Image enhancement method and infrared imaging system

Abstract

The application relates to the field of image processing, in particular to an image enhancement method, which specifically comprises the following steps: collecting pixel value information of each pixel point in a background image, and obtaining a noise threshold corresponding to each pixel point in the background image; performing Laplace edge detection on an original input image; performing threshold judgment on a detection image obtained after the Laplace edge detection, removing pixel values of pixel points with pixel values less than or equal to the noise threshold in the detection image, so as to obtain a threshold judgment image; and superimposing the threshold judgment image and the original input image to obtain a superimposed image; so that the image edge can be better enhanced, the noise can be suppressed, the image edge is highlighted, and the image definition is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of image processing, in particular to an image enhancement method and an infrared imaging system. BACKGROUND

[0002] Image enhancement is an important part of the field of digital image processing, and its purpose is to improve the visual quality of the image, making it easier to analyze or enhance specific features. Laplacian sharpening is a commonly used image enhancement method, which extracts edge information in all directions through a second-order derivative operator, that is, the edge information image is obtained through Laplacian edge detection, and then the edge information image and the original input image are superimposed, which is excellent in detecting lines and isolated points.

[0003] However, its excellent performance can also cause noise points to be enhanced, resulting in a decrease in signal-to-noise ratio, referring to Figure 5 , which is the main reason why Laplacian sharpening cannot be directly used in high-noise environments such as infrared image processing. SUMMARY

[0004] In order to denoise and enhance low signal-to-noise ratio images to highlight image edges and improve image clarity, the present application provides an image enhancement method and an infrared imaging system.

[0005] The image enhancement method and the infrared imaging system provided by the present application adopt the following technical solutions.

[0006] An image enhancement method, specifically comprising the following steps.

[0007] S1, acquiring pixel value information of each pixel point in the background image to obtain a noise threshold corresponding to each pixel point of the background image;

[0008] S2, performing Laplacian edge detection on the original input image;

[0009] S3, performing threshold judgment on the detection image obtained after the Laplacian edge detection to remove pixel values of pixel points in the detection image whose pixel values are less than or equal to the noise threshold, to obtain a threshold judgment image;

[0010] S4, superimposing the threshold judgment image and the original input image to obtain a superimposed image.

[0011] By adopting the above technical solutions, the noise threshold is first obtained through the background image, and then the threshold judgment is performed on the detection image to remove non-zero edge data caused by noise in the Laplacian edge detection, and the pixel values of the image itself edge data with larger values and higher probability are retained, so as to realize the denoising and enhancement of low signal-to-noise ratio images, highlight the image edges, and improve the image clarity.

[0012] Optionally, in S1, several frames of background image are collected, and the pixel value information includes the maximum and minimum pixel values ​​experienced by each pixel in the entire background image. The noise threshold is obtained by subtracting the maximum and minimum pixel values ​​counted for each pixel.

[0013] Optionally, the noise threshold The calculation formula is:

[0014]

[0015] in, The number of rows for each pixel. The number of columns for the number of pixels. for OK The maximum pixel value traversed by the column's pixel points. for OK The minimum pixel value traversed by the pixel points of a column.

[0016] By adopting the above technical solution, the noise threshold can be calculated and the possibility of an extreme value causing the noise threshold to be too large, resulting in subsequent threshold judgments treating a large amount of real edge data as noise and removing it, is reduced.

[0017] Optionally, the threshold determination process in S3 is expressed as follows:

[0018]

[0019] in, for OK The pixel values ​​of the column after thresholding. To detect in the image OK The pixel values ​​of the columns of pixels. This is the noise threshold.

[0020] By adopting the above technical solution, the pixel values ​​of pixels that do not meet the standards are changed to 0, making the noise areas turn black, thereby highlighting the image edges.

[0021] Optionally, after step S4 is completed, the superimposed image is subjected to median filtering to obtain the final image.

