CN111815550A - An infrared and visible light image fusion method based on gray level co-occurrence matrix - Google Patents

Abstract

The invention discloses an infrared and visible light image fusion method based on a grayscale co-occurrence matrix. First, the gray-scale co-occurrence matrix analysis is performed on the infrared source image, and the infrared target saliency map is obtained. Secondly, non-subsampling contourlet transform (NSCT) is performed on the visible light and infrared source images, the low-frequency sub-band images obtained by decomposing are fused to preserve the contrast, and the high-frequency sub-band images are fused by the improved Gaussian difference method; The target saliency map is mapped onto the fused low-frequency subband image. Finally, inverse NSCT transform is performed to obtain the final fused image. The invention utilizes the texture analysis characteristic of the grayscale co-occurrence matrix to detect the saliency of the infrared target, can effectively extract the infrared target and retain rich detail information, and improve the quality of the image after fusion. The objective evaluation index of the method proposed by the invention is superior to the existing classical image fusion methods such as wavelet transform and pyramid transform, and has strong robustness.

Description

An infrared and visible light image fusion method based on gray level co-occurrence matrix

technical field

The invention belongs to the technical field of multi-source image fusion, and in particular relates to an infrared and visible light image fusion method based on a grayscale co-occurrence matrix.

Background technique

Infrared and visible light image fusion technology is an important part of image fusion research. The infrared image is the radiation image processed by the sensor of infrared radiation emitted by the target and the background, which can detect hidden or camouflaged targets. Visible light images record the characteristics of visible light reflection of objects, including a lot of details and texture information, which are in line with the characteristics of human vision.

The purpose of the fusion of infrared and visible light images is to obtain a complete image that contains rich detailed information and accurately reflects the infrared target. Therefore, this technology is widely used in nighttime imaging equipment to improve the nighttime activities of people or machines. In addition, because the fused infrared and visible light images are accurate, clear and complete, they are also used in military reconnaissance, biometrics, fields such as medical imaging and remote sensing.

With the continuous development of computer and image processing technology, the most used fusion method is still pixel-level fusion. This method is divided into two categories: spatial domain fusion and transform domain fusion. The representative algorithm of the former is principal component analysis. The representative algorithms of the latter include: pyramid transform, wavelet transform and various multi-scale decomposition algorithms, such as methods based on non-subsampled contourlet transform (NSCT). In addition, other methods such as Compressed Sensing (CS), Sparse Representation (SR), etc. are also included.

Among the above methods, the NSCT method has the characteristics of translation invariance and can effectively overcome the pseudo-Gibbs phenomenon. It is a widely used image analysis tool. However, in the practical application of NSCT, the texture information of the fused image is often emphasized and the infrared target is ignored, or the infrared target is emphasized and the texture details are lost, both of which cannot be taken into account at the same time.

SUMMARY OF THE INVENTION

The purpose of the present invention is to propose an infrared and visible light image fusion method based on grayscale co-occurrence matrix analysis and non-subsampling contourlet transform, so that the fused image can not only retain detailed information but also highlight the infrared target and improve the fusion quality.

In order to realize the above fusion method, the present invention fuses the registered visible light image I ₁ and the infrared image I ₂ , and both images are grayscale images of N _y ×N _x pixels.

The specific fusion steps are as follows:

Step S1: perform grayscale co-occurrence matrix analysis on the infrared image _I2 , extract the infrared target, and obtain the target saliency map;

Step S2: performing NSCT decomposition on the visible light image I ₁ and the infrared image I ₂ to obtain a low-frequency sub-band image and a series of high-frequency sub-band images respectively;

Step S3: fuse all low-frequency sub-band images to maintain contrast information to obtain a fused low-frequency image, and use an improved Gaussian difference method to fuse all high-frequency sub-bands to obtain a fused high-frequency sub-band image;

Step S4: mapping the target saliency map to the fused low-frequency subband image;

Step S5: Perform inverse NSCT transformation on the fused low-frequency sub-band and high-frequency sub-band to obtain a fused image.

