Improved median histogram-based infrared image column stripe non-uniformity correction method

Abstract

The application discloses an infrared image column stripe non-uniformity correction method based on an improved median histogram, and belongs to the field of infrared image non-uniformity correction. The method comprises the following steps: calculating a column histogram of a single infrared image and a column mean value curve after wavelet denoising according to an input infrared image; performing an inverse operation on the obtained column histogram to obtain a column inverse histogram; generating an adaptive window of a bilateral filter according to the column mean value curve; generating the bilateral filter according to the spatial domain distance of the column and the difference in the gray domain of the column mean value curve; filtering the column inverse histogram according to the obtained bilateral filter to obtain a median histogram; and mapping the input infrared image according to the median histogram to obtain an infrared image after stripe removal. The application realizes a better stripe removal effect, avoids the generation of blind points, and omits the step of repeatedly calculating and comparing results to obtain an optimal parameter value in the original algorithm.

Description

Technical Field

[0001] This invention belongs to the field of infrared image non-uniformity correction, and in particular relates to a method for infrared image column stripe non-uniformity correction based on an improved median histogram. Background Technology

[0002] Ideally, the response characteristic curves of each detection unit in an infrared imaging system and the readout amplification characteristics of its readout circuit should be identical. That is, when irradiated by a uniform heat source, the output signal intensity of each detection unit should be completely consistent. However, due to factors such as the materials and manufacturing processes of the devices, and the influence of ambient temperature, noise, and stray radiation in the application environment, the response rates of each detection unit cannot be completely consistent. This ultimately results in non-uniform noise, or fixed-pattern noise, in the generated infrared image. Especially when different channels of the readout circuit are non-uniform, this leads to significant brightness differences between adjacent rows or columns in the output image, resulting in stripe non-uniform noise. The direction of stripe noise is related to the readout method of the pixels. If pixels in the same row share a single readout channel, horizontal stripes will be produced; if pixels in the same column share a single readout channel, vertical stripes will be produced. The presence of stripe non-uniform noise significantly reduces the quality of infrared images, not only reducing image visibility but also adversely affecting subsequent processing tasks such as target detection and identification. To improve the quality of infrared images and eliminate this stripe noise, non-uniformity correction is usually required.

[0003] Non-uniformity correction algorithms can be divided into two categories based on whether a calibration radiation source such as a blackbody is required: calibration methods and multi-point correction methods. Calibration methods require a uniformly radiating blackbody to collect image data of the blackbody at different temperatures. Since the blackbody emits uniform radiation, the global mean of the blackbody image is used as the theoretical true response value. Therefore, by fitting these calibration data, the response curve parameters of each detector element can be obtained. These parameters can be directly stored in the imaging system and accessed during imaging. Calibration methods offer good correction results and are simple to use, meeting real-time requirements. Therefore, many practical real-time correction imaging systems use calibration methods for non-uniformity correction. However, when the operating environment of an infrared imaging system changes, the pixel response characteristics on the infrared focal plane array drift over time. In this case, the non-uniformity correction coefficients obtained by calibration methods under laboratory conditions are no longer applicable and cannot effectively correct the image for non-uniformity. To obtain better non-uniformity correction results, the imaging process must be paused and the correction coefficients recalculated through calibration. However, pausing the imaging process when the detector is performing target detection, tracking, or guidance tasks can easily lead to target loss, severely impacting its usability. Another type is scene-based methods, mainly divided into four categories: statistical non-uniformity correction methods, filtering-based non-uniformity correction methods, optimization-based non-uniformity correction methods, and learning-based non-uniformity correction methods. Scene-based methods obtain the non-uniformity correction coefficients for each detector pixel based on the captured image during the imaging process. This approach eliminates the need for calibration radiation sources such as blackbodies and can adaptively track changes in pixel response characteristics, thus possessing broader application prospects.