[0022] By adopting the above technical solution, the threshold judgment image will produce breakpoint protrusions similar to salt and pepper noise due to threshold segmentation. Therefore, median filtering is performed to enhance the sharpness in order to obtain the final displayed image.

[0023] Optionally, the noise threshold value of each pixel point in S1 generates a noise threshold value table and stores it.

[0024] By adopting the technical scheme, threshold value judgment is performed on each pixel point, and high-precision and high-robustness picture enhancement processing is realized.

[0025] Optionally, the background image acquisition mode in S1 includes a blocking lens and / or an equivalent picture input.

[0026] By adopting the technical scheme, the acquisition mode of multiple backgrounds is selected as needed.

[0027] An infrared imaging system operates according to the image enhancement method, and includes an infrared lens module for acquiring an original input image and / or a background image, an analog-to-digital conversion module for converting an analog signal output by the infrared lens module into a digital signal and outputting the digital signal, a chip module for receiving the digital signal output by the analog-to-digital conversion module and outputting the digital signal after processing, and a display module for receiving the signal output by the chip module and displaying an image after processing.

[0028] By adopting the technical scheme, the image displayed by the corresponding infrared imaging system is clearer.

[0029] Optionally, the chip module includes an FPGA chip.

[0030] By adopting the technical scheme, the FPGA chip has parallel processing advantages, can start Laplace edge detection and threshold value judgment on a second frame image after a first frame image is completed, performs pipeline operation, can improve the running efficiency of the algorithm, and enhances real-time performance and convenience.

[0031] In summary, the present application at least includes the following beneficial effects.

[0032] 1. The noise threshold value is acquired from the background image, and then the detection image is subjected to threshold value judgment, so as to remove non-zero edge data caused by noise in the Laplace edge detection, retain pixel values that are relatively large and have a high probability of being edge data of the image itself, realize noise reduction and enhancement on the image with low signal-to-noise ratio, highlight the image edge, and improve the image clarity.

[0033] 2. The median filter processing is performed to enhance the clarity, remove breakpoints protruding like salt-and-pepper noise in the threshold value judgment image due to threshold value segmentation, and obtain a final image with high clarity. BRIEF DESCRIPTION OF DRAWINGS

[0034] Figure 1 is a flowchart of an image enhancement method of the present application;

[0035] Figure 2This is a template form diagram of the Laplacian operator S2 in an image enhancement method;

[0036] Figure 3 In the image, (a) is the detection image and (b) is a comparison of the threshold judgment image.

[0037] Figure 4 In the image, (a) is the overlay image, and (b) is a comparison of the two images in the final image.

[0038] Figure 5 This is a flowchart of the three-step median method;

[0039] Figure 6 In the image, (a) is the original input image and (b) is a comparison of the two images after unmodified Laplacian sharpening. Detailed Implementation

[0040] The present application will be further described in detail below with reference to the accompanying drawings.

[0041] This application discloses an image enhancement method, referring to... Figure 1 Specifically, it includes the following steps.

[0042] S1. Collect the pixel value information of each pixel in the background image and obtain the noise threshold corresponding to each pixel in the background image.

[0043] Background image acquisition methods include lens occlusion and / or equivalent image input. Lens occlusion refers to taking several frames by blocking the camera lens to obtain images under no-light conditions, thereby capturing the background noise. Equivalent image input refers to digitally inputting several frames of images with constant grayscale values ​​into the image processing chip to obtain system-level fixed-pattern noise. Both methods can provide several frames, such as 50-100 frames, to improve the confidence of the noise threshold.

[0044] Pixel value information includes the maximum and minimum pixel values ​​that each pixel has experienced throughout the entire background image. The noise threshold is obtained by subtracting the maximum and minimum pixel values ​​for each pixel. Noise Threshold The calculation formula is as follows.

[0045]

[0046] in, The number of rows for each pixel. The number of columns for the number of pixels. for OK The maximum pixel value traversed by the column's pixel points. for OK The minimum pixel value traversed by the pixel points of a column.

