Single image non-uniformity correction method and system based on space-frequency combination, and medium
By decomposing and fusing subbands of infrared images, and combining thresholding functions and masking, the problem of vertical stripe noise in infrared images was solved, improving image quality and clarity.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-24
- Publication Date
- 2026-04-24
AI Technical Summary
In infrared imaging technology, the non-uniformity caused by the difference in the responsivity of infrared detectors results in vertical stripe noise superimposed on the image, affecting the image quality and clarity.
By decomposing the infrared image into vertical sub-bands, adjacent high-frequency sub-bands, and low-frequency sub-bands, feature extraction and gradient amplitude fusion are performed to construct a threshold function. Combined with edge and noise masks, image correction is then performed.
It effectively removes vertical stripe non-uniformity, improves infrared image quality and boundary clarity, and ensures image information integrity.
Smart Images
Figure CN121921216A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of image processing technology, and in particular to a method, system, and medium for correcting non-uniformity of a single image based on spatial-frequency coherence. Background Technology
[0002] Infrared imaging technology, developed based on the unique infrared radiation characteristics of natural objects, plays an indispensable role in various fields such as military reconnaissance, aerospace monitoring, civilian facility inspection, and medical diagnosis. However, due to limitations in manufacturing processes, different detector elements in an infrared detector exhibit different responsibilities to the same infrared radiation, resulting in non-uniformity during the imaging process. Specifically, this manifests as a fixed pattern of noise superimposed on the image, severely impacting the overall image quality. This fixed pattern noise takes various forms, with vertical stripe-type non-uniformity being particularly significant. It appears as regular lines in the image, severely affecting image clarity and realism.
[0003] Currently, scholars have proposed calibration-based and scene-based non-uniformity correction methods to address this type of non-uniformity. Typically, a pre-established calibration model is used to adjust and compensate for the response rate differences of each detector element. This method is relatively simple, but in practical applications, it suffers from problems such as inaccurate or outdated calibration models. Summary of the Invention
[0004] Therefore, it is necessary to propose a method, system, and medium for single-image non-uniformity correction based on spatial-frequency joint method to address the above problems.
[0005] A method for non-uniformity correction of a single image based on spatial-frequency joint analysis, the method comprising: Acquire infrared images.
[0006] The infrared image is decomposed into vertical sub-bands, adjacent high-frequency sub-bands, and low-frequency sub-bands.
[0007] The infrared image is subjected to feature extraction processing at at least two specific scales to determine the gradient magnitude at each scale. The gradient magnitude maps at at least two different scales are then fused to obtain a fused gradient map.
[0008] The edge regions of the fused gradient map are compressed to obtain an edge response map. Based on the comparison between the pixel value of each pixel in the edge response map and the preset high threshold and preset low threshold, the effective edge pixels are determined and converted into an edge binary mask.
[0009] A threshold function is constructed based on the local energy of the vertical sub-band and the adjacent high-frequency sub-band, and a noise-dominant region mask is determined based on the local energy of the vertical sub-band.
[0010] The edge binary mask is combined with the noise-dominant region mask, and the vertical sub-band is processed using the threshold function. The processed vertical sub-band, adjacent high-frequency sub-band, and low-frequency sub-band are then inversely transformed to obtain the corrected infrared image.
[0011] Specifically, the process of processing the infrared image at at least two specific scales to determine the gradient magnitude at each scale includes: The infrared image is subjected to feature extraction processing at at least two specific scales to obtain initial images at each scale.
[0012] The initial image at each scale is convolved using preset horizontal and vertical convolution kernels to obtain the horizontal and vertical gradients of the infrared image at each scale.
[0013] The gradient magnitude at each scale is determined based on the horizontal and vertical gradients of the infrared image at each scale.
[0014] The step of fusing at least two gradient magnitude maps of different scales to obtain a fused gradient map specifically includes: according to The gradient magnitude maps at least two different scales are fused to obtain a fused gradient map, wherein... This is a gradient magnitude plot on an s-scale. These are parameters used to control the weight decay rate. To fuse gradient maps.