[0004] Based on the number of image frames used in the correction process, non-uniformity correction methods can be categorized into single-frame-based and multi-frame-based methods. In application, multi-frame-based methods require spatiotemporal information from consecutive frames, thus necessitating certain assumptions about scene motion, such as randomness. When these assumptions are not met, ghosting and artifacts appear in the corrected image, significantly impacting image quality. Single-frame-based methods, on the other hand, utilize only a single frame to complete the correction, generally avoiding ghosting and artifacts caused by scene changes. The median histogram-based infrared image non-uniformity correction algorithm is based on the fact that a single column (or row, depending on the readout system) of the image carries sufficient information for histogram equalization. Because the image is continuous, the difference between two adjacent columns is statistically small, meaning their histograms are almost equal, and since stripe non-uniformity is usually independent. Therefore, stripe non-uniformity noise can be eliminated by transforming the histogram of each column (or row) to the median histogram of the neighboring column (or row) histograms. Median histogram-based infrared image non-uniformity correction algorithms have the advantages of simple principle, low complexity, and strong ability to preserve image details. However, in practice, traditional median histogram-based infrared image stripe non-uniformity correction algorithms use a fixed window determined by the standard deviation in Gaussian filtering. Considering that the histogram characteristics of neighboring columns are not similar at the details and edges of infrared images, the fixed window reduces the accuracy of the median histogram, leading to an increase in correction residuals and blind cells. Furthermore, traditional algorithms need to traverse parameters within a parameter range according to the minimum variation principle to determine the parameters, which wastes computational resources. Summary of the Invention

[0005] To address the issues that traditional infrared image non-uniformity correction algorithms based on median histograms are prone to causing correction residuals at scene details and edges, as well as generating blind pixels due to mapping, and easily leading to wasted computational resources, this invention proposes an infrared image column stripe non-uniformity correction method based on an improved median histogram. By introducing column mean curves denoised by one-dimensional wavelet, adaptive filtering windows, and bilateral filters that simultaneously consider gray-scale domain weights and distance domain weights, the algorithm for infrared image non-uniformity correction based on median histograms is optimized, effectively improving the correction effect, especially showing significant advantages at scene details and edges.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] An improved infrared image non-uniformity correction method based on median histogram includes the following steps:

[0008] Step (1): Based on the input infrared image Calculate the column histogram of a single frame infrared image And the column mean curve after wavelet denoising ;

[0009] Step (2): Based on the correspondence between the histogram and the inverse histogram, the generated column histogram... Obtain the column inverse histogram ;

[0010] Step (3): Set the maximum window size for adaptive windows. and minimum window And based on the column mean curve Adaptive window for generating bilateral filters ;

[0011] Step (4): Based on the obtained adaptive window The spatial distance between columns and the difference in grayscale range of the column mean curves generate a bilateral filter kernel matrix. ;

[0012] Step (5): Based on the obtained bilateral filter kernel matrix Inverse histogram Filtering is performed to obtain the column median histogram. ;

[0013] Step (6) According to the median histogram The infrared image after stripe removal is obtained by mapping. .

[0014] The beneficial effects of this invention are as follows:

[0015] When calculating the column mean curve of an infrared image, considering that stripe noise alters the column mean characteristics of a real infrared image, the column mean curve after one-dimensional wavelet denoising is more likely to reflect the correlation between columns in a real infrared image.

[0016] Considering the gradient of the column mean curve, columns with drastic changes in mean correspond to details and edges in the image, where the correlation between columns is poor, and a smaller window is used. Conversely, columns with gentler changes in mean correspond to uniform regions in the image, where the correlation between columns is strong, and a larger window is used. Compared to the fixed-size Gaussian filter used in the original algorithm, the improved algorithm generates a bilateral filter kernel matrix based on the obtained adaptive window size, the spatial domain distance of the column indices, and the difference in the gray domain of the column mean curves. This results in better correction and avoids the repetitive step of iterative calculation to determine parameters, as required by the original algorithm.

[0017] This invention effectively solves the problems of the original infrared image non-uniformity correction algorithm based on median histogram, which easily leads to correction residuals at scene details and edges, as well as the generation of blind pixels due to mapping, and easily leads to the waste of computing resources. Attached Figure Description

[0018] Figure 1 This is a flowchart of the infrared image stripe non-uniformity correction method based on improved median histogram according to the present invention.

[0019] Figure 2 The images shown are correction results based on the method of the present invention, where (a) is the original image, (b) is the correction result of the original algorithm, and (c) is the correction result of the improved algorithm.

