A non-uniformity correction method for infrared video based on robust estimation

Abstract

The present invention discloses an infrared video non-uniformity correction method based on robust estimation, which belongs to the field of infrared video processing. Aiming at the problems that the current scene-based infrared video non-uniformity correction method has a reduced correction effect and is prone to "ghosting" when the camera movement is insufficient, the present invention is based on the long-tail characteristics of the observed video sequence and uses a mixed distribution model to model the observed video sequence, which can effectively reduce the impact of sharp edges in the image on the non-uniformity estimation, thereby effectively avoiding the generation of "ghosting"; the maximum likelihood estimation is used to estimate the non-uniformity to ensure the quality of the estimation, thereby effectively correcting the non-uniformity in the video. The present invention is suitable for fields such as infrared thermal imaging, and can effectively correct the non-uniformity in infrared videos without interrupting the camera operation, improve the quality of infrared images, and has high application value.

Description

Technical Field

[0001] The invention relates to an infrared video non-uniformity correction method based on robust estimation, and belongs to the field of infrared video processing. Background Art

[0002] Infrared imaging is a new type of imaging technology. Different from traditional RGB imaging or full-color imaging, infrared imaging can directly detect the thermal radiation power of an object, thus having incomparable advantages over traditional imaging in some special fields, such as night vision, temperature measurement, military reconnaissance, etc. In terms of its technical path, infrared imaging detectors can be divided into cooled and uncooled types. The operating temperature of cooled detectors is generally around tens of Kelvin, and special liquid nitrogen refrigeration devices are required for cooling, and they are mostly used in high-end equipment; while uncooled detectors can work normally at room temperature and are widely used in the civilian field.

[0003] Uncooled infrared detectors have a serious non-uniformity problem. The generation of non-uniformity can be attributed to the inconsistency of physical properties such as pixel size, effective absorption area, doping concentration, and resistance caused by defects in the detector processing technology; when receiving radiation of the same intensity, the signal value generated by each pixel is different. At the same time, the response of the pixel is closely related to the ambient temperature. When the ambient temperature changes, the response of each pixel will also change accordingly, causing the non-uniformity to change. In short, the existence of non-uniformity makes the final imaging superimposed with a layer of "noise". For each given detector, this special "noise" is only affected by the ambient temperature: when the ambient temperature is stable, the non-uniformity remains unchanged; when the ambient temperature changes, the non-uniformity also changes.

[0004] The current non-uniformity correction methods can be roughly divided into two categories: one is calibration-based correction methods, and the other is scene-based correction methods. The calibration-based correction method uses the shutter method for correction. During the operation of the camera, the camera shutter is automatically lowered at regular intervals or when the ambient temperature changes, and the response of each pixel is corrected to make its output uniform (see Schulz M, Caldwell L. Nonuniformity correction and correctability of infrared focal plane arrays [J]. Infrared Physics & Technology, 1995, 36 (4): 763-777). The disadvantages of this type of correction method are obvious. First, the shutter will interrupt the normal operation of the camera; second, the shutter is a mechanical structure, which increases the cost and reduces the reliability. The scene-based correction method does not require a shutter structure. In theory, it can complete the non-uniformity correction without interrupting the normal operation of the camera. It is currently receiving more and more attention.

[0005] The scene-based nonuniformity correction method needs to rely on the movement of the camera. The basic idea is that when the camera moves, the scene signal received by each pixel changes rapidly, while the nonuniformity of each pixel changes slowly or even remains unchanged (when the ambient temperature is stable); a low-pass filter can be used to separate the signal from the nonuniformity. The most commonly used filter is the linear recursive filter (see Scribner DA, Sarkady KA, Caulfield JT, et al. Nonuniformity correction for staring IR focal plane arrays using scene-based techniques [C]. Infrared Detectors and Focal Plane Arrays. International Society for Optics and Photonics, 1990, 1308: 224-233). This type of scene-based nonuniformity correction method using a linear filter has obvious limitations: when the camera movement is not sufficient, the correction effect decreases, and even "ghost images" may appear. Summary of the invention

[0006] In view of the problems that the current scene-based non-uniformity correction method has, such as reduced correction effect and generation of "ghosts" when the camera motion is insufficient, the present invention aims to provide an infrared video non-uniformity correction method based on robust estimation. By adopting the robust estimation technology, the non-uniformity can still be effectively corrected even if the camera motion is insufficient, and the generation of "ghosts" can be effectively avoided.

