Infrared image non-uniformity correction method based on notch filtering and guide filtering

Abstract

The invention discloses an infrared image non-uniformity correction method based on notch filtering and guide filtering, which comprises the following steps of: inputting an infrared focal plane array image, performing discrete Fourier transform on the input image to obtain an amplitude spectrum and a phase spectrum, predicting a central frequency spectrum area of the amplitude spectrum, estimating a notch filter compensation parameter, and correcting the infrared image non-uniformity. Further constructing a notch filter; enabling the amplitude spectrum to pass through the constructed notch filter to obtain a new amplitude spectrum after filtering, and obtaining a filtered background image layer through inverse discrete Fourier transform by utilizing the new amplitude spectrum after filtering and the original phase spectrum; carrying out subtraction on the original input infrared focal plane array image and the filtered background image layer to obtain a noise image, carrying out weighted improvement on a regularization coefficient by utilizing anisotropic guided filtering, and extracting a part of detail image layers from the noise image; and superposing all the detail layers, and outputting a correction result. The method can effectively overcome the problem that image details are lost in the non-uniform correction process.

Description

Technical Field

[0001] The invention relates to the technical field of infrared image processing, and in particular to an infrared image non-uniformity correction method based on notch filtering and guided filtering. Background Art

[0002] Infrared images have important applications in the fields of medical treatment, agricultural production, etc., mainly including thermal imagers, night vision devices, satellite imaging and other technologies. Unlike traditional optical imaging technology, in the COMS readout circuit of infrared thermal imaging technology, each column unit shares an amplifier. Since the parameter details of the amplifier cannot be completely consistent, significant stripe noise appears in the infrared image. In view of this, non-uniformity correction technology has emerged.

[0003] At present, there are two mainstream non-uniformity correction technologies. One is a calibration-based method, which requires data acquisition in a known environment and then reference correction. This method is simple in principle and relatively mature in technology, but it requires continuous correction to cope with the non-uniform time drift of the detector. Each correction requires interrupting the imaging process, which poses a major problem in application. The other method is a scene-based method, which does not require a reference environment but processes the image. It includes two methods, multi-frame and single-frame. The multi-frame correction method requires the analysis of multi-frame data, occupies a lot of resources, and is difficult to implement. The single-frame processing method only relies on a single picture, has fast timeliness, occupies less resources, overcomes the shortcomings of multi-frame, and is the current research focus.

[0004] At present, the algorithms with better effects in removing stripe noise include the intermediate histogram equalization algorithm based on grayscale statistics and the one-dimensional guided filtering algorithm based on filter estimation. The grayscale statistics algorithm adjusts the current pixel by calculating the local histogram to eliminate stripes; the one-dimensional guided filtering algorithm based on filter estimation achieves noise removal through local linear estimation.

[0005] Although the above algorithms can eliminate stripes, they have poor ability to preserve image details, and most of the image details will be lost during the processing. Therefore, there is a lack of a targeted infrared image uniformity correction algorithm that preserves image details well. Summary of the invention

[0006] In view of the shortcomings of the prior art, the present invention proposes an infrared image non-uniformity correction method based on notch filtering and guided filtering, which is used to accurately correct the non-uniform noise existing in the infrared image, while overcoming the problem of image detail loss during the non-uniform correction process and retaining a large amount of image details.

[0007] In order to solve the above technical problems, the technical solution of the present invention is:

[0008] The infrared image non-uniformity correction method based on notch filtering and guided filtering includes the following steps:

[0009] Step 1: Input an infrared focal plane array image, perform discrete Fourier transform on the input image, obtain an amplitude spectrum and a phase spectrum, estimate the compensation parameters of the notch filter by predicting the central spectrum area of ​​the amplitude spectrum, and then construct a notch filter;

[0010] Step 2: Pass the amplitude spectrum through the constructed notch filter to obtain a new amplitude spectrum after filtering, and use the new amplitude spectrum after filtering and the original phase spectrum to obtain the filtered background layer through inverse discrete Fourier transform;

[0011] Step 3: Subtract the original input infrared focal plane array image from the filtered background layer to obtain a noise map, and use anisotropic guided filtering to regularize the regularization coefficient. Weighted improvement, extracting some detail layers from the noise map;

[0012] Step 4: Overlay all detail layers and output the correction results.

[0013] Preferably, in step 1, the input infrared focal plane array image I(i) is defined as having a size of M×N. The infrared focal plane array image I(i) is subjected to discrete Fourier transform to obtain an amplitude spectrum G(u, v) bit spectrum. The expression is as follows:

[0014]

[0015] Among them, F[*] is the calculation of the amplitude spectrum, and Θ[*] is the calculation of the phase spectrum.

