An infrared image enhancement method based on double histogram and laplacian sharpening

Abstract

The application discloses an infrared image enhancement method based on average preserving dual histogram equalization and Gaussian Laplacian sharpening fusion. The method first constructs an infrared image enhancement system, the constructed system comprises an average preserving dual histogram equalization module, a Gaussian Laplacian sharpening module and an image fusion module, the average preserving dual histogram equalization module is used for performing average preserving dual histogram equalization on an infrared image to obtain an equalized image, the Gaussian Laplacian sharpening module is used for sharpening the infrared image to obtain a sharpened image, and the image fusion module is used for weightedly fusing the equalized image and the sharpened image to obtain an enhanced infrared image. The application realizes contrast enhancement of the infrared image, enhances the detail information in the infrared image, and solves the problems of excessive enhancement of noise and background in the image.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of image processing, and particularly to an infrared image enhancement method based on double histogram and Laplacian sharpening. BACKGROUND

[0002] As an important branch of infrared image processing technology, infrared image enhancement technology has a very important role in improving the quality of infrared imaging. However, the traditional histogram equalization method is prone to over-enhancing the background and noise. The traditional histogram method uses a single conversion function to change the distribution range of all gray values in the original image, so that the image gray probability density function almost obeys uniform distribution. The image enhancement is achieved by increasing the gray dynamic range of the image. Since the equalization is operated on the global image, and the target and background are stretched according to the same degree criterion, these processes lead to a decrease in the number of gray levels, a decrease in information entropy, and a tendency to underexposure or overexposure, that is, over-enhancement of the background and noise and loss of local details. In addition, the traditional histogram method does not consider the edge information of the image and cannot be used for detail enhancement. SUMMARY

[0003] The purpose of the present application is to provide an infrared image enhancement method based on double histogram and Laplacian sharpening, which solves the problem of over-enhancement of the background and noise caused by the traditional enhancement method.

[0004] An infrared image enhancement method based on double histogram and Laplacian sharpening, the method comprising:

[0005] obtaining an infrared image and attribute information of the infrared image;

[0006] based on the attribute information, performing average-preserving double histogram equalization on the infrared image to obtain an equalized image;

[0007] performing sharpening processing on the infrared image to obtain a sharpened image;

[0008] performing weighted fusion on the equalized image and the sharpened image to obtain an enhanced infrared image.

[0009] In one embodiment, the average-preserving double histogram equalization on the infrared image to obtain an equalized image comprises:

[0010] calculating the average gray value of the infrared image according to the height h and the width w of the infrared image;

[0011] dividing the infrared image into a low gray value image and a high gray value image according to the average gray value of the infrared image;

[0012] obtain a first conversion function according to the low gray value image cumulative probability density function, and obtain a second conversion function according to the high gray value image probability density function;

[0013] obtain a first conversion function according to the low gray value image cumulative probability density function, and obtain a second conversion function according to the high gray value image probability density function;

[0014] obtain a first conversion function according to the low gray value image cumulative probability density function, and obtain a second conversion function according to the high gray value image probability density function;

[0015] obtain a first conversion function according to the low gray value image cumulative probability density function, and obtain a second conversion function according to the high gray value image probability density function;

[0016] obtain a first conversion function according to the low gray value image cumulative probability density function, and obtain a second conversion function according to the high gray value image probability density function.

[0017] In one embodiment, the average gray value of the infrared image is calculated according to the height h and the width w of the infrared image by formula (1);

[0018]

[0019] wherein I mean is the average gray value of the infrared image, I(x, y) is the gray value of the pixel point at position (x, y) in the infrared image I, x is the row position of the pixel point, and y is the column position of the pixel point.

[0020] In one embodiment, the infrared image is divided into a low gray value image and a high gray value image according to the average gray value of the infrared image by formula (2);

[0021]

[0022]

[0023] wherein I L (x, y) is the low gray value image, and I U (x, y) is the high gray value image.

[0024] In one embodiment, the probability density functions of the low gray value image and the high gray value image are respectively obtained by formula (3), and the low gray value image probability density function and the high gray value image probability density function are obtained,

[0025] p l(r1) = hist(r1) / n l , r1 = 0... I mean

[0026] p u (r2) = hist(r2) / n u , r2 = I mean + 1... L - 1... equation (3)

[0027] where p l (r1) is the probability density function of the low gray value image I L , p u (r2) is the probability density function of the high gray value image, hist(r1), hist(r2) represent the number of pixels of the low gray value image I L , high gray value image I U with gray value r1, r2 respectively, n l , n u are the total number of pixels in the low gray value image I L , high gray value image I U respectively, and L is the gray level of the image I.

