Adaptive Selection Method of Regularization Parameter for Digital Detail Enhancement Based on Guided Filter

Abstract

The present invention discloses a method for adaptively selecting regularization parameters for digital detail enhancement based on guided filtering, including: performing guided filtering on an input image to obtain a background layer image, subtracting the background layer image from the input image to obtain a detail layer image, calculating the local variance of the detail layer image, and performing normalization processing to screen out target regions where the local variance of the detail layer image is lower than a threshold; calculating the local variance of the input image, and taking the mean of the product of the local variance gray value of the input image and the gray value of the target region as the regularization parameter. The present invention calculates corresponding regularization parameters for different infrared images, enabling technicians not to manually adjust the regularization parameters when using digital detail enhancement technology, reducing the operation difficulty, making the digital detail enhancement technology only amplify the details of the target region, significantly improving the quality of infrared images, and enhancing the scene adaptability of the technology.

Description

Technical Field

[0001] The present invention relates to the field of infrared image processing. More specifically, the present invention relates to a method for adaptively selecting regularization parameters for digital detail enhancement based on guided filtering. Background Art

[0002] Infrared images have a large dynamic range, and their image bit width is greater than that of existing displays. Therefore, the original infrared must undergo dynamic range compression before it can be displayed. If the compression method is improper, the detail information in the infrared image will be lost.

[0003] Compared with linear stretching and histogram equalization, digital detail enhancement technology can enhance the details of infrared images while compressing the dynamic range, and is widely used in the industrial field. Digital detail enhancement technology is based on a hierarchical idea, that is, the original infrared image is divided into a background layer and a detail layer by an edge-preserving smoothing filter, and then different layers are processed differently, and finally the processed background layer and detail layer are merged together. In order to ensure real-time performance and suppress gradient flipping, guided filtering is often used as the edge-preserving smoothing filter in digital detail enhancement technology. Among them, the regularization parameter of guided filtering has a great influence on the final processing result of infrared images. However, when actually using digital detail enhancement technology, technicians need to repeatedly select the regularization parameter according to experience many times to achieve the ideal effect, resulting in a cumbersome operation process and poor scene adaptability of the technology. How to adaptively select the regularization parameter according to the scene, reduce the operation difficulty of digital detail enhancement technology, and increase the scene adaptability is still a problem to be explored. Summary of the Invention

[0004] An object of the present invention is to solve at least the above problems and provide at least the advantages described hereinafter.

[0005] Another object of the present invention is to provide a method for adaptively selecting regularization parameters for digital detail enhancement based on guided filtering, which calculates corresponding regularization parameters for different infrared images, so that technicians do not need to manually adjust the regularization parameters when using digital detail enhancement technology, reducing the operation difficulty, making the digital detail enhancement technology only amplify the details of the target area, significantly improving the quality of infrared images, and enhancing the scene adaptability of the technology.

[0006] To achieve these objects and other advantages of the present invention, there is provided a method for adaptively selecting regularization parameters for digital detail enhancement based on guided filtering, including:

[0007] Performing guided filtering on the input image to obtain a background layer image, subtracting the background layer image from the input image to obtain a detail layer image, calculating the local variance of the detail layer image, and performing normalization processing, and screening to obtain a target area where the local variance of the detail layer image is lower than a threshold;

[0008] Calculate the local variance of the input image, and use the mean value of the product of the local variance gray value of the input image and the gray value of the target area as the regularization parameter.

[0009] Preferably, the local variance var of the input image I The mean value of the product of the gray value and the gray value of the target area ObjectMap is the regularization parameter ε.

[0010]

[0011] Preferably, select the filtering radius r, traverse the input image I, and calculate the local variance var of the input image I I ,

[0012]

[0013] where i and j are image coordinates, and m and n are the actual calculation point positions.

[0014] Preferably, select the filtering radius r, traverse the detail layer image dt, and calculate the local variance var of the detail layer image dt dt ,

[0015]

[0016] where i and j are image coordinates, and m and n are the actual calculation point positions.

[0017] Preferably, perform guided filtering on the input image I to obtain the background layer image bg.

[0018] bg = GF r,ε′ (I, I)

[0019] where r is the filtering radius and ε′ is the set regularization parameter.