[0047] The noise threshold for each pixel is used to generate a noise threshold table and stored in DDR.

[0048] S2. Perform Laplacian edge detection on the original input image.

[0049] The specific process involves using the Laplacian operator to perform edge detection on the input image. The Laplacian operator is an isotropic, rotation-invariant differential operator, satisfying the requirements for detecting image edges with different orientations. If the Laplacian operator is written in template form, it can be found here. Figure 2 As shown.

[0050] Generally, the image is convolved with the above operators to achieve the edge detection process. For image edge data, padding with zeros or copying neighboring values ​​is used.

[0051] S3. Threshold the detection image obtained after Laplacian edge detection, and remove the pixel values ​​of pixels in the detection image that are less than or equal to the noise threshold to obtain the threshold judgment image.

[0052] The threshold determination process can be represented by the formula below.

[0053]

[0054] in, for OK The pixel values ​​of the column after thresholding. To detect in the image OK The pixel values ​​of the columns of pixels. This is the noise threshold.

[0055] Reference Figure 3 The obtained noise threshold is compared with the data obtained from Laplacian edge detection. Data exceeding the noise threshold is retained as image details, while data below the threshold is considered noise interference and its value is set to 0. This ensures that the pixel values ​​of the final superimposed image remain unchanged, preventing noise enhancement. In summary, thresholding effectively removes noise from the detected image, retaining pixel values ​​of pixels with a higher probability of being edge data, and setting the pixel values ​​of non-zero edge data caused by noise to 0 for smaller value removal. This highlights the edges of the thresholded image and improves its clarity.

[0056] S4. Overlay the threshold judgment image and the original input image to obtain the overlay image.

[0057] Reference Figure 4The superposition process is addition of pixel values, and the edge detection result after addition will make the original edge value of the image larger or smaller, having a highlighting effect. The edge of the superposition image is enhanced, and most noise is greatly suppressed after threshold judgment, avoiding noise enhancement.

[0058] S5, performing median filtering processing on the superposition image to obtain a final image.

[0059] Referring to Figure 5 The median filtering processing is specifically in the form of a sliding block to obtain a 3X3 pixel window, median data is obtained as output by a parallel three-step median method, and is replaced pixel by pixel. The three-step median method is to compare each row of three pixels two by two to obtain a maximum value Max, an intermediate value Mid, and a minimum value Min; then the minimum value of the three maximum values Max_Min, the intermediate value of the three intermediate values Mid_Mid, and the maximum value of the three minimum values Min_Max are calculated; then the intermediate value of Max_Min, Mid_Mid, and Min_Max is calculated, which is the final result, and then the pixel value at the center of the pixel window is replaced by the final result. In addition, the pixel points at the edge of the superposition image cannot become the center of the pixel window, so the corresponding pixel values remain unchanged, and can also be processed according to a preset boundary strategy, such as copying the pixel values of similar pixel points or clearing zero.

[0060] Referring to Figure 6 The final image after median filtering processing is less likely to have the salt and pepper noise in the superposition image, so that the image clarity is further improved. In order to more intuitively see the effect of the image enhancement method, the original image processed by the improved Laplacian sharpening and the image processed by the image enhancement method are quantitatively analyzed in terms of MSE and PSNR.

[0061] MSE is mean square error, that is, the average square value of the pixel value difference between the original image and the processed image, and the lower the MSE value, the more similar to the original image. The MSE calculation formula is as follows.

[0062] MSE = [ ∑ i = 1 w ∑ j = 1 h ( p L ( i , j ) − p Y ( i , j )) 2 ] / h / w ;

[0063] wherein, is the length of the image, is the height of the image, is the data of the i-th row and j-th column image after enhancement, is the data of the i-th row and j-th column original image.