Further, the method for extracting the infrared image target in the step S1 is as follows:

(1) Carry out preliminary target extraction, and take the absolute value of the difference between the pixel value of the source infrared image and its gray mean value as the preliminary target extraction image SalPre;

(2) Calculate the gray-level co-occurrence matrix coMat of SalPre, which is a symmetric matrix. If both a and b are the grayscale values of the SalPre image, then (a, b) is a grayscale value pair. An element value in the grayscale co-occurrence matrix is each pixel value a, and the number of pixel values b in the neighborhood range of its size w is w=3 in the present invention;

(3) Process the gray-level co-occurrence matrix coMat to obtain a modified gray-level co-occurrence matrix; it specifically includes the following steps: first, normalize the gray-level co-occurrence matrix coMat; then use a logarithmic function for processing; The average value is subtracted from the elements larger than the average value, and the elements smaller than or equal to the average value are taken as zero, and then the modified grayscale co-occurrence matrix Sal(a, b) is obtained;

(4) Map the modified grayscale co-occurrence matrix to the preliminary target extraction image according to the following formula:

In the formula, U(a, b) is the average value of (a, b) pixel pairs in the w×w neighborhood, SalMap(a, b) is the saliency detection map of the same size as the source image, and then the image is normalized. unification;

(5) Combine SalMap and SalPre to obtain a target extraction image with a more prominent infrared target and a smoother background. The specific formula is:

SalFinal(a,b)=SalMap(a,b).*SalPre(a,b)

Among them, .* represents the multiplication of corresponding values.

Further, the NSCT decomposition rules in the step S2 are as follows: the number of directions for each level of decomposition is 8, and the number of decomposition scales is adaptive to the size of the image, and the formula is:

l=[log ₂ (min(N _y , N _x ))-7]

In the formula, l is the number of decomposition scales, and [] is rounded up.

Further, the fusion rule of the low-frequency sub-band coefficients in the step S3 is set as follows: the coefficients of the low-frequency sub-band of the visible light image and the low-frequency sub-band of the infrared image are respectively different from their average values to obtain coefficient matrices w ₁ , w ₂ , then The fusion weight matrix of the infrared image is w=(w ₂ -w ₁ )×0.5+0.5, and the fusion weight matrix of the visible light image is 1-w. Finally, the low frequency subband coefficients after fusion are obtained by multiplying the weight by the low frequency subband.

Further, the high-frequency sub-band coefficient fusion rules in the step S3 are set as follows: for each layer of high-frequency sub-band coefficients, the coefficients of the visible light image high-frequency sub-band and the infrared image high-frequency sub-band are respectively made difference with their average values. , obtain the preliminary fusion coefficients a ₁ , a ₂ ; then perform Gaussian filtering on the original high-frequency sub-band coefficients, where the filter template size is 11×11 and the standard deviation is 5, then the filtered high-frequency sub-band coefficients are b ₁ , b ₂ , the fusion weight of the visible light image is s ₁ =b ₁ -a ₁ , and similarly, the fusion weight of the infrared image is s ₂ =b ₂ -a ₂ . According to the weight, the high-frequency coefficients after fusion are selected.

Further, in the step S4, an addition method is used to map the target saliency map to the fused low-frequency subband image.

Compared with the prior art, the present invention has the following beneficial effects:

First, before fusing the two source images, the present invention uses the texture analysis characteristics of the grayscale co-occurrence matrix to detect the saliency of the infrared target and map it to the fused image. This step makes the infrared target in the source image in the It is more pronounced in the fused image and is well preserved in the fused image.

Second, the present invention adopts a low-frequency fusion rule that maintains contrast information, so that the fused image can better meet the visual characteristics of human eyes.

Third, for high-frequency sub-band images containing a large amount of detailed information, the present invention adopts an improved Gaussian difference algorithm, so that the detailed information in the source image can be retained more completely, and the halo around the infrared target can be effectively reduced.

Finally, experiments show that the objective evaluation index of the image fusion method proposed by the present invention is better than the existing popular image fusion methods such as wavelet transform and pyramid transform, and the fused image is more in line with human visual requirements.

Description of drawings

Fig. 1 is the fusion flow chart of the present invention;

Fig. 2 is a flowchart of infrared target extraction;

Figure 3 is a structural frame diagram of NSCT;

Fig. 4 is the structure diagram of NSP after being decomposed by 3 scales;

Figure 5 is a segmentation diagram of the image frequency domain of NSDFB at 3 scales;

Fig. 6 is the "Camp" image of Example 1 in the present invention, Fig. 6(a) is a visible light image, and Fig. 6(b) is an infrared image;

Fig. 7 is an image of "Tree" in Example 2 of the present invention, Fig. 7(a) is a visible light image, and Fig. 7(b) is an infrared image.

