A polarized image texture feature representation method and a computer readable storage medium

Abstract

The application discloses a polarization image texture feature representation method and a computer readable storage medium, and belongs to the technical field of polarization spectral imaging and image processing. θ The application solves the problem of selecting fixed direction parameters in the traditional texture feature representation method by combining actual polarization directions; the polarization gray level range co-occurrence matrix P θ is established, polarization texture feature change image of four different polarization directions is obtained, the mutual relationship between the pixel gray level and the gray level change degree is described, and the problem that single gray level relationship cannot sufficiently describe the polarization texture feature information of artificial targets and natural backgrounds is overcome; the weight coefficients ω θ of the polarization texture feature change images of the four different polarization directions are respectively calculated, the total polarization texture feature change image I F is obtained, and the polarization texture feature difference in the direction of artificial targets and natural backgrounds is enhanced, thereby providing a new idea for various model representation methods of the traditional texture features.

Description

Technical Field

[0001] This invention belongs to the field of polarization spectral imaging and image processing technology, specifically relating to a method for representing polarization image texture features and a computer-readable storage medium. Background Technology

[0002] Traditional light intensity imaging is ineffective for detecting and identifying man-made targets in cluttered environments or harsh weather conditions. Representing texture features in images presents challenges such as the ambiguity of texture feature definition, the structural differences within local regions, and the periodicity of the overall structure. Furthermore, traditional texture feature representation methods suffer from drawbacks, such as selecting fixed directional parameters and using a single gray-level relationship to describe the texture features of man-made targets and natural backgrounds. While polarization imaging, which obtains polarization features dependent on the object's inherent characteristics, can overcome some of the problems of traditional light intensity imaging, the commonly used weighted averaging algorithms weaken the differences in polarization texture features between man-made targets and natural backgrounds across different polarization directions. Summary of the Invention

[0003] The purpose of this invention is to provide a polarization image texture feature representation method, which solves the problem of selecting fixed direction parameters in traditional texture feature representation methods; overcomes the problem that a single gray-level relationship cannot fully describe the polarization texture feature information of artificial targets and natural backgrounds; enhances the difference in polarization texture features between artificial targets and natural backgrounds in the direction, and provides new ideas for various model representation methods of traditional texture features.

[0004] The technical solution adopted by the present invention to achieve the above objective is: a method for representing polarization image texture features, the method comprising the following steps, which are performed sequentially:

[0005] Step 1: Calculate and obtain fully polarized light intensity images for four different polarization directions.

[0006] By using a camera to acquire polarization images, polarization intensity images of four different polarization directions are obtained. θ θ = [0°, 45°, 90°, 135°], and then polarization orthogonal difference operation is performed to obtain fully polarized light intensity images in four different polarization directions.

[0007] Step 2: Calculate and obtain polarization texture feature change images for four different polarization directions.

[0008] Calculate the intensity images of fully polarized light in four different polarization directions. The polarization grayscale range co-occurrence matrix P θθ = [0°, 45°, 90°, 135°], obtaining polarization texture feature variation images in four different polarization directions.

[0009] Step 3: Calculate and obtain the total polarization texture feature change image I F

[0010] Calculate the polarization texture feature changes in images with four different polarization directions. Weighting coefficient ω θ θ = [0°, 45°, 90°, 135°], to obtain the total polarization texture feature change image I. F .

[0011] Furthermore, in step one, the polarization orthogonal difference operation is performed as follows:

[0012]

[0013] in, I represents the intensity image of fully polarized light in the θ polarization direction. θ and I θ+π / 2 Representing θ and Image of polarized light intensity along the polarization direction.