[0007] In order to achieve the above purpose, the present invention adopts the following technical solutions:

[0008] The present invention discloses an infrared video non-uniformity correction method based on robust estimation. Based on the long-tail characteristics of the observed video sequence, a mixed distribution model is used to model the observed video sequence, and finally the non-uniformity correction algorithm proposed by the present invention is derived by using maximum likelihood estimation. Even if the camera movement is not sufficient, the non-uniformity can still be effectively corrected, and the generation of "ghosting" can be effectively avoided.

[0009] The present invention discloses an infrared video non-uniformity correction method based on robust estimation, comprising the following steps:

[0010] Step 1: Initialize the non-uniformity of the infrared video to B0=0.

[0011] B0 records the value of each pixel offset initialized. The size of B0 is consistent with the size of the infrared image. Assuming that the size of the infrared image is m rows and n columns, then B0 is an m×n matrix. Initialize each element of B0 to zero.

[0012] Step 2: Calculate the k-th frame of infrared video Y k The local mean of Where k = 1, 2, 3, etc.

[0013] Y k and Both are m×n matrices. The i-th row and j-th column element in corresponds to Y k The average value of the element in the i-th row and j-th column and its surrounding elements.

[0014] Step 3: Convert the kth frame of infrared video Y k Subtract its local mean Get the residual image

[0015] is a matrix of size m×n. The specific calculation method is

[0016]

[0017] Step 4: Update the non-uniformity of the infrared video to obtain B k .

[0018] Using the mixed distribution model Modeling and calculation of B k The maximum likelihood estimate of , we get the following cost function

[0019]

[0020] Where μ is a positive real number, which can be taken as The standard deviation of the cost function (2) is k-1 Derivative, we get B k The update equation

[0021]

[0022] Step 5: Subtract the non-uniformity to obtain the corrected k-th frame of infrared video X k , that is, infrared video non-uniformity correction is achieved based on robust estimation.

[0023] The X k Calculated by the following formula

[0024] X k =Y k -B k(4)

[0025] Beneficial effects:

[0026] 1. The present invention discloses an infrared video non-uniformity correction method based on robust estimation. Based on the long-tail characteristics of video sequences, a mixed distribution model is used to model the video sequences, thereby effectively avoiding the generation of "ghost images".

[0027] 2. The present invention discloses an infrared video non-uniformity correction method based on robust estimation, which uses maximum likelihood estimation to estimate non-uniformity, ensures the quality of the estimation, and effectively corrects the non-uniformity in the video.

[0028] 3. The present invention discloses an infrared video non-uniformity correction method based on robust estimation, which realizes robust estimation on the basis of achieving beneficial effects 1 and 2. Even if the camera movement is not sufficient, it can still effectively correct the non-uniformity and effectively avoid the generation of "ghost images". BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 The invention discloses a flowchart of an infrared video non-uniformity correction method based on robust estimation.

[0030] Figure 2 The correction effect of the method of the present invention is compared with that of the existing method.

[0031] Figures a and c are the correction effects of the existing method, and Figures b and d are the correction effects of the method of the present invention. DETAILED DESCRIPTION

[0032] The present invention will be described in detail below with reference to the accompanying drawings and embodiments. The technical problems solved by the technical solution of the present invention and the beneficial effects are also described. It should be noted that the described embodiments are only intended to facilitate the understanding of the present invention and do not have any limiting effect on it.