[0028] In one embodiment, the low gray value image cumulative probability density function is calculated according to the low gray value image probability density function by equation (4), and the high gray value image cumulative probability density function is calculated according to the high gray value image probability density function,

[0029]

[0030]

[0031] c l (r1) is the low gray value image cumulative probability density function, c u (r2) is the high gray value image cumulative probability density function.

[0032] In one embodiment, the first conversion function is obtained according to the low gray value image cumulative probability density function by equation (5), and the second conversion function is obtained according to the high gray value image probability density function,

[0033] s l (r1) = I mean · c l (r1)

[0034] s u (r2) = I mean + (L - 1 - I mean ) · c u (r2)... equation (5)

[0035] s l (r1) is a first conversion function, s u (r2) is a second conversion function.

[0036] In one embodiment, the low gray value image is equalized according to the first conversion function by formula (6) to obtain a low gray value equalized image, the high gray value image is equalized according to the second conversion function to obtain a high gray value equalized image; the low gray value equalized image and the high gray value equalized image are merged by formula (7) to obtain an equalized image,

[0037] I BBHEL (x,y) = s l (I L (x,y))

[0038] I BBHEU (x,y) = s u (I U (x,y)) …… formula (6)

[0039] I BBHE (x,y) = I BBHEL (x,y) ∪ I BBHEU (x,y) …… formula (7)

[0040] I BBHEL (x,y) is a low gray value equalized image, I BBHEU (x,y) is a high gray value equalized image, I BBHE (x,y) is an equalized image.

[0041] In one embodiment, the sharpening processing of the infrared image to obtain a sharpened image comprises:

[0042] A Gaussian Laplace filter f LG (m,n) is constructed by formula (8):

[0043]

[0044]

[0045] Wherein σ>0, σ is a spatial standard deviation parameter of the Gaussian Laplace filter, when σ is close to 1, relatively fine edge features in the image can be detected; the height and width of the filter are 2h f +1, 2w f +1, respectively, f LG (m,n) is the value of the filter at position (m,n), wherein m and n are the row and column positions of the elements in the filter, respectively, and satisfy 1≤m≤2hf +1, 1≤n≤2w f +1.

[0046] through a Gaussian Laplacian filter f LG (m,n) to the infrared image, the filtered image I LGF (x,y) is obtained through equation (9).

[0047] I LGF (x,y) = conv(I(x,y), f LG (m,n)) …… equation (9)

[0048] wherein conv represents convolution, i.e. filtering operation of I(x,y) with filter f LG (m,n);

[0049] The filtered image is superimposed according to the sharpening coefficient through equation (10) to obtain the sharpened image I LGS (x,y):

[0050] I LGS (x,y) = I(x,y) + aI LGF (x,y) …… equation (10)

[0051] wherein a>1 is the sharpening coefficient.

[0052] In one embodiment, the weighting fusion of the equalized image and the sharpened image to obtain the enhanced infrared image comprises:

[0053] The sharpened image I LGS (x,y) obtained through equation (11) is fused with the equalized image I BBHE (x,y) at pixel level to obtain the enhanced infrared image I BBHELG (x,y):

[0054] I BBHELG (x,y) = βI BBHE (x,y) + (1-β)I LGS (x,y) …… equation (11)

[0055] wherein 0≤β≤1 is the fusion coefficient, β=0 is the enhanced image obtained only by Gaussian Laplacian sharpening, β=1 is the enhanced image obtained only by average histogram equalization, β is selected according to the actual infrared image characteristics, and it is usually recommended that β=0.5.

[0056] The application realizes the improvement of image contrast and target signal-to-noise ratio in the image by appropriately enhancing the original infrared image through the enhancement means, selectively emphasizing and highlighting the region of interest in the scene, removing or suppressing the unimportant part in the image, not only can increase the effective distance of the infrared detection imaging system, but also can provide more accurate information for the subsequent target detection and recognition task. Since the infrared image usually has low contrast, the gray value distribution is relatively concentrated, and the gray level distribution is not rich, the brightness preservation characteristics of the average preserving double histogram equalization and the detail and edge enhancement characteristics of the Gaussian Laplace sharpening are combined, and the equalization and sharpening results of the infrared image are fused at the pixel level, so that the infrared image is more effectively and reasonably enhanced. The application adopts the double histogram technology, combines the brightness preservation characteristics of the average preserving double histogram equalization and the detail and edge enhancement characteristics of the Gaussian Laplace sharpening, and fuses the equalization and sharpening results of the infrared image at the pixel level, so that the infrared image is more effectively and reasonably enhanced. BRIEF DESCRIPTION OF DRAWINGS

[0057] Figure 1 It is a flowchart of an infrared image enhancement method based on double histogram and Laplace sharpening. DETAILED DESCRIPTION

[0058] The application will be described in detail below in combination with the drawings and specific embodiments, and the advantages and characteristics of the application will be more apparent according to the following description and claims.