[0016] Preferably, in step 1, the method for estimating the compensation parameter of the notch filter is:

[0017] First, according to the size of the infrared focal plane array image I(i), the amplitude spectrum is obtained. The amplitude spectrum is an M×N matrix, and the expression is as follows:

[0018]

[0019] Then, the amplitude spectrum is equally divided into the upper and lower halves by row;

[0020] For the upper part, set the parameter K to represent the specific area of ​​the notch filter. Select row B through the data outside the K neighborhood. Estimate the K rows of data column by column and make a linear fit to calculate a. i , b i parameter:

[0021]

[0022] in, x i is the number of rows, G(i,v) is the data of the amplitude graph;

[0023] Using the estimated parameter a i , b i Fit and calculate the data in the K-neighborhood to estimate the compensation parameters of the notch filter;

[0024] The same calculations are performed for the lower half of the amplitude spectrum as for the upper half.

[0025] Preferably, in step 1, the notch filter constructed is as follows:

[0026]

[0027] Among them, Ω K×N Represents a K×N-shaped box, where K is the width of the rectangular box and is used to attenuate the noise energy in the horizontal center area. This is compensation after attenuation.

[0028] Preferably, in step 2, the amplitude spectrum G(u,v) is multiplied by the notch filter H(u,v), and finally Ω K×N Compensation within the area C i,j , forming a new amplitude spectrum G', and then the new amplitude spectrum G' and the original phase spectrum are inversely transformed by discrete Fourier transform to obtain the background layer.

[0029] Preferably, the anisotropic guided filtering method is: on the basis of guided filtering, by modifying the regularization coefficient, it can maintain the maximum and minimum intensity boundaries in all directions, so as to achieve better noise reduction effect at the edge details.

[0030] Preferably, in step 3, the method for extracting the detail layer by anisotropic guided filtering is expressed as follows:

[0031]

[0032]

[0033] q i =a i I i +b i

[0034] where μ k and σ k 2 Yes k The mean and variance of the pixel data of the guidance image within the window, λ is the regularization coefficient, and |ω| is ω k The number of pixels within Yes k The mean of the input image, By calculating the parameter a of the anisotropic guided filter i ,b i , then get the stripe layer, and subtract the stripe layer from the noise layer to get the detail layer, completing the extraction of detail information.

[0035] The present invention has the following characteristics and beneficial effects:

[0036] By adopting the above technical solution, the stripe information in the image can be effectively removed, the non-uniformity correction of the image can be completed, and a large amount of detail information can be retained through anisotropic guided filtering, thereby avoiding image distortion and effectively preventing image distortion. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0038] Figure 1 The present invention is a flowchart of an infrared image non-uniformity correction method based on notch filtering and guided filtering according to an embodiment of the present invention.

[0039] Figure 2 It is a schematic diagram of outputting the correction result according to an embodiment of the present invention.

[0040] Figure 3 It is a correction flow chart of a real-time example of the present invention. DETAILED DESCRIPTION

[0041] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0042] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and the like are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, features defined as "first", "second", and the like may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "multiple" means two or more.

[0043] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood by specific circumstances.

[0044] The present invention provides an infrared image non-uniformity correction method based on notch filtering and guided filtering, such as Figure 1-Figure 3 As shown, the specific steps are as follows:

[0045] Step 1: Input the infrared focal plane array image, perform discrete Fourier transform on the input image to obtain the amplitude spectrum and phase spectrum, use the amplitude spectrum data to predict the central spectrum area, classify the filtering problem as a regression problem, estimate the notch filter compensation parameters, and construct the notch filter.

[0046] Specifically: Input infrared focal plane array I(i), image size is M×N, discrete Fourier transform is performed to obtain amplitude spectrum G(u, v) and Right now

[0047]

[0048] Where I(i) is the original input infrared focal plane array, F[*] is the calculation of the amplitude spectrum, and Θ[*] is the calculation of the phase spectrum.

[0049] Using the amplitude spectrum data, we predict the central spectrum area, classify the filtering problem as a regression problem, estimate the compensation parameters of the notch filter, and construct the notch filter. The specific process is:

[0050] In the first part, according to a large number of image analyses, the energy of the stripe noise is mainly concentrated in the middle area of ​​the horizontal axis of the amplitude spectrum. By constructing a notch filter, the spectrum energy in the middle area is attenuated to achieve the purpose of removing noise and retain some image details. The amplitude spectrum is an M×N matrix. It is divided into the upper and lower parts, and the same operation needs to be performed separately. The parameter K is set to represent the specific area of ​​notch filtering (energy attenuation). Through the data outside the K neighborhood, select row B, estimate the data of row B, do linear fitting, and calculate a i , b i parameter.