[0020] Figure 3 The images shown are corrected stripe non-uniformity noise maps based on the method of this invention, where (a) is the correction result map of the original algorithm, (b) is the correction result map of the improved algorithm, (c) is an enlarged view of the correction result of the original algorithm, and (d) is an enlarged view of the correction result of the improved algorithm. Detailed Implementation

[0021] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0022] like Figure 1 As shown, the infrared image column stripe non-uniformity correction method based on improved median histogram provided by the present invention includes the following steps:

[0023] Step 1: Based on the input infrared image Calculate column histogram And the column mean curve after wavelet denoising of one-dimensional curves The method is as follows:

[0024] infrared images Let m be the row number of any row. [1, M], M 2;

[0025] infrared images The column number of any column is denoted as n, n [1, N], N 2;

[0026] infrared images The pixel value of any pixel (m, n) is denoted as ;

[0027] Histogram The calculation method is as follows:

[0028] ,

[0029] in, This represents the nth column of the column histogram H, i.e., the histogram generated from the nth column of the infrared image. l represents the gray level, which defaults to 0-255.

[0030] The mean of each column is calculated, and one-dimensional wavelet denoising is applied to remove fluctuations in the column mean caused by column stripes in the infrared image, thus better restoring the inherent fluctuations in the pixel value column mean in the real image. The formula is as follows:

[0031] ,

[0032] For one-dimensional signals Perform wavelet transform to obtain wavelet coefficients. Where a is the scale factor and b is the translation factor, the formula is as follows:

[0033] ,

[0034] in It is a wavelet basis function, and t represents the integration variable.

[0035] For wavelet coefficients A threshold is applied to remove noise. Common thresholding methods include hard thresholding and soft thresholding. Here, soft thresholding is used to achieve better noise reduction results. The threshold is set to λ, and the processed coefficients are... for:

[0036] ,

[0037] The processed coefficients are then subjected to an inverse transform to obtain the denoised column mean curve. The formula is as follows:

[0038] .

[0039] Step 2: Based on the correspondence between the histogram and the inverse histogram, generate the column histogram. Obtain the column inverse histogram The specific steps for calculating the inverse histogram are as follows:

[0040] ,

[0041] Where z represents the normalized frequency of the cumulative gray levels.

[0042] Step 3: Based on the column mean curve after wavelet denoising And set the maximum window size for adaptive windows. and minimum window Further generate an adaptive window for the bilateral filter. The specific steps for generating an adaptive window are as follows:

[0043] Let the maximum value of the adaptive window be denoted as max_window.

[0044] Let the minimum value of the adaptive window be denoted as min_window.

[0045] Calculate the column mean curve The gradient values ​​are calculated and sorted in order of size. The adaptive windows for the columns corresponding to the maximum 10% and the minimum 10% are set to min_window and max_window, respectively.

[0046] The remaining columns of the window are linearly stretched based on the maximum and minimum values ​​to obtain an adaptive window. .

[0047] Step 4: Based on the obtained adaptive window size The spatial distance between columns and the difference in grayscale range of the column mean curves generate a bilateral filter kernel matrix. The specific steps for generating the bilateral filter kernel matrix are as follows:

[0048] Spatial domain weights:

[0049] ,

[0050] in Let Variance be the variance of the spatial domain. Indicates the column number currently being processed.

[0051] Intensity domain weights:

[0052] ,

[0053] in Let be the variance of the intensity domain.

[0054] Based on the obtained adaptive window size For each column, the kernel of the bilateral filter is set to an adaptive window size, and a normalization operation is performed:

[0055] ,

[0056] in For example The neighborhood, For example The weight values ​​of the bilateral filter in the nth column of the neighborhood are in total. For each column, the weight values ​​of the bilateral filter are calculated as described above and then normalized, ultimately forming the bilateral filter kernel matrix. .

[0057] Step 5: Based on the obtained bilateral filter kernel matrix Inverse histogram Filtering is performed to obtain the column median histogram. The specific steps to generate a histogram of the median values ​​of a column are as follows:

[0058] ,

[0059] Step 6: Analyze the original noisy image based on the median histogram. The infrared image after stripe removal is obtained by mapping. The specific steps for generating the destriped infrared image are as follows:

[0060] in Any pixel value can be represented as:

[0061] ,

[0062] At this point, the destriped infrared image is obtained.

