A monolithic diffractive infrared system image processing method

Abstract

The application discloses an image processing method of a monolithic diffractive infrared system, which is based on a local maximum gradient prior mode and used for solving the image degradation and blurring problem of the monolithic diffractive infrared system. Research finds that the maximum gradient in the neighborhood range of a pixel point of system imaging will decrease after the system imaging process, which leads to the degradation of system imaging and conforms to the degradation process of system imaging. The method is characterized in that the local maximum gradient is directly recovered as a prior theory. The method process is that a two-dimensional matrix of a degraded image is input, a target function of system image restoration is established, and finally an iterative solution is estimated to obtain a final clear image. In the analysis of the image degradation process of the monolithic diffractive infrared system, an inherent prior theory is found, and the application restores the image by using the prior, and has the advantages of simple realization, good restoration effect and the like.

Description

Technical Field

[0001] This invention relates to the fields of image processing and optical engineering technology, specifically to the field of image restoration and enhancement of monolithic diffractive infrared systems, and particularly to an image processing method for monolithic diffractive infrared systems based on a local maximum gradient prior approach. Background Technology

[0002] The use of lighter diffractive lenses is a future trend in optical imaging. However, due to manufacturing processes and inherent imaging characteristics, diffractive lenses can affect diffraction efficiency, ultimately leading to image degradation and blurring. This is especially true in monolithic diffraction systems, where the lack of other correction elements exacerbates the image quality degradation. Therefore, real-time image deblurring from the image-end perspective would greatly expand the application scenarios of monolithic diffraction systems.

[0003] In monolithic diffraction infrared systems, the final image is a grayscale image, so there's no need to consider chromatic aberration, making it more suitable for image processing at the image end. The main problem to be solved is image blurring. Currently, the mainstream solution is Wiener filtering. Wiener filtering treats the image as a two-dimensional stationary continuous signal with independent noise and zero mean, aiming to find the minimum mean square error between the estimated sharp image and the original sharp image. However, this method requires accurate knowledge of the degradation process parameters of the optical system's imaging, as well as a good estimate of the ratio of the noise power spectrum to the power spectrum of the ideal image; otherwise, the image restoration effect is far from ideal. Therefore, the characteristics of Wiener filtering limit its application scenarios. Summary of the Invention

[0004] To address the problems existing in the current technology, this invention proposes an image processing method for monolithic diffraction infrared systems. During the analysis of system imaging, it was found that the maximum gradient value within the neighborhood of each pixel in the image decreases after the system imaging process, leading to a degradation in the image quality. Therefore, this is proposed as an inherent prior for system image restoration, and an image processing method for monolithic diffraction infrared systems is designed accordingly.

[0005] The technical solution adopted by the method of the present invention is: an image processing method for a monolithic diffraction infrared system, comprising the following steps:

[0006] Step 1: Input a blurred image from a monolithic diffraction infrared system. The image is a two-dimensional grayscale matrix of size i*j.

[0007] Step 2: Establish the objective function for image restoration. The objective function is as follows:

[0008]

[0009] Among them, the first item In the diagram, I represents the estimated sharp image, and k represents the blur kernel. The first term represents the convolution process, where B represents the input blurred image. This term uses the L2 norm as a constraint to minimize the loss of the blurred image after convolving the sharp image with the blurred kernel. The second term, ||2-LMG(I)||1, represents the local maximum gradient value of the image. This term uses the L1 norm as a constraint to maximize the local maximum gradient of the image. The third term... middle, The fourth term represents the gradient matrix of the image. This term is constrained using the L0 norm to eliminate local small gradient regions while retaining large gradient regions; In this context, the L2 norm is used as a constraint to regularize the solution of the fuzzy kernel; α, β, and λ are the weight parameters, respectively.

[0010] Step 3: Initialize the blur kernel for the system's imaging degradation process. The blur kernel is a two-dimensional matrix of size m*m that causes system degradation. The blur kernel initialization process is as follows: Create an m*m matrix with all values ​​set to 0 according to the input blur kernel size m, and then set the values ​​of the elements at row position (m-1 / 2) and column position (m-1 / 2) to 0.1.