[0020] Preferably, the normalization process is specifically:

[0021] Find the maximum value max of the local variance var of the detail layer image dt and the minimum value min dt and perform a linear transformation. dt ,

[0022]

[0023] The present invention has at least the following beneficial effects:

[0024] Taking the mean of the local variances of the target region in the infrared image as the regularization parameter, calculating the corresponding regularization parameter for different infrared images, enabling technicians to use the digital detail enhancement technology without manually adjusting the regularization parameter, reducing the operation difficulty, making the digital detail enhancement technology only amplify the details of the target region, significantly improving the quality of the infrared image, and enhancing the scene adaptability of the technology.

[0025] Other advantages, objectives, and features of the present invention will be partially reflected by the following description and partially understood by those skilled in the art through the research and practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a flowchart of the method for adaptively selecting the regularization parameter in a technical solution of the present invention;

[0027] Figure 2 It is a flowchart of the method for obtaining the target region in a technical solution of the present invention;

[0028] Figure 3 It is a result graph of the dynamic range compression of the original infrared image by the regularization parameter in an example of the indoor scene of the present invention;

[0029] Figure 4 It is a result graph of the dynamic range compression of the original infrared image by the regularization parameter in an example of the outdoor scene of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0030] The following further describes the present invention in detail with reference to the accompanying drawings, so that those skilled in the art can implement it according to the description in the specification.

[0031] It should be understood that the terms such as "having", "including", and "comprising" used herein do not exclude the presence or addition of one or more other elements or their combinations.

[0032] It should be noted that the experimental methods described in the following implementation plans are all conventional methods unless otherwise specified, and the reagents and materials can be obtained from commercial channels unless otherwise specified.

[0033] As Figure 1 shown, the present invention provides a method for adaptively selecting the regularization parameter for digital detail enhancement based on guided filtering, including:

[0034] Using the input image I as the guidance image, performing guided filtering on the input image I to obtain the background layer image bg, subtracting the background layer image bg from the input image I to obtain the detail layer image dt, traversing the detail layer image dt for each frame of the image for calculation, and calculating the local variance var of each pixel of the detail layer image dt dt, and perform normalization. Find the maximum and minimum values of all pixel points in an image, calculate the normalization result for each pixel point, search for the target area, and filter out the target areas where the local variance of the detail layer image dt is lower than the threshold. There will be multiple adjacent or non - adjacent areas that meet the judgment criteria;

[0035] Traversing the input image I means performing operations on each frame of the image, and calculating the local variance var of each pixel in the input image I I , and use the local variance var of the input image I I The mean value of the product of the gray - scale value and the gray - scale value of the target area ObjectMap is used as the regularization parameter.

[0036] The specific implementation of the above technical solution is as follows:

[0037] Step 1, calculate the local variance var of the input image I I :

[0038] Select the filtering radius r, let the filtering radius r = 1, traverse the input image I, and calculate the local variance var of each pixel in the input image I within the filtering radius r I :

[0039]

[0040] Among them, i and j are image coordinates, m and n are the positions of the actually calculated points, m ranges from i - r to i + r, and n ranges from j - r to j + r.

[0041] Step 2, as Figure 2 shown, obtain the target area;

[0042] Step 2.1, perform guided filtering on the input image I to obtain the background layer image bg.

[0043] Use the input image I as the guidance image, perform guided filtering on the input image I to obtain the background layer image bg, that is

[0044] bg = GF r,ε′ (I, I) Equation (2)

[0045] In the formula, r is the filtering radius; ε′ is the regularization parameter, and a sufficiently small value should be selected;

[0046] Among them, the filtering radius r = 1, the regularization parameter ε′ = 1.0, and taking 1 as the initial value for iteration, the actual value will be obtained after iteration;

[0047] Step 2.2, subtract the background layer image bg from the input image I to obtain the detail layer image dt

[0048] dt = I - bg Equation (3)

[0049] Step 2.3, calculate the local variance of the detail layer:

[0050] Select the filtering radius r, where r = 1. Traverse the detail layer image dt, and calculate the local variance var of each pixel in the detail layer image dt within the filtering radius r dt ,

[0051]

[0052] where i and j are the image coordinates, and m and n are the actual calculation point positions.