[0015] Specifically, determining effective edge pixels based on a comparison between the pixel value of each pixel in the edge response map and a preset high threshold and a preset low threshold, and converting the effective edge pixels into an edge binary mask, includes: The pixels whose pixel values in the edge response map are greater than a preset high threshold are defined as strong edge pixels, and these strong edge pixels are the first valid edge pixels.
[0016] The pixels whose pixel values in the edge response map are less than or equal to the preset high threshold and greater than the preset low threshold are selected as candidate edge pixels.
[0017] The second effective edge pixel is determined based on the connectivity between the strong edge pixel and the candidate edge pixel.
[0018] The first and second effective edge pixels are converted into a binary edge mask.
[0019] Specifically, determining the effective edge pixels based on the connectivity between the strong edge pixels and the candidate edge pixels includes: If the strong edge pixel is connected to the candidate edge pixel, then the candidate edge pixel is the second valid edge pixel.
[0020] If the strong edge pixel and the candidate edge pixel are not connected, then the candidate edge pixel is a pseudo edge pixel, and the pseudo edge pixel is discarded.
[0021] The step of constructing a threshold function based on the local energy of the vertical sub-band, the local energy of the adjacent high-frequency sub-band, and the local energy of the low-frequency sub-band specifically includes: according to Determine the local energy of the vertical sub-band, where, Let K be the local energy of the vertical subband, K be the half-size of the window in the horizontal direction, L be the half-size of the window in the vertical direction, k be the downsampling scale level, and l be the weight index during gradient fusion. For vertical sub-bands.
[0022] according to Construct a threshold function, where T(x,y) is the threshold function, and α and β are empirical weights. For the local energy of the vertical subband, for Local energy of adjacent high-frequency subbands in the direction, for Local energy of adjacent high-frequency subbands in the direction.
[0023] Specifically, determining the noise-dominant region mask based on the local energy of the vertical sub-band includes: according to Determine the mask for the noise-dominant region, where, For noise-dominant regions, For the local energy of the vertical subband, This is the energy threshold.
[0024] Specifically, the step of combining the edge binary mask image with the noise-dominant region mask and processing the vertical sub-band using the threshold function includes: according to The vertical sub-strips are processed; in, The vertical sub-band coefficient after processing. For edge binary mask, Let T(x,y) be the mask for the noise-dominant region, and T(x,y) be the threshold function. This represents the sub-band coefficient in the vertical direction.
[0025] A single-image non-uniformity correction system based on spatial-frequency joint method, the system comprising: The image acquisition module is used to acquire infrared images.
[0026] The image decomposition module is used to decompose the infrared image into vertical sub-bands, adjacent high-frequency sub-bands, and low-frequency sub-bands.
[0027] An edge binary mask acquisition module is used to perform feature extraction processing on the infrared image at least two specific scales, determine the gradient magnitude at each scale, fuse the gradient magnitude maps at at least two different scales to obtain a fused gradient map, compress the edge region of the fused gradient map to obtain an edge response map, determine the effective edge pixels based on the comparison between the pixel value of each pixel in the edge response map and the preset high threshold and preset low threshold, and convert the effective edge pixels into an edge binary mask.
[0028] The noise-dominant region mask module is used to construct a threshold function based on the local energy of the vertical sub-band, the local energy of the adjacent high-frequency sub-band, and the local energy of the low-frequency sub-band, and to determine the noise-dominant region mask based on the local energy of the vertical sub-band.
[0029] The infrared image correction module is used to combine the edge binary mask image with the noise-dominant region mask, and use the threshold function to process the vertical sub-band. The processed vertical sub-band, adjacent high-frequency sub-band and low-frequency sub-band are inversely transformed to obtain the corrected infrared image.
[0030] A computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to perform the steps of the method described above.