Detailed ways

In order to facilitate the understanding and implementation of the present invention by those of ordinary skill in the art, the present invention will be described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the embodiments described herein are only used to illustrate and explain the present invention, and not to limit the scope of the present invention. in the present invention.

A flowchart of a method for fusion of infrared and visible light images based on a grayscale co-occurrence matrix provided by the present invention is shown in FIG. 1 . The method firstly extracts the infrared target from the source infrared image to obtain the target saliency map, and then performs non-subsampling contourlet transform (NSCT) multi-scale decomposition on the source infrared and visible light images, respectively, to obtain a low-frequency subband image and a series of High-frequency sub-band images; then adopt the fusion method that preserves the contrast information for the low-frequency sub-band images, and use the improved Gaussian difference method to fuse the high-frequency sub-band coefficients to obtain the fused low-frequency sub-band images and a series of high-frequency sub-bands image; again, the target saliency map is mapped to the low-frequency subband image; finally, the fused image is obtained by inverse NSCT transform. Specific steps are as follows:

Step S1. Perform infrared target extraction on the source infrared image to obtain a target saliency map, as shown in Figure 2. The specific steps are as follows:

(1) Preliminary target extraction. Let I ₂ be the infrared image, and mean(I ₂ ) be the grayscale average value of the infrared image, then the preliminary target extraction image can be calculated by the following formula:

SalPre=I ₂ -mean(I ₂ )

In the formula, SalPre is the preliminary target extraction image, and its size is N _y ×N _x ;

After preliminary extraction, although the contrast of the infrared target has been enhanced, there are still many objects in the background image that interfere with the judgment of the infrared target, and the infrared target needs to be further extracted.

(2) Set the gray value range of I ₂ to be {0, 1, 2, ..., Q-1}, and calculate the gray level co-occurrence matrix of the SalPre image. The formula is as follows:

coMat=F(a,b)

In the formula, coMat is a symmetric matrix with a size of Q×Q, a and b are the grayscale values of the SalPre image, and (a, b) is a grayscale value pair. In the SalPre image, for each pixel value a, the number of pixel values b is calculated within the neighborhood range of size w (in the embodiment of the present invention, w=3), and the calculation result is stored in the coMat matrix middle.

(3) Process the gray-level co-occurrence matrix coMat to obtain a modified gray-level co-occurrence matrix; specifically, the following steps are included:

First, normalize the coMat matrix with the following formula:

In the formula, P(a, b) is the probability that the gray value pair (a, b) appears in the coMat matrix;

The Gray Level Co-occurrence Matrix (GLCM) refers to the joint probability distribution of the simultaneous occurrence of two gray pixels with a distance of d in an image. For an image of size M×N, the calculation steps of the gray level co-occurrence matrix are as follows:

Let the gray value of the pixel located at (x, y) be g ₁ , and the gray value of the pixel located at (x+i, y+j) to be g ₂ ; set (x, y) on the entire image Move, you will get different (g ₁ , g ₂ ) gray value pairs, count the number of occurrences of each (g ₁ , g ₂ ), if the gray value series of the whole image is l, then all ( The gray values of g ₁ , g ₂ ) are combined into an l×l matrix, and finally the total number of occurrences of (g ₁ , g ₂ ) is normalized to the probability of occurrence P(g ₁ , g ₂ ), Such a matrix is the gray-level co-occurrence matrix.

For a scan window of size 3×3, when i=1, j=0, the pixel pair is horizontal, i.e. 0° scan; when i=1, j=1, the pixel pair is located on the upper right diagonal, i.e. 45° scan, and so on, we can get the gray level co-occurrence matrix in a specific direction. In the method used in the present invention, 8 directions (0°, 45°, 90°, 135°, 180°, 225°, 270° and 315°) were calculated. For images with slow gray value changes, such as background images, the value on the diagonal of the gray co-occurrence matrix is larger, while for images with drastic changes in gray values, such as details, sudden changes, and gray co-occurrence matrix The value on the diagonal is smaller and the values on the sides are larger.

Secondly, since the smaller the value in P(a, b), the more severe the gray value change here, the larger the value, the slower the gray value change, and the smaller the value in P(a, b), so the In order to facilitate the analysis, the following formula is used for logarithmic processing:

L _p (a, b)=2×[−ln(P(a, b))] ²

In this way, the value in L _p (a, b) is proportional to the change in grayscale value, and the value is larger.

