An infrared image sharpening method based on detail preservation

Abstract

The application discloses an infrared image sharpening method based on detail preservation, which comprises the following steps: reading an infrared scene image and a background image, and correcting to obtain a corrected image nucimg; performing spatial filtering on the image nucimg to obtain a filtered image nimg; adopting a hybrid dimming algorithm on the image nimg to obtain a hybrid dimming image Mimg; performing image segmentation on the image nimg to obtain a detail image Dimg; performing measurement on the image Dimg by using a noise suppression function, and performing noise suppression and detail preservation operations on the extracted detail image to obtain a final detail image Eimg; and superimposing the image Eimg on the image Mimg to obtain a final sharpened image. The application solves the problem of detail loss in the infrared image sharpening process, and preserves the original image details.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of infrared image processing, in particular to an infrared image sharpening method based on detail preservation. BACKGROUND

[0002] Infrared imaging technology has the following advantages: high temperature detection sensitivity, can distinguish 100mk or even lower temperature difference; good concealment, not easy to be affected by external electromagnetic interference; high reliability, can work all-weather at night and in very bad weather environment, with strong environmental adaptability. Due to these unique advantages, infrared imaging technology has been widely used in military applications and national social economy.

[0003] With the continuous improvement of the imaging mechanism of infrared image itself, detector material and sensitivity, infrared image has the characteristics of high imaging dynamic range, low contrast, image blur and non-uniformity, etc. Therefore, image sharpening of infrared image can effectively improve image contrast, suppress image noise and improve image visual effect. However, in the current sharpening process of infrared image, a lot of image details are often lost. SUMMARY

[0004] In order to solve the problem of detail loss in the sharpening process of infrared image, the present application provides an infrared image sharpening method based on detail preservation, which comprises the following steps:

[0005] S101: reading infrared scene image and background image, and correcting to obtain corrected image nucimg;

[0006] S102: performing spatial filtering on the image nucimg to obtain the filtered image nimg;

[0007] S103: using a hybrid dimming algorithm on the image nimg to obtain a hybrid dimming image Mimg;

[0008] S104: performing image segmentation on the image nimg to obtain a detail image Dimg;

[0009] S105: using a noise suppression function to measure the image Dimg, and performing noise suppression and detail preservation operation on the extracted detail image to obtain a final detail image Eimg;

[0010] S106: superimposing the image Eimg on the image Mimg to obtain a final sharpening image.

[0011] The present application provides the beneficial effect of solving the problem of detail loss in the sharpening process of infrared image, and preserving the original image details. BRIEF DESCRIPTION OF DRAWINGS

[0012] Figure 1 is a schematic diagram of the method flow of the present application. DETAILED DESCRIPTION

[0013] In order to make the purpose, technical scheme and advantages of the present application more clear, the embodiments of the present application will be further described below with reference to the drawings.

[0014] Please refer to Figure 1 , Figure 1 is a schematic diagram of the method flow of the present application.

[0015] The present application provides an infrared image sharpening method based on detail preservation, which comprises the following steps:

[0016] S101: reading an infrared scene image and a background image, and correcting to obtain a corrected image nucimg;

[0017] It should be noted that in step S101, the correction is specifically two-point correction, as follows:

[0018] nucimg=(img-base)*K (1)

[0019] Wherein, img is an infrared scene image collected by aligning the scene, base represents the background image, and K is a correction coefficient.

[0020] S102: performing spatial filtering on the image nucimg to obtain a filtered image nimg;

[0021] It should be noted that in step S102, the spatial filtering specifically refers to non-local mean filtering.

[0022] In the non-local mean filtering algorithm, three parameters need to be manually adjusted: noise suppression coefficient h, sliding block M*M and search neighborhood window S*S.

[0023] In the present method, in order to reduce the calculation amount in FPGA and save resources, the sliding block is fixedly set to 3*3, and the search neighborhood window is fixedly set to 7*7, which remain unchanged.

[0024] The noise suppression coefficient h is then adaptively adjusted, so that it is automatically adjusted according to the scene without manual adjustment, and the specific calculation is shown in formula (2):

[0025]

[0026]

[0027] Wherein, Cner is a convolution kernel, Conv2 is a two-dimensional convolution operation on the image, conv_img(i,j) represents the pixel value of the i-th row and j-th column, and μ represents the mean value.

[0028] S103: Adopting a hybrid dimming algorithm on the image nimg to obtain a hybrid dimming image Rimg.

[0029] It should be noted that in step S103, the hybrid dimming algorithm is specifically:

[0030] S201: Filtering out valid gray values from the image nimg by using an image binarization operation.

[0031] The image is subjected to a binarization operation, which is specifically shown in formula (3):

[0032]

[0033] Wherein, nx represents the image gray value. The purpose of the threshold T is to improve the overall contrast, which can reduce the influence of small abnormal values and change the global intensity of the display. When nx is greater than the threshold, it should be retained in the output. In the method of the present application, T is selected as one-third of the sum of all pixel values and the mean value of all pixels.

[0034] S202: Obtaining a dimmed image Aimg by using histogram equalization.

[0035] It should be noted that the pixel values of 0 in the image are removed, histogram statistics are performed, the maximum and minimum gray values of the image are calculated with a cutoff threshold of 0.01, and linear mapping is performed to obtain the dimmed image Aimg.