[0063] To verify the effectiveness of the method, this invention tests real infrared images containing non-uniformity of column stripes. Images are selected from the image library provided by the original algorithm, such as... Figure 2 The original image shown in (a) and the image corrected by the original algorithm are as follows: Figure 2 As shown in (b), the image after correction by the improved algorithm is as follows: Figure 2 As shown in (c), it can be seen that the improved algorithm significantly reduced column stripes, resulting in a substantial improvement in image quality. Through... Figure 3 As can be seen from the correction effect of another infrared image shown, the improved method of the present invention overcomes the phenomenon of blind elements generated by mapping in the original algorithm and achieves a better stripe removal effect. Among them, (a) is the correction result of the original algorithm, (b) is the correction result of the improved algorithm, (c) is an enlarged view of the correction result of the original algorithm, and (d) is an enlarged view of the correction result of the improved algorithm.

[0064] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for correcting non-uniformity of column stripes in infrared images based on an improved median histogram, characterized in that, Includes the following steps: Step (1): Based on the input infrared image Calculate the column histogram of a single frame infrared image And the column mean curve after wavelet denoising ; Step (2): Based on the correspondence between the histogram and the inverse histogram, the generated column histogram... Obtain the column inverse histogram ; Step (3): Set the maximum window size for adaptive windows. and minimum window And based on the column mean curve Adaptive window for generating bilateral filters ; Step (4): Based on the obtained adaptive window The spatial distance between columns and the difference in grayscale range of the column mean curves generate a bilateral filter kernel matrix. ; Step (5): Based on the obtained bilateral filter kernel matrix Inverse histogram Filtering is performed to obtain the column median histogram. ; Step (6): Input infrared image According to the median histogram The infrared image after stripe removal is obtained by mapping. .

2. The method for correcting infrared image stripe non-uniformity based on improved median histogram according to claim 1, characterized in that, Step (1) includes: infrared images Let m be the row number of any row. [1, M], M 2; infrared images The column number of any column is denoted as n, n [1, N], N 2; infrared images The pixel value of any pixel (m, n) is denoted as ; Histogram The calculation method is as follows: ， in, This represents the nth column of the column histogram H. Indicates grayscale level. [0, 255]; Calculate the mean of each column: ， For one-dimensional signals Perform wavelet transform to obtain wavelet coefficients. ,in It is a scale factor. It is the translation factor: ， in These are wavelet basis functions, where t is the integration variable; For wavelet coefficients Apply soft thresholding for noise reduction, setting the threshold to... The processed coefficients for: ， The processed coefficients are then subjected to an inverse transform to obtain the denoised column mean curve. : 。 3. The method for correcting infrared image column stripe non-uniformity based on improved median histogram according to claim 2, characterized in that, In step (2), the column inverse histogram is calculated. The method is as follows: ， Where z represents the normalized frequency of the cumulative gray levels.

4. The method for correcting infrared image stripe non-uniformity based on improved median histogram according to claim 3, characterized in that, Step (3) includes: Calculate the column mean curve for denoising. The gradient values ​​are calculated and sorted in order of magnitude. The adaptive windows corresponding to the columns with the largest and smallest 10% values ​​are set to [values]. and ; The remaining columns of the window are linearly stretched based on the maximum and minimum values ​​to obtain an adaptive window. .

5. The method for correcting infrared image column stripe non-uniformity based on improved median histogram according to claim 4, characterized in that, Step (4) includes: Calculate the spatial domain weights: ， in Let Variance be the variance of the spatial domain. Indicates the column number currently being processed; Calculate the intensity domain weights: ， in The variance of the intensity domain; Set the kernel of the bilateral filter for each column to an adaptive window size and perform a normalization operation: ， in For example The neighborhood, For example The weights of the bilateral filter used in the nth column of the neighborhood are in total. For each column, calculate the weight values ​​of the bilateral filter and normalize them to form the bilateral filter kernel matrix. .

6. The method for correcting the non-uniformity of infrared image column stripes based on an improved median histogram according to claim 5, characterized in that, In step (5), the median histogram of the columns. The first in Any pixel value in a column is represented as: 。 7. The method for correcting the non-uniformity of infrared image stripes based on an improved median histogram according to claim 6, characterized in that, In step (6), the infrared image after stripe removal Any pixel value is represented as: 。

Images