An infrared image enhancement method based on polarization information

Abstract

The application relates to an image enhancement method, in particular to an infrared image enhancement method based on polarization information. The method comprises the following steps: 1, acquiring four polarization images of a target scene at different angles in an infrared wave band; 2, acquiring Stokes vector images S0, S1 and S2, a polarization degree image and a polarization angle image; 3, acquiring a temperature image T; step 4, acquiring a material image M; 5, acquiring a texture image C; step 6, combining the material image M, the temperature image T and the texture image C to generate an HSV enhanced image of the target scene, and realizing infrared image enhancement. The four angle polarization images of the target scene in the infrared wave band are acquired by an infrared polarization camera, polarization information is utilized to realize detection and sensing of the temperature, material and texture of the environment target, and then the three-dimensional physical information is fused through an HSV color space fusion method to obtain an HSV enhanced image with rich contrast and texture information.

Description

TECHNICAL FIELD

[0001] The present application relates to an image enhancement method, in particular to an infrared image enhancement method based on polarization information. BACKGROUND

[0002] For night vision needs, the current infrared imaging means is mostly used. Infrared thermal radiation is an ideal signal source to achieve night vision needs. According to Planck's law, all objects with temperature radiate heat signals. Thermal radiation is scattered and propagated, and diffused in every corner of the world. Therefore, thermal radiation not only has the advantages of "natural harmlessness", "passive", "concealment", and "night vision", but also carries important physical information such as object temperature, material, and surface texture. However, the traditional infrared imaging has a typical "ghost" effect. The infrared image has low contrast and lacks texture details. These defects are the main obstacles for using infrared imaging to extract information at night, and are also the main bottleneck for its wide application in visual navigation and fine target detection.

[0003] At present, the main means for enhancing infrared images is of two types: one is to improve the signal-to-noise ratio based on the infrared image itself in the spatial or frequency domain to complete image enhancement; the other is to fuse and enhance the intensity direction by combining multi-dimensional information (such as spectral dimension). The common point of the two types is that whether based on the infrared image itself or multi-dimensional information fusion, it is reflected in the enhancement of image intensity information, and lacks the application of multiple physical information in the infrared band, resulting in the problems of low contrast and lack of texture information in the obtained infrared image. SUMMARY

[0004] The purpose of the present application is to solve the technical problems of low contrast and lack of texture information in the infrared image obtained by the existing infrared image enhancement method, and to provide an infrared image enhancement method based on polarization information.

[0005] To achieve the above purpose, the technical scheme adopted by the present application is:

[0006] An infrared image enhancement method based on polarization information, characterized in that it comprises the following steps:

[0007] Step 1: obtaining four polarization images of a target scene at different angles in the infrared band;

[0008] Step 2: obtaining Stokes vector images S0, S1 and S2 based on the four polarization images at different angles, and obtaining a polarization degree image and a polarization angle image based on the Stokes vector images S0, S1 and S2;

[0009] Step 3, converting the obtained Stokes vector image S0 into pixel temperature information, and then normalizing the pixel temperature information to convert the temperature image T with a gray value of 0-1;

[0010] Step 4, performing cluster analysis on the polarization degree image and the polarization angle image, taking the intersection of the two image categories to determine the number of material categories in the polarization degree image and the polarization angle image, and generating a material image M with a gray value of 0-1 based on a preset correspondence between the number of material categories and the gray value;

[0011] Step 5, normalizing the four polarization images at different angles to obtain a normalized image set, and then converting the normalized image set into gray images C1, C2, C3 and C4 with a gray value of 0-1, and then taking the average of the gray images C1, C2, C3 and C4 to obtain a texture image C;

[0012] Step 6, combining the material image M, the temperature image T and the texture image C to generate an HSV enhanced image of the target scene to realize infrared image enhancement; wherein the material information in the material image M is the hue H in the HSV enhanced image, the temperature information in the temperature image T is the saturation S in the HSV enhanced image, and the texture information in the texture image C is the brightness V in the HSV enhanced image.