[0059] Although the application can be extended in various forms of modification and replacement, some specific implementation examples are listed in the specification and described in detail; it should be understood that the starting point of the inventor is not to limit the application to the specific examples described, on the contrary, the starting point of the inventor is to protect all improvements, equivalent replacements and modifications made within the spirit or scope defined by the present right declaration.

[0060] In one embodiment, please refer to Figure 1 An infrared image enhancement method based on double histogram and Laplace sharpening, the method comprises:

[0061] S110, acquiring an infrared image and attribute information of the infrared image;

[0062] S120, based on the attribute information, performing average preserving double histogram equalization on the infrared image to obtain an equalized image;

[0063] S130, performing sharpening processing on the infrared image to obtain a sharpened image;

[0064] S140, performing weighted fusion on the equalized image and the sharpened image to obtain an enhanced infrared image.

[0065] In one embodiment, the averaging the infrared image keeps the double histogram equalization, and the equalization image comprises:

[0066] According to the height h and the width w of the infrared image, an average gray value of the infrared image is calculated;

[0067] According to the average gray value of the infrared image, the infrared image is divided into a low gray value image and a high gray value image;

[0068] Probability density functions of the low gray value image and the high gray value image are respectively calculated to obtain a low gray value image probability density function and a high gray value image probability density function;

[0069] According to the low gray value image probability density function, a low gray value image cumulative probability density function is calculated, and according to the high gray value image probability density function, a high gray value image cumulative probability density function is calculated;

[0070] According to the low gray value image cumulative probability density function, a first conversion function is obtained, and according to the high gray value image probability density function, a second conversion function is obtained;

[0071] According to the first conversion function, the low gray value image is equalized to obtain a low gray value equalization image, and according to the second conversion function, the high gray value image is equalized to obtain a high gray value equalization image;

[0072] The low gray value equalization image and the high gray value equalization image are combined to obtain an equalization image.

[0073] In one embodiment, according to the height h and the width w of the infrared image, the average gray value of the infrared image is calculated by formula (1);

[0074]

[0075] wherein, I mean is the average gray value of the infrared image, I(x, y) is the gray value of the pixel point at position (x, y) in the infrared image I, x is the row position of the pixel point, and y is the column position of the pixel point.

[0076] In one embodiment, according to the average gray value of the infrared image, the infrared image is divided into a low gray value image and a high gray value image by formula (2);

[0077]

[0078]

[0079] wherein I L (x,y) is a low gray value image, I U (x,y) is a high gray value image.

[0080] In one embodiment, the probability density functions of the low gray value image and the high gray value image are obtained by formula (3) respectively,

[0081] p l (r1) = hist(r1) / n l , r1 = 0...I mean

[0082] p u (r2) = hist(r2) / n u , r2 = I mean +1...L-1...formula (3)

[0083] wherein p l (r1) is the probability density function of the low gray value image I L , p u (r2) is the probability density function of the high gray value image, hist(r1), hist(r2) respectively represent the pixel number of the low gray value image I L , the high gray value image I U with gray value r1, r2, n l , n u respectively are the total pixel number in the low gray value image I L , the high gray value image I U , and L is the gray level of the image I.

[0084] In one embodiment, the cumulative probability density function of the low gray value image is calculated according to the low gray value image probability density function by formula (4), and the cumulative probability density function of the high gray value image is calculated according to the high gray value image probability density function,

[0085]

[0086]

[0087] c l (r1) is the cumulative probability density function of the low gray value image, c u (r2) is the cumulative probability density function of the high gray value image.

[0088] In one embodiment, a first conversion function is obtained according to the probability density function of the low gray value image by formula (5), a second conversion function is obtained according to the probability density function of the high gray value image,

[0089] s l (r1) = I mean · c l (r1)

[0090] s u (r2) = I mean + (L - 1 - I mean ) · c u (r2) …… formula (5)

[0091] s l (r1) is the first conversion function, s u (r2) is the second conversion function.

