analyzed using connected component analysis. The specific process is as follows: Step S341: Using the binary image obtained by thresholding, pixels with gradient values greater than the threshold are marked as significant change regions, while other pixels are marked as the background. The binary image formula is as follows: ； Among them, represents a binary image; Step S342: By performing connected component labeling on the binary image, the quantity is counted. Connected component labeling calculates the area and centroid position of each connected component. The area of each connected component is the number of pixels it contains. The calculation formulas for the area and centroid position are as follows: ； ； Among them, represents the area of the connected component; represents at the position the label on; represents the centroid position of the connected component; is an indicator function, which is 1 when the condition is true and 0 otherwise; represents the substitution coefficient; Step S343: By counting through connected component labeling, the number of connected regions is obtained. The expression for the number of connected components is as follows: ； Among them, represents the number of connected components; represents the number of connected component labels.

Images