[0011] Step 4: Calculate the local maximum gradient prior of the image and use it as the L1 norm for sparsity constraints. The local maximum gradient refers to the maximum gradient within the neighborhood of each pixel in the image, calculated as follows:

[0012]

[0013] Where x represents the position of any pixel; P(x) represents the neighborhood of pixel x; and y represents the pixel in the neighborhood P(x) of pixel x. This represents the gradient value of pixel y; the gradient represents the sum of the rates of change of a pixel in the horizontal and vertical directions, and is calculated as follows:

[0014]

[0015] Among them, g x and g y These represent the rates of change in the horizontal and vertical directions, respectively.

[0016] Step 5: Decompose the objective function from Step 3 into subproblems. The objective function is decomposed into the following subproblems:

[0017] Formula 1:

[0018] Formula 2:

[0019] Equation 1 is used to solve for the clear image I; Equation 2 is used to solve for the blur kernel k.

[0020] Step Six: Iteratively solve for the estimated blur kernel and sharp image. The solution process uses the semi-quadratic splitting method.

[0021] The image processing method for a monolithic diffractive infrared system proposed in this invention, which uses a local maximum gradient prior approach, has the following advantages and innovations compared to existing technologies:

[0022] 1. The method reveals that image degradation during the imaging process of a monolithic diffractive infrared system conforms to the characteristics of a local maximum gradient prior.

[0023] 2. By employing the inherent local maximum gradient prior, image restoration and enhancement can be performed directly without measuring and estimating the strict parameters of the optical system, making it more practical and capable of generalization.

[0024] 3. The proposed method incorporates a small gradient constraint term into its objective function, which can be used to eliminate small gradients and retain larger gradients, thus achieving the effect of removing imaging noise. Attached Figure Description

[0025] Figure 1 This is a flowchart of an image processing method for a monolithic diffraction infrared system according to the present invention.

[0026] Figure 2 A target for imaging with a monolithic diffraction infrared system.

[0027] Figure 3 This is a photographic optical path diagram of a monolithic diffraction infrared system.

[0028] Figure 4 The images are captured and the images restored by this method, wherein, Figure 4 (a) An image captured by the system. Figure 4 (b) is the processed image. Detailed Implementation

[0029] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0030] This invention proposes an image processing method for a monolithic diffraction infrared system. During the analysis of the system imaging process, it was found that the maximum gradient value within the neighborhood of each pixel in the image decreases after the system imaging process, leading to a degradation in the image quality. Therefore, this method is proposed as an inherent prior for system image restoration, and an image processing method for monolithic diffraction infrared systems is designed. This invention enables blind restoration without requiring specific prior knowledge of the optical system.

[0031] like Figure 1 As shown, the present invention provides an image processing method for a monolithic diffraction infrared system, comprising the following steps:

[0032] Step 1: Input a blurred image from a monolithic diffraction infrared system. The image is a two-dimensional grayscale matrix of size i*j.

[0033] Step 2: Establish the objective function for image restoration. The objective function is as follows:

[0034]

[0035] Among them, the first item In the diagram, I represents the estimated sharp image, and k represents the blur kernel. The first term represents the convolution process, where B represents the input blurred image. This term uses the L2 norm as a constraint to minimize the loss of the blurred image after convolving the sharp image with the blurred kernel. The second term, ||2-LMG(I)||1, represents the local maximum gradient value of the image. This term uses the L1 norm as a constraint to maximize the local maximum gradient of the image. The third term... middle, The fourth term represents the gradient matrix of the image. This term is constrained using the L0 norm to eliminate local small gradient regions while retaining large gradient regions; In this context, the L2 norm is used as a constraint to regularize the solution of the fuzzy kernel; α, β and λ are weight parameters, initialized to 0.01, 0.008 and 0.004 respectively.

[0036] Step 3: Initialize the blur kernel for the system's imaging degradation process. The blur kernel is a two-dimensional matrix of size m*m that causes system degradation. The blur kernel initialization process is as follows: Create an m*m matrix with all values ​​set to 0 according to the input blur kernel size m, and then set the values ​​of the elements at row position (m-1 / 2) and column position (m-1 / 2) to 0.1.