[0053] Step 2.4, normalize the local variance of the detail layer:

[0054] The specific normalization process is as follows:

[0055] Find the maximum value max dt and the minimum value min dt of the local variance var of the detail layer image dt. Perform a linear transformation on the local variance var of the detail layer image dt dt , and map the maximum value max dt and the minimum value min dt to 1 and 0 respectively. dt

[0056]

[0057] Step 2.5, compare each pixel of the local variance var of the detail layer image dt dt with the threshold TH, and obtain the target region ObjectMap with a smaller local variance var of the detail layer image dt dt , where the threshold TH = 0.01.

[0058]

[0059] Step 3, calculate the mean value of all pixels in the entire input image I that meet the requirements of Step 2.5, and calculate the local variance var of the input image I I The mean value of the product of the grayscale value and the grayscale value of the target region ObjectMap is the required regularization parameter ε.

[0060]

[0061] The present invention uses the mean value of the local variance of the target region in the infrared image as the regularization parameter, calculates the corresponding regularization parameter for different infrared images, enables technicians not to manually adjust the regularization parameter when using the digital detail enhancement technology, reduces the operation difficulty, makes the digital detail enhancement technology only amplify the details of the target region, significantly improves the quality of the infrared image, and enhances the scene adaptability of the technology.​​​

[0062] The existing technology selects the regularization parameter by the local variance statistical histogram of the original image (only the local method with statistics less than 6000) , but the disadvantage is that it cannot be applied to the situation where the local variance distribution range is very wide, such as outdoor (local variance ranges from 0 to 10 6 ), so that the local variance statistics histogram cannot count some local variances, resulting in a small regularization parameter, which increases the noise of the enhanced image. Using the method of the present invention, in indoor scenes, the digital detail enhancement technology uses the regularization parameter calculated by the embodiment of the present invention to compress the dynamic range of the original infrared image. Figure 3 As shown, in an outdoor scene, the result of the digital detail enhancement technology using the regularization parameters calculated by the embodiment of the present invention to compress the dynamic range of the original infrared image is as follows: Figure 4 As shown, the present invention enables digital detail enhancement to correctly select matching regularization parameter values for different scene conditions, and has obvious effects for situations where the temperature varies over a wide range.

[0063] The number of devices and processing scales described here are used to simplify the description of the present invention. Applications, modifications and variations of the present invention will be obvious to those skilled in the art.

[0064] Although the embodiments of the present invention have been disclosed as above, they are not limited to the applications listed in the specification and the implementation modes, and they can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily implemented. Therefore, without departing from the general concept defined by the claims and the scope of equivalents, the present invention is not limited to the specific details and the illustrations shown and described herein.

Claims

1. A method for adaptively selecting regularization parameters for digital detail enhancement based on guided filtering, characterized in that Including: Performing guided filtering on the input image to obtain a background layer image, subtracting the background layer image from the input image to obtain a detail layer image, calculating the local variance of the detail layer image, and performing normalization processing to screen out target regions where the local variance of the detail layer image is lower than a threshold; Calculating the local variance of the input image, and taking the mean of the product of the local variance gray value of the input image and the target region as the regularization parameter; Calculate the local variance var of the input image I I Specifically: Select the filtering radius r, traverse the input image I, and calculate the local variance var of the input image I I , where i and j are image coordinates, and m and n are the actual calculated point positions; Calculate the local variance var of the detail layer image dt dt Specifically: Select the filtering radius r, traverse the detail layer image dt, and calculate the local variance var of the detail layer image dt dt , where i and j are image coordinates, and m and n are the actual calculated point positions; Local variance var of the input image I The mean value of the product of the gray value and the target region ObjectMap is the regularization parameter ε 2. The method for adaptively selecting regularization parameters for digital detail enhancement based on guided filtering according to claim 1, wherein Performing guided filtering on the input image I to obtain a background layer image bg, bg = GF r,ε′ (I, I) where r is the filtering radius and ε′ is the set regularization parameter.

3. The method for adaptively selecting regularization parameters for digital detail enhancement based on guided filtering according to claim 1, characterized in that, The normalization processing specifically is: Find the local variance var of the detail layer image dt of the maximum value max dt and the minimum value min dt , linear transformation

Images