[0031] The embodiments of the present invention have the following beneficial effects: This invention decomposes infrared images into vertical sub-bands, adjacent high-frequency sub-bands, and low-frequency sub-bands, thus better preserving the directional information of the signal. Since non-uniformity is mainly manifested as stripe noise in the vertical direction, this invention effectively extracts information in this direction. Furthermore, by performing feature extraction processing on the infrared image at at least two specific scales, the scale sensitivity of the image is increased, the gradient magnitude at each scale is determined, and the gradient magnitude maps at at least two different scales are fused to obtain a fused gradient map. This fusion process captures edge features at different scales. Further, the edge regions of the fused gradient map are compressed to obtain an edge response map. Based on the comparison of the pixel value of each pixel in the edge response map with preset high and low thresholds, effective edge pixels are determined. These effective edge pixels are converted into a binary edge mask, improving the clarity of the image boundary regions and making the effects of non-uniformity more apparent, facilitating subsequent processing.
[0032] Furthermore, a threshold function is constructed based on the local energy of the vertical sub-band and adjacent high-frequency sub-bands. This threshold function can effectively highlight and process noisy areas without affecting the true information of the image, thereby improving the overall image quality.
[0033] Furthermore, a noise-dominant region mask is determined based on the local energy of the vertical sub-bands, and the noise region is identified through this mask. The edge binary mask is combined with the noise-dominant region mask, and a threshold function is used to process the vertical sub-bands. The processed vertical sub-bands, adjacent high-frequency sub-bands, and low-frequency sub-bands are then inversely transformed to obtain the corrected infrared image, effectively removing vertical stripe non-uniformity. This invention can significantly improve the overall quality of infrared images while ensuring the integrity of image information, thus better meeting practical application needs. Attached Figure Description
[0034] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0035] in: Figure 1 This is a flowchart illustrating an embodiment of a single-image non-uniformity correction method based on space-frequency joint method provided by the present invention. Figure 2 This is a flowchart illustrating an embodiment of a single-image non-uniformity correction method based on space-frequency joint method provided by the present invention. Figure 3A schematic diagram of an embodiment of a single-image non-uniformity correction system based on space-frequency joint method provided by the present invention; Figure 4 A schematic diagram of the structure of an embodiment of the medium provided by the present invention. Detailed Implementation
[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0037] like Figure 1 As shown, Figure 1 This is a flowchart illustrating an embodiment of a single-image non-uniformity correction method based on spatial-frequency joint analysis provided by the present invention. The method includes: S101: Acquire infrared image.
[0038] S102: Decompose the infrared image into vertical sub-bands, adjacent high-frequency sub-bands, and low-frequency sub-bands.
[0039] For example, the infrared image I(x,y) is decomposed into L-level asymmetric dual-tree complex wavelet transform to extract sub-band information in each direction, as shown in the following equation: ; in, This represents the high-frequency subband in the l-th layer and d-th direction. This represents the low-frequency subband of the Lth layer.
[0040] The focus is on acquiring the vertical sub-band and its adjacent high-frequency sub-bands in the adjacent directions, while retaining the low-frequency sub-band as the basic information of the image.
[0041] S103: Perform feature extraction processing on the infrared image at at least two specific scales, determine the gradient magnitude at each scale, fuse the gradient magnitude maps at at least two different scales, and obtain a fused gradient map.
[0042] For example, feature extraction processing is performed on the infrared image at at least two specific scales, the specific scale selection... (respectively, original, downsampled ×2, downsampled ×4), after passing through an anti-aliasing Gaussian filter ( After downsampling, initial images at each scale are obtained.
[0043] Furthermore, the initial image Is is convolved with the horizontal convolution kernel Sx and the vertical convolution kernel Sy respectively to obtain the horizontal gradient Gx=Is. Sx and the vertical gradient Gy=Is Sy. Among them, The convolution operation is represented by the following formula: ; The vertical convolution kernel Sy is shown in the following formula: .
[0044] Furthermore, the gradient magnitude of each pixel is determined according to the following formula: ; in, This is a gradient magnitude map at scale s. The gradient magnitude reflects the edge intensity distribution of the image at this scale. The larger the value, the more prominent the edge.