Finally, in order to highlight the saliency area, that is, to increase the difference between the saliency area and the background image, the following provisions are made:

In the formula, ENT is the mean value of L _p (a, b). In this way, the modified grayscale co-occurrence matrix Sal(a, b) is obtained.

(4) Map the corrected grayscale co-occurrence matrix to the preliminary detection image SalPre:

Among them, U(a, b) is the average value of (a, b) pixel pair in the neighborhood of w × w, SalMap(a, b) is the saliency detection map of size N _y × N _x , and then The image is normalized.

(5) Combine SalMap and SalPre to obtain a target extraction image with a more prominent infrared target and a smoother background. The specific formula is:

SalFinal(a,b)=SalMap(a,b).*SalPre(a,b)

Among them, .* represents the multiplication of corresponding values.

Step S2. Perform NSCT decomposition on the visible light image I ₁ and the infrared image I ₂ to obtain a low-frequency sub-band image and a series of high-frequency sub-band images respectively. The formula is as follows:

和

为源图像，

和

分别为可见光图像与红外图像的低频子带系数，

和

为高频子带系数，k为每级分解的方向数，本发明的实施例中，规定k＝8，l为分解尺度数量，l采用如下公式计算，使其自适应于图像尺寸：In the formula,

and

is the source image,

and

are the low-frequency subband coefficients of the visible light image and the infrared image, respectively,

and

is the high-frequency subband coefficient, k is the number of directions for each level of decomposition, in the embodiment of the present invention, it is specified that k=8, l is the number of decomposition scales, and l is calculated by the following formula to make it adaptive to the image size:

l=[log ₂ (min(N _y , N _x ))-7]

In the formula, [] is rounded up.

NSCT consists of two parts: non-subsampling pyramid (NSP) and non-subsampling directional filter (NSDFB). The decomposition diagram of NSCT is shown in Figure 3.

The source image is decomposed into low-frequency sub-band and high-frequency sub-band through NSP. After this layer is decomposed, the low-frequency image of this layer continues to be decomposed by NSP. After N-layer decomposition, a low-frequency sub-band sum (2 ¹ +2 ² +...+2 ^N ) high frequency subbands, as shown in FIG. 4 .

NSDFB is used to complete the multi-directional decomposition of high-frequency sub-band images, and synthesize the singular points in the same direction into the coefficients of NSCT. The decomposition diagram is shown in Figure 5.

Compared with the traditional CT algorithm, NSCT not only retains multi-directionality and anisotropy, but also reduces the distortion in the sampling process because it does not use up-sampling and down-sampling, so that the algorithm has translation invariance.

Step S3. Adopt the fusion rule of maintaining contrast for the low frequency sub-band, and the specific steps are as follows:

分别为可见光图像与红外图像低频子带系数的均值，首先对原低频子带系数进行如下处理：(1) set

are the mean values of the low-frequency sub-band coefficients of the visible light image and the infrared image, respectively. First, the original low-frequency sub-band coefficients are processed as follows:

(2) Then for the processed coefficients w ₁ and w ₂ , we design the following fusion rules:

w=(w ₂ -w ₁ )×0.5+0.5

where w is the final low-frequency subband fusion coefficient;

为融合后的低频子带图像。In the formula,

is the fused low-frequency subband image.

Step S4. The method of improving the Gaussian difference for the high-frequency sub-band, and the fusion steps are as follows:

分别为可见光图像与红外图像第l层第k个方向上的高频子带系数的均值，首先进行如下处理：(1) set

are the mean values of the high-frequency subband coefficients in the k-th direction of the lth layer of the visible light image and the infrared image, respectively. First, the following processing is performed:

In the formula, a ₁ and a ₂ are preliminary weight coefficients;

(2) Gaussian filtering is performed on the original high-frequency subband coefficients:

In the formula, b ₁ and b ₂ are the high-frequency subband images after Gaussian filtering, G is the Gaussian filter used to smooth the image, the template size hsize is 11×11, and the standard deviation sigma is 5;

(3) Finally, we set the following fusion rules:

s ₁ =b ₁ -a ₁

s ₂ =b ₂ -a ₂

In the formula, s ₁ and s ₂ are the final fusion weights, and we fuse them according to the following formula:

为融合后的高频子带系数。In the formula,

is the fused high-frequency subband coefficients.