[0014] Furthermore, step two specifically includes:

[0015] ① Calculate the intensity images of fully polarized light in four different polarization directions. The polarization grayscale range co-occurrence matrix P θ :

[0016]

[0017]

[0018] Among them, P θ (i,j) represents the polarization grayscale range co-occurrence matrix P. θ The value of the (i,j)th element, i.e., the image of the fully polarized light intensity along the θ polarization direction. The probability that a gray value i and its gray value change j occur in pairs; (i,j) represents the intensity image of fully polarized light in the θ polarization direction. The situation where the gray value i and its gray value change level j appear in pairs is simply called a polarized gray value-gray value change level pair; I V Image representing the intensity of fully polarized light The degree of grayscale change of the upper pixel (x,y) in the θ polarization direction; and These are images of the fully polarized light intensity along their respective θ polarization directions. The pixel values ​​at (x,y), (x,y+1), (x-1,y+1), (x-1,y), and (x-1,y-1); S represents the image with fully polarized light intensity. The set of polarization gray-level - gray-level change degree pairs (i,j) in a 5*5 target region centered on the upper pixel (x,y); Let S represent the number of occurrences of each polarization gray level-gray level change pair (i,j) in set S, and count{S} represent the total number of occurrences of each polarization gray level-gray level change pair (i,j) in set S.

[0019] ② Obtain images showing the changes in polarization texture features in four different polarization directions. The formula is as follows:

[0020]

[0021] in, Image showing the variation of polarization texture features in the θ polarization direction The pixel value at (x,y), i.e., the statistical polarization grayscale range co-occurrence matrix P in the 5*5 target region. θ The polarization texture features are obtained and assigned to the center pixel (x, y) of the target region; k represents the gray level size of the original image after dimensionality reduction, with a value of 8, and the polarization gray-level range co-occurrence matrix P θ Its size is k*k.

[0022] Furthermore, step three specifically includes:

[0023] ① Calculate the polarization texture feature changes in images with four different polarization directions. Weighting coefficient ω θ The formula is as follows:

[0024]

[0025]

[0026] Where, ω θ This represents the image showing the change in polarization texture features along the θ polarization direction. Weighting coefficient values; I A Image representing changes in polarization texture features The values ​​of adjacent pixels (x, y) in the θ polarization direction; and Images showing the changes in polarization texture features along their respective θ polarization directions. Pixel values ​​at (x,y), (x,y+1), (x-1,y+1), (x-1,y), and (x-1,y-1); image showing polarization texture feature changes. The size is m*n;

[0027] ② Obtain the total polarization texture feature change image I F The formula is as follows:

[0028]

[0029] A computer-readable storage medium storing computer instructions that, when executed, cause the computer to perform the steps of the polarization image texture feature representation method.

[0030] Through the above design scheme, the present invention can bring the following beneficial effects: The present invention provides a method for representing polarization image texture features, which eliminates the natural light component through polarization orthogonal difference operation and obtains fully polarized light intensity images in four different polarization directions. Using the actual polarization direction, establish the polarization grayscale range co-occurrence matrix P. θ The polarization texture features were calculated to obtain images showing the changes in polarization texture features in four different polarization directions. Considering the changes in polarization texture features in the image Due to the anisotropy of polarization, the changes in polarization texture features in four different polarization directions were calculated. Weighting coefficient ω θ Obtain the total polarization texture feature change image I F This invention solves the problem of selecting fixed direction parameters in traditional texture feature representation methods by incorporating actual polarization direction; it establishes a polarization grayscale range co-occurrence matrix P. θ Images showing the changes in polarization texture features in four different polarization directions were obtained. This method describes the relationship between pixel grayscale and the degree of grayscale variation, overcoming the problem that a single grayscale relationship cannot fully describe the polarization texture feature information of artificial targets and natural backgrounds; it calculates the polarization texture feature changes in four different polarization directions separately. Weighting coefficient ω θ Obtain the total polarization texture feature change image I F This enhances the difference in polarization texture features between artificial targets and natural backgrounds, providing new ideas for various model representation methods of traditional texture features. Attached Figure Description

[0031] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and are used to understand the invention. They do not constitute an improper limitation of the invention. In the drawings:

[0032] Figure 1 A flowchart of a method for representing texture features in polarized images;

[0033] Figure 2 The polarization gray-scale range co-occurrence matrix P in polarization image texture feature representation methods θ A schematic diagram showing the positional relationship of pixels in four different polarization directions;

[0034] Figure 3 In this embodiment of the method for representing texture features of polarized images, the image shows the intensity of fully polarized light in the 0° polarization direction. A schematic diagram of the grayscale value distribution of pixels within a 7x7 region after dimensionality reduction;

[0035] Figure 4 In this embodiment of the polarization image texture feature representation method, the polarization grayscale range co-occurrence matrix P in the 0° polarization direction is... θ Schematic diagram. Detailed Implementation

[0036] To make the objectives, features, and advantages of this invention more apparent and understandable, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, this invention is not limited to the following embodiments, and specific implementation methods can be determined according to the technical solutions of this invention and actual circumstances. To avoid obscuring the essence of this invention, well-known methods, processes, and procedures are not described in detail.