[0092] In one embodiment, a low gray value equalization image is obtained by equalizing the low gray value image according to the first conversion function by formula (6), a high gray value equalization image is obtained by equalizing the high gray value image according to the second conversion function, and the low gray value equalization image and the high gray value equalization image are merged by formula (7) to obtain an equalization image,

[0093] I BBHEL (x, y) = s l (I L (x, y))

[0094] I BBHEU (x, y) = s u (I U (x, y)) …… formula (6)

[0095] I BBHE (x, y) = I BBHEL (x, y) ∪ I BBHEU (x, y) …… formula (7)

[0096] I BBHEL (x, y) is the low gray value equalization image, I BBHEU (x, y) is the high gray value equalization image, and I BBHE (x, y) is the equalization image.

[0097] In one embodiment, the sharpening processing of the infrared image to obtain a sharpened image comprises:

[0098] A Gaussian Laplace filter f LG (m, n) is constructed by formula (8):

[0099]

[0100]

[0101] where σ>0, σ is a spatial standard deviation parameter of the Laplacian of Gaussian filter, when σ is close to 1, the relatively fine edge features in the image can be detected; the height and width of the filter are 2h and 2w respectively f +1,2w f +1,1≤n≤2w LG (m,n) is the value of the filter at position (m,n), where m and n are the row and column positions of the element in the filter respectively, and satisfy 1≤m≤2h f +1,1≤n≤2w f +1。

[0102] The filtered image I LG (x,y) is obtained by filtering the infrared image with the Laplacian of Gaussian filter f LGF (m,n) according to formula (9)

[0103] I LGF (x,y) = conv(I(x,y),f LG (m,n)) …… formula (9)

[0104] where conv represents convolution, that is, filtering operation of I(x,y) with the filter f LG (m,n);

[0105] The filtered image is superimposed according to the sharpening coefficient to obtain the sharpened image I LGS (x,y) according to formula (10)

[0106] I LGS (x,y) = I(x,y) + aI LGF (x,y) …… formula (10)

[0107] where a>1 is a sharpening coefficient.

[0108] In one embodiment, the weighting fusion of the equalized image and the sharpened image to obtain the enhanced infrared image comprises:

[0109] The sharpened image I LGS (x,y) obtained is fused with the equalized image I BBHE (x,y) at the pixel level according to formula (11) to obtain the enhanced infrared image I BBHELG (x,y):

[0110] I BBHELG (x,y) = βI BBHE(x, y) + (1 - β)I LGS (x, y) …… equation (11)

[0111] Wherein 0≤β≤1 is a fusion coefficient, when β=0, it is an enhanced image obtained by using only the Laplacian of Gaussian sharpening, when β=1, it is an enhanced image obtained by using only the average histogram equalization, and β is selected according to the actual infrared image characteristics, and it is usually recommended that β=0.5.

[0112] In one embodiment, a computer readable storage medium is provided, and a computer program is stored on the computer readable storage medium, and the computer program is executed by a processor to implement the steps of the method in any one of the above embodiments. It can be understood by those skilled in the art that the processes of implementing all or part of the above-mentioned embodiment methods can be completed by instructing the relevant hardware through the computer program, and the computer program can be stored in a non-volatile computer readable storage medium, and when the computer program is executed, the processes of the above-mentioned embodiments of the method can be included.

[0113] A computer device is provided, and the computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the steps of the method in any one of the above embodiments when executing the computer program. Specifically, the computer device can be a server, and the computer device includes a processor, a memory, a network interface and a database connected by a system bus.

[0114] The present application realizes the contrast of the image and the signal-to-noise ratio of the target in the image by appropriately enhancing the original infrared image through the enhancement means, selectively emphasizing and highlighting the region of interest in the scene, removing or suppressing the unimportant part in the image, not only can increase the effective distance of the infrared detection imaging system, but also can provide more accurate information for the subsequent target detection and recognition task. Since the infrared image usually has low contrast, the gray value distribution is relatively concentrated, and the gray level distribution is not rich, combined with the brightness preservation characteristics of the average preserving double histogram equalization and the detail and edge enhancement characteristics of the Laplacian of Gaussian sharpening, the equalization and sharpening results of the infrared image are fused at the pixel level, and the infrared image is more effectively and reasonably enhanced.