[0036] S203: Enhancing the contrast of the image nimg by using a limited contrast histogram to obtain an image Cimg.

[0037] It should be noted that the image Cimg is obtained by using CLAHE to enhance the contrast of the filtered image.

[0038] S204: Weighted addition of the image Aimg and the image Cimg to obtain a hybrid dimming image Mimg.

[0039] In step S103, the dimmed image and the contrast-enhanced image are weighted and added to obtain Mimg, which is specifically shown in formula (4).

[0040] Mimg=w*Aimg+(1-w)*Cimg (4)

[0041] w is a weight parameter, which can be preset.

[0042] S104: image segmentation is performed on the image nimg to obtain an image detail image Dimg;

[0043] It should be noted that the image filtering is subjected to image segmentation to obtain image details, and specifically, the image after two-point correction (nuc) is subtracted from the image after filtering to obtain a detail image.

[0044] S105: the image Simg is measured by using a noise suppression function, noise suppression and detail preservation operations are performed on the extracted detail image to obtain a final detail image Eimg;

[0045] In step S105, the image filtering is subjected to image segmentation to obtain image details, and the image after nuc is subtracted from the image after filtering to obtain a detail image.

[0046] Generally, the extracted details belong to high-frequency information, and the high-frequency information contains both targets and noise, so noise suppression needs to be performed on the extracted details while the image details are preserved, and then the noise visibility function is defined as:

[0047]

[0048]

[0049] M(i,j)=max(weight(Ω)) (5)

[0050] wherein f(i,j) represents noise sensitivity, wherein theta is an adjustable parameter. Omega represents a neighborhood pixel set of the current point, std(Omega) represents a local standard deviation of the current point, The Euclidean distance between the current point and the points in the neighborhood pixel set is represented by d(i,j), weight(Omega) represents the neighborhood pixel weight, and max(weight(Omega)) takes the maximum value of the neighborhood weight, and in the method, Omega takes a 21*21 image block centered at (i,j).

[0051] Then the final detail image is calculated as shown in formula (6):

[0052] E(i,j)=(1-f(i,j))*D(i,j) (6)

[0053] S106: the image Eimg is superimposed on the image Mimg to obtain a final sharpened image.

[0054] In step 106, the image Mimg after hybrid dimming and the improved detail layer image Eimg are combined according to a certain weight to obtain a final sharpened image.

[0055] The present application has the beneficial effects of solving the problem of detail loss in the infrared image sharpening process and preserving the original image details.

[0056] The above description is only the preferred embodiment of the present application, and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A detail-preserving infrared image sharpening method, characterized in that: Includes the following steps: S101: Reads the infrared scene image and background image, performs correction, and obtains the corrected image. ; S102: Image Spatial filtering is performed to obtain the filtered image. ; S103: Image A hybrid dimming algorithm is used to obtain the image after hybrid dimming. Mimg ; S104: For the image Image segmentation to obtain image details ; S105: Using a noise suppression function to process the image The extracted detail image is then quantified, and noise suppression and detail preservation operations are performed to obtain the final detail image. Eimg ; S106: Transfer image Eimg Overlay on image The final sharpened image is obtained; In step S103, the hybrid dimming algorithm specifically includes: S201. Perform image binarization operation on the image. Filter out the valid grayscale values; S202. Obtain the adjusted image by using histogram equalization. ; S203, Regarding the image Image contrast enhancement is achieved using a limited contrast histogram. ; S204, Image and images The image after weighted summation is obtained by mixing and dimming. Mimg ; In step S201, the image is binarized, as shown in the following formula: in, nx This represents the image grayscale value; the threshold is... T， It is the sum of all pixel values. all , and the mean of the sum of all pixel values all_aver 1 / 3 of the sum; In step S105, the noise suppression function is as follows: （5） in The noise sensitivity is represented by θ, which is an adjustable parameter. It represents the set of neighboring pixels of the current point. This represents the local standard deviation with respect to the current point. This represents the Euclidean distance between the current point and the points in the neighboring pixel set. Indicates the weight of neighboring pixels. Take the maximum value of the neighborhood weight; In step S105, the final detailed image Eimg The calculation formula is as follows: ： （6） in E ( i , j (Image) Eimg One pixel, D ( i , j (Image) Dimg A single pixel.

2. The infrared image sharpening method based on detail preservation as described in claim 1, characterized in that: In step S101, the correction specifically employs a two-point correction, as shown in the following formula: （1） in, img It involves capturing infrared scene images of the scene. base Represents the background image. K It is the correction factor.

3. The infrared image sharpening method based on detail preservation as described in claim 1, characterized in that: In step S102, the spatial domain filtering specifically refers to nonlocal mean filtering.

4. The infrared image sharpening method based on detail preservation as described in claim 3, characterized in that: The nonlocal mean filtering includes three adjustment parameters, namely the noise suppression coefficient. h Sliding block M * M and the search neighborhood window S * S The slider and search neighborhood window are preset values; the noise suppression coefficient is... h Adaptive parameters are used.

5. The infrared image sharpening method based on detail preservation as described in claim 1, characterized in that: Step S106 specifically involves: transferring the image Mimg and images Eimg The image is then merged according to certain weights to obtain the final sharpened image.

Images