[0037] Step 4: Calculate the local maximum gradient prior of the image and use it as the L1 norm for sparsity constraints. The local maximum gradient refers to the maximum gradient within the neighborhood of each pixel in the image, calculated as follows:

[0038]

[0039] Where x represents the position of any pixel; P(x) represents the neighborhood of pixel x; and y represents the pixel in the neighborhood P(x) of pixel x. This represents the gradient value of pixel y; the gradient represents the sum of the rates of change of a pixel in the horizontal and vertical directions, and is calculated as follows:

[0040]

[0041] Among them, g x and g y These represent the rates of change in the horizontal and vertical directions, respectively, and the calculation range is 2 pixels.

[0042] Step 5: Decompose the objective function from Step 3 into subproblems. The objective function is decomposed into the following subproblems:

[0043] Formula 1:

[0044] Formula 2:

[0045] Equation 1 is used to solve for the clear image I; Equation 2 is used to solve for the blur kernel k.

[0046] Step Six: Iteratively solve for the estimated blur kernel and sharp image. The solution process uses the semi-quadratic splitting method.

[0047] Step 7: Output the final blur kernel, with a size of m*m; output the final clear image, with a size of i*j.

[0048] An image of the target object is obtained by using a monolithic diffraction infrared system. The target object is a checkerboard target, such as... Figure 2 As shown, a single diffractive lens is used for imaging. The target object is within the imaging range, and the infrared detector is on the focal plane. The captured image of the target is obtained through a monitor. Figure 3 As shown, the image is then processed using the method of the present invention to obtain the following result. Figure 4 The system shown directly captures images, and the images processed by the method are shown. It can be intuitively seen that the contrast and clarity of the images are greatly improved.

[0049] It should be understood that the above description of the preferred embodiments is quite detailed, but it should not be considered as a limitation on the scope of protection of this invention. Those skilled in the art, under the guidance of this invention, can make substitutions or modifications without departing from the scope of protection of the claims of this invention, and all such substitutions or modifications fall within the scope of protection of this invention. The scope of protection of this invention should be determined by the appended claims.

Claims

1. An image processing method for a monolithic diffraction infrared system, characterized in that: Includes the following steps: Step 1: Input the blurred image from the monolithic diffraction infrared system; Step 2: Establish the objective function for image restoration; the objective function model for image restoration is: Among them, the first item In the diagram, I represents the estimated sharp image, and k represents the blur kernel. The first term represents the convolution process, where B represents the input blurred image. This term uses the L2 norm as a constraint to minimize the loss between the sharp image and the blurred image after convolution with the blurred kernel; the second term... In this context, LMG represents the local maximum gradient value of the image. This term uses the L1 norm as a constraint to maximize the local maximum gradient of the image; the third term... middle, The fourth term represents the gradient matrix of the image. This term is constrained using the L0 norm to eliminate local small gradient regions while retaining large gradient regions; In this context, the L2 norm is used as a constraint to regularize the solution of the fuzzy kernel; , and These are the weight parameters; Step 3: Initialize the blur kernel for the system's imaging degradation process; Step 4: Calculate the local maximum gradient prior of the image and apply sparsity constraints using the L1 norm; the local maximum gradient refers to the maximum gradient value within the neighborhood of each pixel in the image, calculated as follows: Where x represents the position of any pixel; P(x) represents the neighborhood of pixel x; and y represents the pixel in the neighborhood P(x) of pixel x. This represents the gradient value of pixel y; Step 5: Decompose the objective function from Step 3 into subproblems; Step 6: Iteratively solve for the estimated blur kernel and sharp image.

2. The image processing method for a monolithic diffraction infrared system according to claim 1, characterized in that: The fuzzy kernel refers to the function that causes system degradation. It is represented by a two-dimensional matrix. The fuzzy kernel initialization process in step two is as follows: according to the fuzzy kernel size m set by the input, create a matrix m with all values ​​of 0. Take an m-matrix and set the values ​​of the elements at row position (m-1 / 2) and column position (m-1 / 2) to 0.

1.

3. The image processing method for a monolithic diffraction infrared system according to claim 1, characterized in that: The gradient represents the sum of the rates of change of a pixel in the horizontal and vertical directions, and is calculated as follows: Among them, g x and g y These represent the rates of change in the horizontal and vertical directions, respectively.

4. The image processing method for a monolithic diffraction infrared system according to claim 1, characterized in that: In step five, the objective function is decomposed into the following subproblems: Formula 1: Formula 2: Equation 1 is used to solve for the clear image I; Equation 2 is used to solve for the blur kernel k.

Images