[0045] Furthermore, the Sigmoid weighting function is used to fuse gradient magnitudes at multiple scales to obtain a fused gradient map, which can be represented as follows: ; in, This parameter, used to control the weight decay rate, is set to 1.5. To fuse gradient maps.
[0046] The fused gradient map integrates multi-scale information, preserving detailed edges (from the fine scale) while suppressing noise and false edges (verified by the coarse scale), providing more robust gradient features for subsequent edge extraction.
[0047] S104: Compress the edge region of the fused gradient map to obtain the edge response map. Based on the comparison between the pixel value of each pixel in the edge response map and the preset high threshold and preset low threshold, determine the effective edge pixels and convert the effective edge pixels into an edge binary mask.
[0048] For example, by combining the "thermal radiation gradient direction information" of the fused gradient map with that of the infrared image, the gradient map is "direction-sensitively enhanced"—highlighting edges that are consistent with the direction of thermal radiation (such as the real edges of the interface between hot and cold objects) and suppressing irrelevant false gradients.
[0049] Furthermore, the enhanced gradient map is compressed using non-maximum suppression, and the size of the suppression window is dynamically adjusted according to the local curvature. Along the edge normal direction, only local maxima are retained, while other points are suppressed, thereby achieving edge refinement and obtaining a single-pixel-wide edge response map.
[0050] Furthermore, a dual-threshold processing is performed on the edge response map: a preset high threshold is set as the 90th percentile of the edge response map, identifying strong edges; a preset low threshold is set as the 50th percentile of the edge response map, identifying candidate edges. Pixels in the edge response map whose pixel values are greater than the preset high threshold are designated as strong edge pixels, and these strong edge pixels are the first effective edge pixels. Pixels in the edge response map whose pixel values are less than or equal to the preset high threshold and greater than the preset low threshold are designated as candidate edge pixels. If a strong edge pixel and a candidate edge pixel are connected, the candidate edge pixel is designated as the second effective edge pixel; if they are not connected, the candidate edge pixel is designated as a false edge pixel and is discarded. Further, the continuity and geometric consistency of the first and second effective edge pixels are checked, isolated points and short edge fragments (length < 5 pixels) are removed, and the first and second effective edge pixels after removing isolated points and short edge fragments are converted into a binary edge mask.
[0051] S105: Construct a threshold function based on the local energy of the vertical sub-band and adjacent high-frequency sub-bands, and determine the noise-dominant region mask based on the local energy of the vertical sub-band.
[0052] For example, the local energy of the vertical sub-band is calculated using the noise energy map, i.e.: ; The window size (2K+1)×(2L+1) is set according to the width of the noise stripes. Let K be the local energy of the vertical subband, K be the half-size of the window in the horizontal direction, L be the half-size of the window in the vertical direction, k be the downsampling scale level, and l be the weight index during gradient fusion. For vertical sub-bands.
[0053] For the energy map in the vertical direction, a threshold function needs to be set to remove stripe non-uniformity. The threshold function is designed to take into account both the energy in the vertical direction and the contrast between adjacent directions, i.e.: ; Where T(x,y) is the threshold function, and α and β are empirical weights that control the contributions of absolute energy and relative contrast, respectively. For the local energy of the vertical subband, for Local energy of adjacent high-frequency subbands in the direction, for Local energy of adjacent high-frequency subbands in the direction.
[0054] The threshold function has adaptive properties: in the vertical stripe region, E90° is significantly higher than that in the adjacent direction, and T(x,y) is larger; in the normal texture region, the energy in each direction is similar, and T(x,y) is smaller.
[0055] Furthermore, the local energy distribution obtained based on the above calculations... The noise-dominant region is extracted through threshold segmentation. Specifically, the energy values of all pixels in the entire image are statistically analyzed, and the 95th percentile of the global energy distribution is used as the threshold. This threshold effectively distinguishes noise-dominant regions from normal texture regions. The resulting noise-dominant region mask is as follows: ; in, For noise-dominant regions, For the local energy of the vertical subband, This is the energy threshold.
[0056] The mask identifies the main distribution areas of vertical stripe non-uniformity in the image, providing spatial location information for subsequent soft thresholding processing.