[0037] like Figure 1 As shown, the present invention provides a method for representing texture features of polarized images, the method comprising the following steps:

[0038] Step S1: Calculate and obtain fully polarized light intensity images for four different polarization directions.

[0039] The specific process is as follows: Four polarization intensity images in different polarization directions are acquired by rotating the polarizer in front of the polarization camera. θ θ = [0°, 45°, 90°, 135°]; for any polarization direction θ, the polarization intensity image I θ Both can be decomposed into natural light components and polarized light components. The natural light component is eliminated by polarization orthogonal difference operation, resulting in fully polarized light intensity images with four different polarization directions. It contains richer polarization feature information, reducing the adverse effects of cluttered environments or severe weather conditions on the detection and recognition of man-made targets.

[0040] The polarization orthogonal difference operation method is as follows:

[0041]

[0042] In the formula, This represents the intensity image of fully polarized light in the θ polarization direction; Iθ and I θ+π / 2 Representing θ and Image of polarized light intensity along the polarization direction.

[0043] Step S2: Calculate and obtain polarization texture feature change images for four different polarization directions.

[0044] To reduce computational load, the intensity images of fully polarized light in four different polarization directions are used. The dimensions were reduced to 8 gray levels, where the gray level size is represented by k; then, the intensity images of the fully polarized light in four different polarization directions were taken after the dimension reduction. For any pixel (x, y) and its adjacent pixels along the θ polarization direction, calculate the grayscale change value j between these two pixels. Let the grayscale value of pixel (x, y) be i. Then (i, j) is the image of the fully polarized light intensity along the θ polarization direction. The reduced grayscale value i and its grayscale variation level j appear in pairs, and (i,j) is simply referred to as a polarization grayscale-grayscale variation level pair. When a pixel (x,y) moves within a 5*5 target area centered on it, k*k different combinations of polarization grayscale-grayscale variation level pairs (i,j) are obtained. For each 5*5 target area, the number of occurrences of each polarization grayscale-grayscale variation level pair (i,j) is counted, and then compared with the total number of occurrences of all polarization grayscale-grayscale variation level pairs (i,j) to obtain the probability of occurrence of each polarization grayscale-grayscale variation level pair (i,j). These probabilities are then arranged into a k*k square matrix, which is the polarization grayscale variation range co-occurrence matrix P. θ , θ=[0°, 45°, 90°, 135°];

[0045]

[0046]

[0047] Among them, P θ (i,j) represents the polarization grayscale range co-occurrence matrix P. θ The value of the (i,j)th element, i.e., the image of the fully polarized light intensity along the θ polarization direction. The probability that a gray value i and its gray value change j occur in pairs; (i,j) represents the intensity image of fully polarized light in the θ polarization direction. The situation where the gray value i and its gray value change level j appear in pairs is simply called a polarized gray value-gray value change level pair; I V Image representing the intensity of fully polarized light The degree of grayscale change of the upper pixel (x,y) in the θ polarization direction; and These are images of the fully polarized light intensity along their respective θ polarization directions. The pixel values ​​at (x,y), (x,y+1), (x-1,y+1), (x-1,y), and (x-1,y-1); S represents the image with fully polarized light intensity. The set of polarization gray-level - gray-level change degree pairs (i,j) in a 5*5 target region centered on the upper pixel (x,y); Let S represent the number of occurrences of each polarization gray level-gray level change pair (i,j) in set S, and count{S} represent the total number of occurrences of each polarization gray level-gray level change pair (i,j) in set S.