The formula for mapping the target saliency map to the fused low-frequency subband image is:

Step S5. The fused image subband coefficients after performing NSCT inverse transformation on the fused high and low frequency subband coefficients are:

In the formula, f _F (x, y) is the complete image after fusion.

The present invention uses the following six objective evaluation indicators for image fusion to verify the effectiveness of the fusion method.

(1) Mutual Information (MI)

Mutual information is used to calculate how much information from the source image is transferred to the fused image. The larger the mutual information value, the better the fusion effect. Defined as follows:

MI = MI _AF + MI _BF

where P _A (a) and P _B (b) are the edge probability densities of source images A and B, respectively, and P _F (f) is the probability density of fusion image F. P _FA (f, a) and P _FB (f, b) are the joint probability densities of the fusion image F and the source images A and B, respectively. MI _AF and MI _BF are the mutual information between the source image and the fused image.

(2) Average Gradient (AG)

The average gradient reflects the ability of the fused image to express small detail contrasts and texture changes, as well as the sharpness of the image. The larger the average gradient value, the clearer the fused image. Defined as follows:

In the formula, ΔI _x =f(x, y)-f(x-1, y), and ΔI _y =f(x, y)-f(x, y-1).

(3) Gray standard deviation (Standard Deviation, SD)

The gray standard deviation reflects the dispersion of the image gray value relative to the average gray value. The larger the standard deviation, the more scattered the gray level distribution, the greater the image contrast, and the more information available. Defined as follows:

为融合图像的灰度平均值。In the formula,

is the gray average value of the fused image.

(4) Information Entropy (Information Entropy, IE)

Information entropy can measure the richness of image information. The larger the information entropy, the richer the information contained in the fused image, and the better the fusion effect. Defined as follows:

In the formula, L represents the number of gray levels, and p _i is the distribution probability of each gray level.

(5) Spatial Frequency (SF)

Spatial frequency reflects the overall activity of the spatial domain of the fused image, and can reflect the ability of the fused image to describe the contrast of small details. The larger the SF, the clearer the fused image.

in,

(6) Visual Information Fidelity for Fusion (VIFF)

VIFF can measure the amount of information in the source image in the fused image, and can also accurately reflect the degree of distortion and enhancement in the fused image. The larger the VIFF value, the better the image fusion effect.

In order to illustrate the superiority of the present invention, the method proposed by the present invention is compared and analyzed with 7 existing classical methods, including: mean method, discrete wavelet transform (DWT), principal component analysis (PCA), gradient pyramid transform method (Grad), methods based on anisotropic diffusion and Calhoun-Loeff transform (KL), methods based on adaptive sparse representation (ASR), and methods based on multi-sensor image fusion (EDF) based on fourth-order partial differential equations .

The present invention provides two embodiments for illustration, and uses the above-mentioned six kinds of objective evaluation indexes to evaluate the fused image. The simulation is carried out using Matlab software.

Example 1: The "Camp" image used in this experiment is shown in Fig. 6, Fig. 6(a) is a visible light image, and Fig. 6(b) is an infrared image. The experimental results are shown in Table 1:

Table 1 "Camp" image fusion evaluation index

Example 2: The "Trees" image used in this experiment is shown in Figure 7, Figure 7(a) is a visible light image, and Figure 7(b) is an infrared image. The experimental results are shown in Table 2:

Table 2 "Trees" image fusion evaluation index

From the results in Tables 1 and 2, it can be seen that compared with the other seven classical fusion algorithms, the fusion method of the present invention has the best performance in terms of objective evaluation indicators, especially the gray standard deviation (SD) and spatial frequency (SF). and Visual Information Fidelity (VIFF), the three indicators perform more prominently and significantly outperform other methods. That is to say, the method of the present invention not only highlights the infrared target, but also takes into account the processing of texture details, which is more conducive to human eye observation, and the image quality after fusion is higher. The six objective evaluation indexes selected in the above two embodiments are to evaluate the fused image from different angles such as information amount, image clarity and visual effect. The evaluation results are comprehensive and can fully illustrate the superiority of the method of the present invention.