Claims

1. An infrared image enhancement method based on double histogram and Laplacian sharpening, characterized in that, The method comprises: acquiring an infrared image and attribute information of the infrared image; Based on the attribute information, the infrared image is subjected to average-preserving double histogram equalization to obtain an equalized image; The infrared image is subjected to sharpening processing to obtain a sharpened image; The equalized image and the sharpened image are subjected to weighted fusion to obtain an enhanced infrared image, Wherein, the equalization of the infrared image comprises: calculating the average gray value of the infrared image according to the height h and the width w of the infrared image; dividing the infrared image into a low gray value image and a high gray value image according to the average gray value of the infrared image; respectively calculating the probability density functions of the low gray value image and the high gray value image to obtain the probability density function of the low gray value image and the probability density function of the high gray value image; calculating the cumulative probability density function of the low gray value image according to the probability density function of the low gray value image, and calculating the cumulative probability density function of the high gray value image according to the probability density function of the high gray value image; obtaining a first conversion function according to the cumulative probability density function of the low gray value image, and obtaining a second conversion function according to the probability density function of the high gray value image; equalizing the low gray value image according to the first conversion function to obtain a low gray value equalized image, and equalizing the high gray value image according to the second conversion function to obtain a high gray value equalized image; and merging the low gray value equalized image and the high gray value equalized image to obtain the equalized image; The average gray value of the infrared image is calculated according to the height h and the width w of the infrared image through formula (1); where I mean is the average gray value of the infrared image, I(x, y) is the gray value of the pixel point at position (x, y) in the infrared image I, x is the row position of the pixel point, and y is the column position of the pixel point. The infrared image is divided into a low gray value image and a high gray value image according to the average gray value of the infrared image through formula (2); where I L (x,y) is a low gray value image, I U (x,y) is a high gray value image; The probability density functions of the low gray value image and the high gray value image are respectively calculated through formula (3) to obtain the probability density function of the low gray value image and the probability density function of the high gray value image, Where, p l (r1) is the low grayscale image I. L The probability density function, p u (r2) is the probability density function of the high grayscale image, and hist(r1) and hist(r2) represent the low grayscale image I, respectively. L High grayscale image I U The number of pixels with grayscale values ​​r1 and r2 respectively, n l n u Image I, representing low grayscale values L High grayscale image I U The total number of pixels in the image, where L is the gray level of image I; The cumulative probability density function of the low gray value image is calculated according to the probability density function of the low gray value image through formula (4), and the cumulative probability density function of the high gray value image is calculated according to the probability density function of the high gray value image, c l (r1) is a cumulative probability density function of low gray value images, c u (r2) is a cumulative probability density function of high gray value images; The first conversion function is obtained according to the cumulative probability density function of the low gray value image through formula (5), and the second conversion function is obtained according to the probability density function of the high gray value image, s l (r1) is a first conversion function, s u (r2) is a second conversion function.

2. The method of claim 1, wherein, The low gray value equalized image is obtained by equalizing the low gray value image according to the first conversion function through formula (6), and the high gray value equalized image is obtained by equalizing the high gray value image according to the second conversion function; the equalized image is obtained by merging the low gray value equalized image and the high gray value equalized image through formula (7), I BBHE (x,y) = I BBHEL (x,y) U I BBHEU (x,y)... equation (7) I BBHEL (x,y) is a low gray value equalized image, I BBHEU (x,y) is a high gray value equalized image, I BBHE (x,y) is an equalized image.

3. The method of claim 2, wherein, The infrared image is subjected to sharpening processing to obtain a sharpened image, The Gaussian Laplacian filter f is constructed by equation (8) LG (m, n): where σ > 0, σ is a Gaussian Laplacian filter spatial standard deviation parameter, when σ is close to 1, relatively fine edge features in the image can be detected; the height and width of the filter are 2h respectively f +1, 2w f +1, f LG (m, n) is the value of the filter at position (m, n), where m, n are the row and column positions of the element in the filter respectively, and satisfy 1≤m≤2h f +1, 1≤n≤2w f +1; by a gaussian laplacian filter f LG (m, n) pairs of the infrared image, the filtered image I is obtained by formula (9) LGF (x, y): I LGF (x,y) = conv(I(x,y), f LG (m,n)) …… Equation (9) where conv denotes a convolution, i.e. the filtering of I(x,y) with the filter f LG the pair (m, n) filters I(x,y); The filtered image is superimposed according to the sharpening coefficient by formula (10) to obtain a sharpened image I LGS (x,y): I LGS (x,y) = I(x,y) + a I LGF (x,y) …… Equation (10) Wherein, α>1 is a sharpening coefficient.

4. The method of claim 3, wherein, The weighting fusion of the equalization image and the sharpened image obtains an enhanced infrared image, and the weighting fusion comprises the following steps: The sharpened image I obtained by formula (11) LGS (x,y) and the equalized image I BBHE (x,y) are fused at the pixel level to obtain the enhanced infrared image I BBHELG (x,y): I BBHELG (x,y) = βI BBHE (x,y) + (1 - β)I LGS (x,y) …… Equation (11) Wherein 0≤β≤1 is a fusion coefficient, β=0 is an enhanced image obtained by using only Gaussian Laplacian sharpening, β=1 is an enhanced image obtained by using only average histogram equalization, and β is selected according to actual infrared image characteristics.

Images