[0048] like Figure 2 As shown, images of fully polarized light intensity in different polarization directions. The polarization grayscale range co-occurrence matrix P θ They are different. Addressing the issue of selecting fixed direction parameters in traditional texture feature representation methods, this invention combines the actual polarization direction to calculate fully polarized light intensity images in four different polarization directions. The probability of occurrence of the polarization gray-level change pair (i,j) between adjacent pixels along their respective θ polarization directions is used to establish the polarization gray-level range co-occurrence matrix P for each pixel. θ Image of fully polarized light intensity in the 0° polarization direction. Polarization grayscale range co-occurrence matrix It calculates the probability of the occurrence of the polarization gray level-gray level change pair (i,j) between pixels (x,y) and (x,y+1), representing a fully polarized light intensity image along a 45° polarization direction. The polarization grayscale range co-occurrence matrix P θ It calculates the probability of the occurrence of the polarization gray-level-gray-level change pair (i,j) between pixels (x,y) and (x-1,y+1), representing the intensity image of fully polarized light in the 90° polarization direction. The polarization grayscale range co-occurrence matrix P θ It calculates the probability of occurrence of the polarization gray level-gray level change pair (i,j) between pixel points (x,y) and (x-1,y), representing a fully polarized light intensity image along a 135° polarization direction. The polarization grayscale range co-occurrence matrix P θ It calculates the probability of the occurrence of the polarization gray level - gray level change ratio (i,j) between pixel points (x,y) and (x-1,y-1).

[0049] like Figure 3 As shown, a 7x7 image of the fully polarized light intensity in the 0° polarization direction after dimensionality reduction is presented. In a 5x5 target area, for any pixel (x, y) and its adjacent pixel (x, y+1) in the 0° polarization direction, calculate the grayscale change value j between these two pixels. The resulting polarization grayscale-grayscale change pair is (i, j). For example, pixel (2, 3) has a grayscale value of 0, and its adjacent pixel (2, 4) in the 0° polarization direction has a grayscale value of 3. Therefore, the grayscale change value between these two pixels is 3, and the polarization grayscale-grayscale change pair is (0, 3). In the entire 5x5 target area, this polarization grayscale-grayscale change pair (0, 3) appears only once, and all polarization grayscale-grayscale change pairs (i, j) appear a total of 20 times. Therefore, the probability of this polarization grayscale-grayscale change pair (0, 3) appearing is... That is, the polarization grayscale range co-occurrence matrix P of the 5*5 target region. θ The element at position (0,3) has a value of like Figure 4 As shown, the probability of each polarization grayscale-grayscale change level pair (i,j) is statistically calculated, resulting in a k*k polarization grayscale range co-occurrence matrix P. θ , θ=0°.

[0050] Texture is a key visual feature in images. The polarization texture features of polarization images exhibit certain differences in local areas, but show a certain periodicity overall. This can be determined by statistically analyzing the polarization grayscale range co-occurrence matrix P in a 5x5 target region. θ polarization texture feature quantity This measures the difference in polarization texture features within a local region; a larger value indicates a smaller difference in polarization texture features. The polarization grayscale range co-occurrence matrix P is calculated for each 5x5 target region in the entire image. θ polarization texture feature quantity Obtain images of polarization texture feature changes in four different polarization directions.

[0051]

[0052] in, Image showing the variation of polarization texture features in the θ polarization direction The pixel value at (x,y), i.e., the statistical polarization grayscale range co-occurrence matrix P in the 5*5 target region. θ The polarization texture features are obtained and assigned to the center pixel (x, y) of the target region; k represents the gray level size of the original image after dimensionality reduction, with a value of 8, and the polarization gray-level range co-occurrence matrix P θ Its size is k*k.

[0053] Step S3: Calculate and obtain the total polarization texture feature change image I F .

[0054] Polarization texture features often exhibit a certain periodicity overall. The magnitude of this periodicity is related to the correlation between a pixel (x, y) within a local region and its adjacent pixels along the θ polarization direction. The changes in polarization texture features along four different polarization directions were calculated. The correlation between a pixel (x, y) and its adjacent pixels along its respective θ polarization direction is used as the image of polarization texture feature changes along that polarization direction. Weighting coefficient ω θ , θ=[0°, 45°, 90°, 135°].

[0055]

[0056]

[0057] Where, ω θ This represents the image showing the change in polarization texture features along the θ polarization direction. Weighting coefficient values; I A Image representing changes in polarization texture features The values ​​of adjacent pixels (x, y) in the θ polarization direction; and Images showing the changes in polarization texture features along their respective θ polarization directions. Pixel values ​​at (x,y), (x,y+1), (x-1,y+1), (x-1,y), and (x-1,y-1); image showing polarization texture feature changes. The size is m*n.