[0013] Further, step 1 is specifically:

[0014] Obtaining polarization images of the target scene at four different angles of 0°, 45°, 90° and 135° in the infrared band using an infrared polarization camera, and denoted as I(0°), I(90°), I(45°) and I(135°) respectively;

[0015] Step 2 is specifically:

[0016] Based on the polarization images I(0°), I(90°), I(45°) and I(135°), the Stokes vector images S0, S1 and S2 are calculated by the following formula:

[0017]

[0018] Based on the Stokes vector images S0, S1 and S2, the polarization degree image and the polarization angle image are calculated by the following formula:

[0019]

[0020] In the formula: DoLP is the polarization degree image, and AoLP is the polarization angle image;

[0021] Step 5 is specifically:

[0022] The polarization images I (0°), I (90°), I (45°) and I (135°) are normalized to obtain a normalized image set, and the normalized image set is converted into gray scale images C1, C2, C3 and C4 with gray scale values of 0-1, and then the gray scale images C1, C2, C3 and C4 are subjected to mean value processing to obtain a texture image C.

[0023] Further, in step 4, the preset correspondence between the number of material categories and the gray scale value is specifically:

[0024] When the number of material categories in the polarization degree image is 1, the gray scale value of the material image M is all 1;

[0025] When the number of material categories in the polarization degree image is 2, the gray scale value of the material image M is set to 0 and 1 according to the material temperature from low to high;

[0026] When the number of material categories in the polarization degree image is N, N≥3, the gray scale value of the material image M is set to 0, 1.

[0027] Further, in step 4, the polarization degree image and the polarization angle image are subjected to clustering analysis by a K-means clustering algorithm, and the initial K value of the K-means clustering algorithm is estimated by an elbow method.

[0028] Further, in step 6, the material image M, the temperature image T and the texture image C are combined to generate an HSV enhanced image of the target scene by an HSV color space fusion method.

[0029] Further, the obtained Stokes vector image S0 is converted into temperature information of each pixel according to the calibration information of the infrared polarization camera.

[0030] The present application has the following advantages:

[0031] 1. The present application provides an infrared image enhancement method based on polarization information, which obtains four angle polarization images of a target scene in an infrared band by an infrared polarization camera, realizes the detection and perception of environmental target temperature, material and texture by using polarization information, and then obtains an HSV enhanced image with rich contrast and texture information by an HSV color space fusion method.

[0032] 2. The present application can greatly improve the visual detection accuracy in low-illumination environments such as at night, and has great application value in the fields of unmanned driving and target detection. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1 The present application is a flowchart of an embodiment of an infrared image method based on polarization information. DETAILED DESCRIPTION

[0034] In order to make the objects, advantages and features of the present application clearer, a kind of infrared image enhancement method based on polarization information is further described in detail below in conjunction with the drawings and specific embodiments.The advantages and features of the present application will be clearer according to the following detailed description.

[0035] As shown in the drawings, the embodiment of the present application specifically includes the following steps: Figure 1

[0036] Step 1, using an infrared polarization camera to obtain four different angle (0°, 45°, 90°, 135°) polarization images I(0°), I(90°), I(45°), I(135°) in infrared band.

[0037] In this embodiment, an infrared polarization camera is selected as the data acquisition device, which has a size of 320x240 and a bit number of 16, and can measure temperature in the range of-20℃ to +120℃.

[0038] Step 2, based on the four angle polarization images I(0°), I(90°), I(45°), I(135°), the following formula is used to calculate the Stokes (Stokes) vector image (S0, S1, S2), the polarization degree image DoLP and the polarization angle image AoLP:

[0039]

[0040] Step 3, converting the Stokes vector image S0 obtained by the infrared polarization camera in step 2 into pixel temperature information according to the calibration information of the infrared polarization camera, and then normalizing the pixel temperature information to convert it into a temperature image T with a gray value of 0-1.

[0041] Step 4, using K-means clustering algorithm to perform clustering analysis on the polarization degree image DoLP and the polarization angle image AoLP, taking the intersection of the two image categories, and estimating the initial K value of the K-means clustering algorithm by elbow method (also known as elbow method). Determine the number of material categories in the polarization degree image and the polarization angle image, and generate a material image M based on the preset material category number and gray value correspondence relationship, and the gray value of the material image M ranges from 0 to 1.

[0042] Specifically, the aforementioned preset material category number and gray value correspondence relationship is:

[0043] I. When the number of material categories in the polarization degree image DoLP is 1, the gray value of the material image M is all 1;

[0044] ​II. When the number of material types in the polarization degree image DoLP is 2, the gray values of the material image M are set to 0 and 1 respectively from low to high according to the material temperature;

[0045] III. When the number of material types in the polarization degree image DoLP is N, N≥3, the gray values of the material image M are set to 0, 0.34, 0.67 and 1 respectively from low to high according to the material temperature. 1.

[0046] In this embodiment, the number of material types in the polarization degree image DoLP is taken as an example of 4, and the gray values of the material image M are set to 0, 0.34, 0.67 and 1 respectively from low to high according to the material temperature.