[0051] In this embodiment, the least square method is used to predict the central spectrum area and estimate the compensation data of the middle K rows. Specifically, the upper half of the data is taken from Line fetch There are B rows of data in total. The parameter matrices a(i) and b(i) are calculated using the B rows of data to estimate Go to The lower half of the data is selected in the same way as above. arrive row, row B data, calculate parameters a(i), b(i), estimate Go to Row data.

[0052]

[0053] The calculation is performed column by column, where x i is the number of rows, and G(i,v) is the data of the amplitude graph.

[0054] Furthermore, in the second part, the compensation parameters of the notch filter are estimated, using the estimated parameters a i , b i Fitting calculation of the data in the K-neighborhood. The same operation should be divided into two parts, and the calculations should be performed separately.

[0055] C(i)=a(i)x(i)+b(i) (4)

[0056] Where C(i) is the estimated compensation value. The value of i is arrive

[0057] In the third part, the notch filter is constructed as follows:

[0058]

[0059] Among them, Ω K×NRepresents a rectangular box of size K×N, where K is the width of the rectangular box and is used to attenuate the noise energy in the horizontal center area. This is compensation after attenuation.

[0060] It can be understood that the value of the width of the rectangular frame is the same as the value of the parameter K.

[0061] Step 2: Pass the amplitude spectrum through the constructed notch filter to obtain a filtered new amplitude spectrum, and use the filtered amplitude spectrum and the original phase spectrum to obtain the filtered background layer through inverse discrete Fourier transform.

[0062] Specifically, it includes the following two parts:

[0063] In the first part, the amplitude spectrum G(u,v) is multiplied by the notch filter H(u,v) and finally in Ω K×N Compensation within the area C i,j , forming a new amplitude spectrum G'.

[0064] G'(u,v)=G(u,v)×H(u,v)+C (6)

[0065] In the second part, the background layer is obtained by inverse discrete Fourier transform of the filtered amplitude spectrum and the original phase spectrum.

[0066]

[0067] Among them, O1(i) is the background layer, F -1 [*] is the inverse discrete Fourier transform.

[0068] Step 3: Subtract the original input image from the filtered background image to obtain a noise map, and use anisotropic guided filtering to regularize the regularization coefficient. Weighted improvement,extracting some detail layers from the noise map.

[0069] The specific process is as follows:

[0070] In the first part, the input infrared focal plane array and the filtered background layer are separated to achieve image stratification and obtain the noise layer.

[0071] O n (i)=I(i)-O1(i) (8)

[0072] Among them, O n (i) is the obtained noise layer, I(i) is the input infrared focal plane array, and O1(i) is the background layer.

[0073] In the second part, we use anisotropic guided filtering to extract image details from the noise layer and compensate for image details. Anisotropic guided filtering is an improvement on guided filtering. The principle of guided filtering is that there is a linear relationship between the output and the guide map. On the basis of guided filtering, anisotropic guided filtering modifies the regularization coefficient to keep the maximum and minimum intensity boundaries in all directions, so as to achieve better noise reduction effect at the edge details. We use this feature to extract image details from stripe noise.

[0074]

[0075] Formula (9) is the formula for guided filtering, where a k and b k It is a window k The continuous linear constant coefficients, ω k is an N×N rectangular box, I i is the input guidance image, p i is the image to be processed, and formula (10) is k Constraint function within the window.

[0076]

[0077] In a k and b k The solution of μ k and σ k 2 Yes k The mean and variance of the pixel data of the guide image in the window, λ is the regularization coefficient, the purpose is to prevent σ k 2 =0 and a k Too big, |ω is ω k The number of pixels within Yes k The mean of the input images.

[0078]

[0079] q i =a i I i +b i (15)

[0080] Since the original guided filter has poor diffusion performance when processing image edge details and is insufficient in extracting image details, the linear constant coefficients are anisotropically weighted, and formula (13) is the weighting parameter, where Formula (14) is the weighted linear constant coefficient a i and b i, Formula (11) becomes Formula (15). Using anisotropic guided filtering, the guided map and the input map are both the noise layer O obtained in Formula (8) n (i) In the noise layer, the main part of the image is the stripe information, and the anisotropic guided filtering can be used to obtain O s (i), then O de (i) = O n (i)-O s (i) The detailed information of the image can be extracted.

[0081] Step 4: Overlay the layers and output the correction results.