[0057] S106: Combine the edge binary mask with the noise-dominant region mask, and use a threshold function to process the vertical sub-band. Perform inverse transformation on the processed vertical sub-band, adjacent high-frequency sub-band, and low-frequency sub-band to obtain the corrected infrared image.
[0058] For example, when performing soft thresholding, an edge binary mask is used. With threshold function T(x,y) and noise-dominant region mask Together, they achieve refined processing of the vertical sub-band coefficients, as shown in the following formula: ; in, The vertical sub-band coefficient after processing. For edge binary mask, Let T(x,y) be the mask for the noise-dominant region, and T(x,y) be the threshold function. This represents the sub-band coefficient in the vertical direction.
[0059] Furthermore, the corrected infrared image can be obtained by performing an inverse asymmetric dual-tree complex wavelet transform on the processed vertical sub-band, adjacent high-frequency sub-band, and low-frequency sub-band.
[0060] As described above, this invention better preserves the directional information of the signal by decomposing the infrared image into vertical sub-bands, adjacent high-frequency sub-bands, and low-frequency sub-bands. Since non-uniformity is mainly manifested as stripe noise in the vertical direction, this invention effectively extracts information in this direction. Furthermore, by performing feature extraction processing on the infrared image at at least two specific scales to increase the image's scale sensitivity, the gradient magnitude at each scale is determined. The gradient magnitude maps at at least two different scales are fused to obtain a fused gradient map, which is then fused to capture edge features at different scales. Further, the edge regions of the fused gradient map are compressed to obtain an edge response map. Based on the comparison of the pixel value of each pixel in the edge response map with preset high and low thresholds, effective edge pixels are determined. These effective edge pixels are converted into a binary edge mask, improving the clarity of the image boundary regions and making the effects of non-uniformity more apparent, facilitating subsequent processing.
[0061] Furthermore, a threshold function is constructed based on the local energy of the vertical sub-band and adjacent high-frequency sub-bands. This threshold function can effectively highlight and process noisy areas without affecting the true information of the image, thereby improving the overall image quality.
[0062] Furthermore, a noise-dominant region mask is determined based on the local energy of the vertical sub-bands, and the noise region is identified through this mask. The edge binary mask is combined with the noise-dominant region mask, and a threshold function is used to process the vertical sub-bands. The processed vertical sub-bands, adjacent high-frequency sub-bands, and low-frequency sub-bands are then inversely transformed to obtain the corrected infrared image, effectively removing vertical stripe non-uniformity. This invention can significantly improve the overall quality of infrared images while ensuring the integrity of image information, thus better meeting practical application needs.
[0063] like Figure 2 As shown, Figure 2 This is a flowchart illustrating an embodiment of a single-image non-uniformity correction method based on spatial-frequency joint analysis provided by the present invention. The method includes: S201: Acquire infrared image.
[0064] S202: Decompose the infrared image into vertical sub-bands, adjacent high-frequency sub-bands, and low-frequency sub-bands.
[0065] For example, in conjunction with reference Figure 3 , Figure 3This is a framework diagram of the dual-tree complex wavelet transform provided by the present invention. The infrared image I(x,y) undergoes L-level asymmetric dual-tree complex wavelet transform decomposition. In each layer l (l=1,2,...,L), six adjacent high-frequency sub-bands in six directions are extracted, including ±15°, ±45°, and ±75° directions. Specifically, the vertical (90°) sub-band is extracted, as it exhibits high-energy characteristics at locations of vertical stripe inhomogeneity. Simultaneously, the 75° and 105° sub-bands adjacent to the vertical direction are extracted for subsequent contrast calculations. The low-frequency sub-bands of the Lth layer are retained as basic image information.
[0066] It should be noted that the asymmetric dual-tree complex wavelet transform used in this invention breaks the direction selection structure of the original dual-tree complex wavelet transform and enhances the selectivity in the vertical direction to satisfy the characteristics of non-uniformity of vertical stripes.
[0067] S203: Perform feature extraction processing on the infrared image at at least two specific scales to obtain the initial image at each scale.