[0058] Weighting coefficient ω θ The larger the value, the greater the variation in polarization texture features along the θ polarization direction in the image. The larger the period, the greater the variation in polarization texture features in the image. The polarization texture features are coarse. Consider the changes in polarization texture features in the image. Due to directional anisotropy, the total polarization texture feature change image I was calculated and obtained. F ;

[0059]

[0060] A computer-readable storage medium storing computer instructions that, when executed, cause the computer to perform the steps of the polarization image texture feature representation method.

[0061] It is understood that this invention has been described through some embodiments, and those skilled in the art will recognize that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of this invention. Furthermore, based on the teachings of this invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of this invention. Therefore, this invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the protection scope of this invention.

Claims

1. A method for representing texture features in polarized images, characterized in that, The method includes the following steps: And the following steps are performed in sequence: Step 1: Calculate and obtain the intensity images of fully polarized light in four different polarization directions. ; By using a camera to acquire polarization images, polarization intensity images of four different polarization directions are obtained. , Then, polarization orthogonal difference operations are performed to obtain fully polarized light intensity images in four different polarization directions. ; Step 2: Calculate and obtain polarization texture feature change images for four different polarization directions. ; Calculate the intensity images of fully polarized light in four different polarization directions. Polarization grayscale range co-occurrence matrix , Images showing the changes in polarization texture features in four different polarization directions were obtained. ; Step 3: Calculate and obtain the total polarization texture feature change image. ; Calculate the polarization texture feature changes in images with four different polarization directions. Weighting coefficients , Obtain the image of total polarization texture feature changes. ; Step two specifically includes: ① Calculate the intensity images of fully polarized light in four different polarization directions. Polarization grayscale range co-occurrence matrix : ； ； in, Represents the polarization grayscale range co-occurrence matrix No. The value of each element, i.e. Fully polarized light intensity image in polarization direction grayscale value Its grayscale change The probability of them appearing in pairs; express Fully polarized light intensity image in polarization direction grayscale value Its grayscale change When two pairs appear, they are simply referred to as polarization grayscale-grayscale change degree pairs. Image representing the intensity of fully polarized light Pixel exist The degree of grayscale change along the polarization direction; , , , and Each of them Fully polarized light intensity image in polarization direction exist , , , , Pixel value at; Represents the intensity image of fully polarized light. Pixel Centered The degree of polarization gray-level change in the target region A set; express The degree of gray-level change for each polarization in the set The number of occurrences express The degree of gray-level change of various polarizations in the set Total number of occurrences; ② Obtain polarization texture feature change images for four different polarization directions, using the following formula: ， ； in, Indicates in Image showing changes in polarization texture features along the polarization direction exist The pixel value at that location, i.e., the statistical value. Polarization grayscale range co-occurrence matrix in the target region The polarization texture features are then assigned to the center pixel of the target region. ; This represents the grayscale level of the original image after dimensionality reduction, with a value of 8. The size of the polarization grayscale range co-occurrence matrix is... .

2. The polarization image texture feature representation method according to claim 1, characterized in that, In step one, the polarization orthogonal difference operation is performed as follows: ， ； in, Indicates in Intensity image of fully polarized light along the polarization direction. and They represent in and Image of polarized light intensity along the polarization direction.

3. The method for representing polarization image texture features according to claim 1, characterized in that, Step three specifically includes: ① Calculate the polarization texture feature changes in images with four different polarization directions. Weighting coefficients The formula is as follows: ； ； in, Indicates in Image showing changes in polarization texture features along the polarization direction The weighting coefficient value; Image representing changes in polarization texture features Pixel exist Values ​​of adjacent pixels in the polarization direction; , , , and Each of them Image showing changes in polarization texture features along polarization direction exist , , , , Pixel values ​​at the location; image showing changes in polarization texture features. The size is ; ② Obtain the image showing the change in total polarization texture features. The formula is as follows: ， 。 4. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions, which, when executed, cause the computer to perform the steps of the polarization image texture feature representation method according to any one of claims 1-3.

Images