[0047] Step 5, normalize the polarization images I(0°), I(90°), I(45°) and I(135°) to obtain a normalized image set, and then convert the normalized image set to gray images C1, C2, C3 and C4 with gray values of 0-1, and then take the average of the gray images C1, C2, C3 and C4 to obtain a texture image C, the expression is as follows:

[0048]

[0049] Step 6, combine the material image M, the temperature image T and the texture image C to generate an HSV enhanced image Z to realize infrared image enhancement. The material information in the material image M is taken as the hue H in the HSV enhanced image Z, the temperature information in the temperature image T is taken as the saturation S in the HSV enhanced image Z, and the texture information in the texture image C is taken as the brightness V in the HSV enhanced image Z. That is, the material image M is the first dimension, the temperature image T is the second dimension, and the texture image C is the third dimension, and its expression is: Z=[M, T, C].

Claims

1. A method for enhancing an infrared image based on polarization information, characterized in that, The method comprises the following steps: Step 1: obtaining four polarization images of a target scene at different angles in an infrared band; Step 2: obtaining Stokes vector images S0, S1 and S2 based on the four polarization images at different angles, and then obtaining a polarization degree image and a polarization angle image based on the Stokes vector images S0, S1 and S2; Step 3: converting the obtained Stokes vector image S0 into pixel temperature information, and then performing normalization processing on the pixel temperature information to convert the pixel temperature information into a temperature image T with a gray value of 0-1; Step 4: performing cluster analysis on the polarization degree image and the polarization angle image, taking the intersection of the two image categories, determining the number of material categories in the polarization degree image and the polarization angle image, and generating a material image M with a gray value of 0-1 based on a preset material category number-gray value correspondence relationship; Step 5: performing normalization processing on the four polarization images to obtain a normalized image set, converting the normalized image set into gray images C1, C2, C3 and C4 with a gray value of 0-1, and then performing mean value processing on the gray images C1, C2, C3 and C4 to obtain a texture image C; Step 6: combining the material image M, the temperature image T and the texture image C to generate an HSV enhanced image of the target scene, and realizing infrared image enhancement; wherein the material information in the material image M is taken as the hue H in the HSV enhanced image, the temperature information in the temperature image T is taken as the saturation S in the HSV enhanced image, and the texture information in the texture image C is taken as the brightness V in the HSV enhanced image.

2. The method of claim 1, wherein the method is based on polarization information. Step 1 is specifically: Obtaining polarization images of a target scene at four different angles of 0°, 45°, 90° and 135° in an infrared band by using an infrared polarization camera, and denoting the polarization images as I(0°), I(90°), I(45°) and I(135°) respectively; Step 2 is specifically: Based on the polarization images I(0°), I(90°), I(45°) and I(135°), the Stokes vector images S0, S1 and S2 are calculated by the following formulas: Based on the Stokes vector images S0, S1 and S2, the polarization degree image and the polarization angle image are calculated by the following formulas: In the formula: DoLP is the polarization degree image, and AoLP is the polarization angle image; Step 5 is specifically: The polarization images I(0°), I(90°), I(45°) and I(135°) are normalized to obtain a normalized image set, and the normalized image set is converted into gray images C1, C2, C3 and C4 with a gray value of 0-1, and then the gray images C1, C2, C3 and C4 are subjected to mean value processing to obtain a texture image C.

3. The method of claim 1 or 2, wherein the method is based on polarization information. In step 4, the preset material category number-gray value correspondence relationship is specifically: When the number of material categories in the polarization degree image is 1, the gray value of the material image M is all 1; When the number of material categories in the polarization degree image is 2, the gray values of the material image M are set to 0 and 1 according to the material temperature from low to high; When the number of material types in the polarization image is N, N≥3, the gray values ​​of the material image M are set to 0, ...

1.

4. The infrared image enhancement method based on polarization information according to claim 3, characterized in that: In step 4, the polarization degree image and the polarization angle image are analyzed by a K-means clustering algorithm, and the initial K value of the K-means clustering algorithm is estimated by an elbow method.

5. The infrared image enhancement method based on polarization information according to claim 4, characterized in that: In step 6, the material image M, the temperature image T and the texture image C are combined by an HSV color space fusion method to generate an HSV enhanced image of the target scene.

6. The infrared image enhancement method based on polarization information according to claim 2, characterized in that: The obtained Stokes vector image S0 is converted into pixel temperature information according to the calibration information of the infrared polarization camera.

Images