[0068] For example, feature extraction processing is performed on the infrared image at at least two specific scales, the specific scale selection... (respectively, original, downsampled ×2, downsampled ×4), after passing through an anti-aliasing Gaussian filter ( After downsampling, initial images at each scale are obtained.
[0069] S204: Convolve the initial image at each scale using preset horizontal and vertical convolution kernels respectively to obtain the horizontal and vertical gradients of the infrared image at each scale.
[0070] For example, the initial image Is is convolved with the horizontal convolution kernel Sx and the vertical convolution kernel Sy respectively to obtain the horizontal gradient Gx=Is. Sx and the vertical gradient Gy=Is Sy. Among them, The convolution operation is represented by the following formula: ; The vertical convolution kernel Sy is shown in the following formula: .
[0071] S205: Determine the gradient magnitude at each scale based on the horizontal and vertical gradients of the infrared images at each scale, and fuse the gradient magnitude maps at least two different scales to obtain a fused gradient map.
[0072] For example, the gradient magnitude of each pixel is determined according to the following formula: ; in, This is a gradient magnitude map at scale s. The gradient magnitude reflects the edge intensity distribution of the image at this scale. The larger the value, the more prominent the edge.
[0073] Furthermore, the Sigmoid weighting function is used to fuse gradient magnitudes at multiple scales to obtain a fused gradient map, which can be represented as follows: ; in, This parameter, used to control the weight decay rate, is set to 1.5. To fuse gradient maps.
[0074] S206: Compress the edge region of the fused gradient map to obtain an edge response map. Based on the comparison between the pixel value of each pixel in the edge response map and the preset high threshold and preset low threshold, determine the effective edge pixels and convert the effective edge pixels into an edge binary mask.
[0075] S207: Compress the edge regions of the fused gradient map to obtain the edge response map.
[0076] For example, by combining the "thermal radiation gradient direction information" of the fused gradient map with that of the infrared image, the gradient map is "direction-sensitively enhanced"—highlighting edges that are consistent with the direction of thermal radiation (such as the real edges of the hot and cold interface of an object) and suppressing irrelevant false gradients.
[0077] Furthermore, the enhanced gradient map is compressed using non-maximum suppression, and the size of the suppression window is dynamically adjusted according to the local curvature. Along the edge normal direction, only local maxima are retained, while other points are suppressed, thereby achieving edge refinement and obtaining a single-pixel-wide edge response map.
[0078] S208: The pixels whose pixel values in the edge response map are greater than a preset high threshold are taken as strong edge pixels, and the strong edge pixels are the first valid edge pixels.
[0079] S209: Select the pixels whose pixel values in the edge response map are less than or equal to a preset high threshold and greater than a preset low threshold as candidate edge pixels.
[0080] S210: Determine the second effective edge pixel based on the connectivity between strong edge pixels and candidate edge pixels.
[0081] S211: Convert the first effective edge pixel and the second effective edge pixel into an edge binary mask.
[0082] For example, further, a dual-threshold processing is performed on the edge response map: a preset high threshold is set as the 90th percentile of the edge response map to identify strong edges; a preset low threshold is set as the 50th percentile of the edge response map to identify candidate edges. Pixels in the edge response map whose pixel values are greater than the preset high threshold are identified as strong edge pixels, and these strong edge pixels are the first effective edge pixels. Pixels in the edge response map whose pixel values are less than or equal to the preset high threshold and greater than the preset low threshold are identified as candidate edge pixels. If a strong edge pixel and a candidate edge pixel are connected, the candidate edge pixel is identified as a second effective edge pixel; if a strong edge pixel and a candidate edge pixel are not connected, the candidate edge pixel is identified as a pseudo edge pixel and is discarded. Further, the continuity and geometric consistency of the first and second effective edge pixels are checked, isolated points and short edge fragments (length < 5 pixels) are removed, and the first and second effective edge pixels after removing isolated points and short edge fragments are converted into a binary edge mask.