[0082] The specific operations are:

[0083] O ou (i) = O1(i) + O de (i) (16)

[0084] Among them, ou (i) is the final correction result output, O1(i) is the background layer obtained by formula (7), O de (i) is the extracted detail information.

[0085] pass Figure 2 It can be seen that the correction result in this embodiment is relatively clear, and the resolution and contrast in the image are significantly improved. (a) is the original input image, (b) is the background layer, (c) is the detail layer, and (d) is the correction result.

[0086] In order to highlight the correction effect of the algorithm of the present invention in the actual image, based on the above description, the overall correction process Figure 3 shown.

[0087] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. For those skilled in the art, various changes, modifications, substitutions and variations of these embodiments including components are made without departing from the principles and spirit of the present invention, and still fall within the scope of protection of the present invention.

Claims

1. The infrared image non-uniformity correction method based on notch filtering and guided filtering is characterized by: The following steps are involved: Step 1: Input an infrared focal plane array image, perform discrete Fourier transform on the input image, obtain an amplitude spectrum and a phase spectrum, estimate the compensation parameters of the notch filter by predicting the central spectrum area of ​​the amplitude spectrum, and then construct a notch filter; Step 2: Pass the amplitude spectrum through the constructed notch filter to obtain a new amplitude spectrum after filtering, and use the new amplitude spectrum after filtering and the original phase spectrum to obtain the filtered background layer through inverse discrete Fourier transform; Step 3: Subtract the original input infrared focal plane array image from the filtered background layer to obtain a noise map, and use anisotropic guided filtering to regularize the regularization coefficient. Weighted improvement, extracting some detail layers from the noise map; Step 4: Overlay all detail layers and output the correction results.

2. The infrared image non-uniformity correction method based on notch filtering and guided filtering according to claim 1 is characterized in that: In the step 1, the input infrared focal plane array image I(i) is defined as M×N infrared focal plane array image I(i) is subjected to discrete Fourier transform to obtain the amplitude spectrum G(u,v) bit spectrum The expression is as follows: Among them, F[*] is the calculation of the amplitude spectrum, and Θ[*] is the calculation of the phase spectrum.

3. The infrared image non-uniformity correction method based on notch filtering and guided filtering according to claim 2, characterized in that: In step 1, the method for estimating the compensation parameters of the notch filter is: First, according to the size of the infrared focal plane array image I(i), the amplitude spectrum is obtained. The amplitude spectrum is an M×N matrix, and the expression is as follows: Then, the amplitude spectrum is equally divided into the upper and lower halves by row; For the upper part, set the parameter K to represent the specific area of ​​the notch filter. Select row B through the data outside the K neighborhood. Estimate the K rows of data column by column and make a linear fit to calculate a. i , b i parameter: in, x i is the number of rows, G(i,v) is the data of the amplitude graph; Using the estimated parameter a i , b i Fit and calculate the data in the K-neighborhood to estimate the compensation parameters of the notch filter; The same calculations are performed for the lower half of the amplitude spectrum as for the upper half.

4. The infrared image non-uniformity correction method based on notch filtering and guided filtering according to claim 3, characterized in that: In step 1, the notch filter constructed is as follows: Among them, Ω K×N Represents a K×N-shaped box, where K is the width of the rectangular box and is used to attenuate the noise energy in the horizontal center area. This is compensation after attenuation.

5. The infrared image non-uniformity correction method based on notch filtering and guided filtering according to claim 4 is characterized in that: In step 2, the amplitude spectrum G(u,v) is multiplied by the notch filter H(u,v), and finally in Ω K×N Compensation within the area C i,j , forming a new amplitude spectrum G', and then the new amplitude spectrum G' and the original phase spectrum are inversely transformed by discrete Fourier transform to obtain the background layer.

6. The infrared image non-uniformity correction method based on notch filtering and guided filtering according to claim 1 is characterized in that: The anisotropic guided filtering method is: on the basis of guided filtering, by modifying the regularization coefficient, it can maintain the maximum and minimum intensity limits in all directions, so as to achieve a better noise reduction effect at the edge details.

7. The infrared image non-uniformity correction method based on notch filtering and guided filtering according to claim 1 is characterized in that: In step 3, the method of extracting the detail layer by anisotropic guided filtering is expressed as follows: q i =a i I i +b i where μ k and σ k 2 Yes k The mean and variance of the pixel data of the guidance image within the window, λ is the regularization coefficient, and |ω| is ω k The number of pixels within Yes k The mean of the input image, By calculating the parameter a of the anisotropic guided filter i ,b i , then get the stripe layer, and subtract the stripe layer from the noise layer to get the detail layer, completing the extraction of detail information.

Images