[0033] This embodiment discloses an infrared video non-uniformity correction method based on robust estimation, which is applied to infrared thermal imaging cameras. The hardware test conditions are: Inter i7 6700, 8G RAM, Matlab 2016a; the infrared camera used is Xenics Gobi-640-GigE; the collected image is 14-bit NUC data; the camera automatic correction function is turned off, and μ is set to 1.

[0034] This embodiment discloses a method for correcting infrared video non-uniformity based on robust estimation, such as Figure 1 As shown, the following steps are included:

[0035] Step 101: Initialize the non-uniformity of the infrared video B0=0.

[0036] The B0 records the value of each pixel offset for initialization. The size of B0 is consistent with the size of the infrared image. The size of the infrared image is 640*512, so B0 is a 640*512 matrix. Each element of B0 is initialized to zero.

[0037] Step 102: Sequentially read each frame of infrared image Y k , and calculate its local mean

[0038] As described in step 102 It is calculated by the following formula.

[0039]

[0040] So Calculated by the following formula

[0041]

[0042] In the formula Represents a two-dimensional convolution computation.

[0043] Step 103: Transform the k-th frame of infrared image Y k Subtract its local mean Get the residual image

[0044] is a matrix of size 640*512. The specific calculation method is

[0045]

[0046] Step 4: Update the non-uniformity of the infrared video to obtain B k .

[0047] Step 104 B k Calculated by the following formula

[0048]

[0049] Wherein μ is a positive real number, and in this embodiment, it is 1; and The output of is a 640*512 matrix, and the specific definitions of its i-th row and j-th column are as follows:

[0050]

[0051]

[0052] Step 105: Output the corrected k-th infrared image X k .

[0053] Step 105 X k Calculated by the following formula

[0054] X k =Y k -B k (4)

[0055] from Figure 2 It can be seen that after adopting the non-uniformity correction algorithm of the present invention, the correction effect is greatly improved, and at the same time, the "ghosting" problem that is prone to occur in the existing algorithm when the camera movement is insufficient is effectively avoided.

[0056] The specific description above further illustrates the purpose, technical solutions and beneficial effects of the invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for infrared video non-uniformity correction based on robust estimation, characterized in that: The following steps are included: Step 1: Initialize the non-uniformity of the infrared video B0=0; Step 2: Calculate the k-th frame of infrared video Y k The local mean of Where k = 1, 2, 3, ...; Step 3: Convert the kth frame of infrared video Y k Subtract its local mean Get the residual image Step 4: Update the non-uniformity of the infrared video to obtain B k ; The implementation method of step 4 is: Using the mixed distribution model Modeling and calculation of B k The maximum likelihood estimate of , we get the following cost function Where μ is a positive real number, which can be taken as The standard deviation of the cost function (2) about B k-1 Derivative, we get B k The update equation Step 5: Subtract the non-uniformity to obtain the corrected k-th frame of infrared video X k , that is, infrared video non-uniformity correction is achieved based on robust estimation.

2. The infrared video non-uniformity correction method based on robust estimation as claimed in claim 1, characterized in that: The implementation method of step one is: B0 records the value of each initialized pixel bias; the size of B0 is consistent with the size of the infrared image; assuming that the size of the infrared image is m rows and n columns, then B0 is an m×n matrix; each element of B0 is initialized to zero.

3. The infrared video non-uniformity correction method based on robust estimation as claimed in claim 1, characterized in that: The implementation method of step 2 is: Y k and All are m×n matrices; The i-th row and j-th column element in corresponds to Y k The average value of the element in the i-th row and j-th column and its surrounding elements.

4. The infrared video non-uniformity correction method based on robust estimation as claimed in claim 1, characterized in that: The implementation method of step three is: is a matrix of size m×n; The specific calculation method is:

5. The infrared video non-uniformity correction method based on robust estimation as claimed in claim 1, characterized in that: The implementation method of step five is: The X k Calculated by the following formula X k =Y k -B k (4)。

Images