The embodiments of the present invention have been described in detail above. However, the present invention is not limited to the above-mentioned embodiments, and various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims (6)

1. a kind of infrared and visible light image fusion method based on gray level co-occurrence matrix, is characterized in that comprising the following steps: Step S1: perform grayscale co-occurrence matrix analysis on the infrared image _I2 , extract the infrared target, and obtain the target saliency map; Step S2: performing NSCT decomposition on the visible light image I ₁ and the infrared image I ₂ to obtain a low-frequency sub-band image and a series of high-frequency sub-band images respectively; Step S3: fuse all low-frequency sub-band images to maintain contrast information to obtain a fused low-frequency image, and use an improved Gaussian difference method to fuse all high-frequency sub-bands to obtain a fused high-frequency sub-band image; Step S4: mapping the target saliency map to the fused low-frequency subband image; Step S5: Perform inverse NSCT transformation on the fused low-frequency sub-band and high-frequency sub-band to obtain a fused image. 2. The infrared and visible light image fusion method based on grayscale co-occurrence matrix as claimed in claim 1, is characterized in that: the method for extracting infrared image target in described step S1 is as follows: (1) Carry out preliminary target extraction, and take the absolute value of the difference between the pixel value of the source infrared image and its gray mean value as the preliminary target extraction image SalPre; (2) Calculate the gray-level co-occurrence matrix coMat of SalPre, which is a symmetric matrix; if a and b are the gray-level values of the SalPre image, then (a, b) is a gray-level value pair; in the gray-level co-occurrence matrix The value of an element of is each pixel value a, and the number of pixel values b in the neighborhood range of its size w, in the present invention, take w=3; (3) Process the gray-level co-occurrence matrix coMat to obtain a modified gray-level co-occurrence matrix; specifically, it includes the following steps: first, normalize the gray-level co-occurrence matrix coMat; then use a logarithmic function to process to obtain L _p (a , b); finally, subtract the average value from the elements in the matrix that are greater than the average value, and take zero for the elements less than or equal to the average value, and then obtain the modified grayscale co-occurrence matrix Sal(a, b); (4) Map the modified grayscale co-occurrence matrix to the preliminary target extraction image according to the following formula:

In the formula, U(a, b) is the average value of (a, b) pixel pairs in the w×w neighborhood, SalMap(a, b) is the saliency detection map of the same size as the source image, and then the image is normalized. unification; (5) Combining SalMap and SalPre to obtain the final target extraction image, the specific formula is: SalFinal(a,b)=SalMap(a,b).*SalPre(a,b) Among them, .* represents the multiplication of corresponding values. 3. the infrared and visible light image fusion method based on gray level co-occurrence matrix as claimed in claim 1, it is characterized in that: in described step S2, NSCT decomposition rule is as follows: the number of directions decomposed at every level is 8, and the number of decomposition scales is adaptive Depending on the size of the image, the formula is: l=[log ₂ (min(N _y , N _x ))-7] In the formula, l is the number of decomposition scales, the size of the image is N _y ×N _x , and [ ] is rounded up. 4. the infrared and visible light image fusion method based on gray level co-occurrence matrix as claimed in claim 1, it is characterized in that: the fusion rule of the low frequency subband coefficient in described step S3 is set as follows: the visible light image low frequency subband and infrared The coefficients of the low-frequency subbands of the image are respectively compared with their average values to obtain coefficient matrices w ₁ and w ₂ ; then the fusion weight matrix of the infrared image is w=(w ₂ -w ₁ )×0.5+0.5, and the fusion weight matrix of the visible light image is 1-w; finally, the weight is multiplied by the low-frequency sub-band to obtain the fused low-frequency sub-band coefficient. 5. The infrared and visible light image fusion method based on gray level co-occurrence matrix as claimed in claim 1, it is characterized in that: in described step S3, the high frequency subband coefficient fusion rule is set as follows: for each layer of high frequency subband coefficients , the coefficients of the high-frequency sub-band of the visible light image and the high-frequency sub-band of the infrared image are respectively different from their average values to obtain the preliminary fusion coefficients a ₁ , a ₂ ; then Gaussian filtering is performed on the original high-frequency sub-band coefficients, wherein the filtering template The size is 11×11 and the standard deviation is 5; the filtered high-frequency subband coefficients are b ₁ and b ₂ , and the fusion weight of the visible light image is s ₁ =b ₁ -a ₁ . Similarly, the fusion weight of the infrared image is is s ₂ =b ₂ -a ₂ ; the fused high-frequency coefficients are selected according to the weight. 6. The infrared and visible light image fusion method based on gray-level co-occurrence matrix as claimed in claim 1, it is characterized in that: in described step S4, adopt the method of addition to map the target saliency map to the low-frequency subband image after fusion .

Images