[0064] PSNR is the peak signal-to-noise ratio, the ratio between the maximum power of the signal (original image) and the power of the introduced noise (error), the higher the PSNR value, the smaller the image is disturbed by noise, the higher the image quality. The PSNR calculation formula is as follows.

[0065] ;

[0066] Wherein depth is the bit width of the pixel, generally 8.

[0067] The comparison results of the two indicators of MSE and PSNR of the original image processed by the unimproved Laplace sharpening and the image processed by the image enhancement method of the application are shown in the following table 1.

[0068] Table 1:

[0069]

[0070] The application also discloses an infrared imaging system, which comprises an infrared lens module, an analog-digital conversion module, a chip module and a display module.

[0071] In the embodiment, the infrared lens module can collect images by using a refrigeration type II superlattice detector, and the image resolution can be 640*512. The analog-digital conversion module can select an ADC chip with a model of ADs6442. The chip module can use an FPGA chip with a model of xc7z030sbg485-2 from Xilinx company, and an image enhancement algorithm of the embodiment is programmed by using a Verilog language, is compiled and synthesized by using a vivado software, is burned into the FPGA chip to form a hardware circuit. The image display module can be an SDI monitor, a data stream of a BT1120 protocol is written by using the Verilog language and is output to the SDI monitor for image display.

[0072] The image enhancement method and the infrared imaging system of the embodiment of the application have the following implementation principles: the noise threshold is obtained through a plurality of background images, then the detection image is subjected to threshold judgment, the noise is suppressed and the edge is highlighted, so that the image processed by using the Laplace edge detection has improved definition and can be applied to infrared imaging.

[0073] The above are preferred embodiments of the application, and do not limit the protection scope of the application, therefore: all equivalent changes made according to the structure, shape, principle of the application should be covered in the protection scope of the application.

Claims

1. An image enhancement method, characterized in that: Specifically, the following steps are included: S1. Collect the pixel value information of each pixel in the background image and obtain the noise threshold corresponding to each pixel in the background image; S2. Perform Laplacian edge detection on the original input image; S3. Threshold the detection image obtained after Laplacian edge detection, and remove the pixel values ​​of pixels in the detection image that are less than or equal to the noise threshold to obtain the threshold judgment image. S4. Overlay the threshold judgment image and the original input image to obtain the overlay image; In S1, several frames of background image are collected. The pixel value information includes the maximum and minimum pixel values ​​experienced by each pixel in the entire background image. The noise threshold is obtained by subtracting the maximum and minimum pixel values ​​of each pixel. The noise threshold The calculation formula is: ； in, The number of rows for each pixel. The number of columns for the number of pixels. for OK The maximum pixel value traversed by the column's pixel points. for OK The minimum pixel value traversed by the pixel points of a column.

2. The image enhancement method according to claim 1, characterized in that: The threshold determination process in S3 is expressed as follows: ； in, for OK The pixel values ​​of the column after thresholding. To detect in the image OK The pixel values ​​of the columns of pixels. This is the noise threshold.

3. The image enhancement method according to claim 1, characterized in that: After step S4 is completed, the superimposed image is subjected to median filtering to obtain the final image.

4. The image enhancement method according to claim 1, characterized in that: In S1, the noise threshold of each pixel is used to generate a noise threshold table and then stored.

5. The image enhancement method according to claim 1, characterized in that: The background image acquisition method in S1 includes occluding the lens and / or inputting equivalent images.

6. An infrared imaging system, operating according to an image enhancement method according to any one of claims 1 to 5, characterized in that: It includes an infrared lens module for acquiring the original input image and / or background image, an analog-to-digital converter module for receiving the analog signal output from the infrared lens module, converting it into a digital signal and outputting it, a chip module for receiving the digital signal output from the analog-to-digital converter module, processing it, and outputting it, and a display module for receiving the signal output from the chip module, processing it, and displaying the image.

7. An infrared imaging system according to claim 6, characterized in that: The chip module includes an FPGA chip.

Images