[0083] S212: Construct a threshold function based on the local energy of the vertical sub-band and adjacent high-frequency sub-bands, and determine the noise-dominant region mask based on the local energy of the vertical sub-band.
[0084] For example, the local energy of the vertical sub-band is calculated using a noise energy map, i.e.: ; The window size (2K+1)×(2L+1) is set according to the width of the noise stripes. Let K be the local energy of the vertical subband, K be the half-size of the window in the horizontal direction, L be the half-size of the window in the vertical direction, k be the downsampling scale level, and l be the weight index during gradient fusion. For vertical sub-bands.
[0085] For the energy map in the vertical direction, a threshold function needs to be set to remove stripe non-uniformity. The threshold function is designed to take into account both the energy in the vertical direction and the contrast between adjacent directions, i.e.: ; Where T(x,y) is the threshold function, and α and β are empirical weights that control the contributions of absolute energy and relative contrast, respectively. For the local energy of the vertical subband, for Local energy of adjacent high-frequency subbands in the direction, for The local energy of adjacent high-frequency subbands in the direction. This threshold function has adaptive characteristics: in the vertical stripe region, E90° is significantly higher than that in the adjacent direction, and T(x,y) is larger; in the normal texture region, the energy in each direction is similar, and T(x,y) is smaller.
[0086] Furthermore, the local energy distribution obtained based on the above calculations... The noise-dominant region is extracted through threshold segmentation. Specifically, the energy values of all pixels in the entire image are statistically analyzed, and the 95th percentile of the global energy distribution is used as the threshold. This threshold effectively distinguishes noise-dominant regions from normal texture regions. The resulting noise-dominant region mask is as follows: ; in, For noise-dominant regions, For the local energy of the vertical subband, This is the energy threshold.
[0087] S213: Combine the edge binary mask with the noise-dominant region mask, and use a threshold function to process the vertical sub-band. Then, perform an inverse transformation between the processed vertical sub-band and the adjacent high-frequency sub-band in the vertical direction to obtain the corrected infrared image.
[0088] For example, an edge binary mask With threshold function T(x,y) and noise-dominant region mask Together, they achieve refined processing of the vertical sub-band coefficients, as shown in the following formula: ; in, The vertical sub-band coefficient after processing. For edge binary mask, Let T(x,y) be the mask for the noise-dominant region, and T(x,y) be the threshold function. This represents the sub-band coefficient in the vertical direction.
[0089] Furthermore, the corrected infrared image can be obtained by performing an inverse asymmetric dual-tree complex wavelet transform on the processed vertical sub-band, adjacent high-frequency sub-band, and low-frequency sub-band.
[0090] like Figure 3 As shown, Figure 3 This is a schematic diagram of an embodiment of a single-image non-uniformity correction system based on space-frequency joint analysis provided by the present invention. A single-image non-uniformity correction system 10 based on space-frequency joint analysis is provided, comprising: Image acquisition module 11 is used to acquire infrared images.
[0091] The image decomposition module 12 is used to decompose the infrared image into vertical sub-bands, adjacent high-frequency sub-bands, and low-frequency sub-bands.
[0092] The edge binary mask acquisition module 13 is used to perform feature extraction processing on the infrared image at least two specific scales, determine the gradient magnitude at each scale, fuse the gradient magnitude maps at least two different scales to obtain a fused gradient map, compress the edge region of the fused gradient map to obtain an edge response map, determine the effective edge pixels based on the comparison between the pixel value of each pixel in the edge response map and a preset high threshold and a preset low threshold, and convert the effective edge pixels into an edge binary mask.
[0093] The noise-dominant region mask module 14 is used to construct a threshold function based on the local energy of the vertical sub-band and adjacent high-frequency sub-bands, and to determine the noise-dominant region mask based on the local energy of the vertical sub-band.
[0094] The infrared image correction module 15 is used to combine the edge binary mask image with the noise-dominant region mask, and to process the vertical sub-band using a threshold function. The processed vertical sub-band, adjacent high-frequency sub-band and low-frequency sub-band are inversely transformed to obtain the corrected infrared image.
[0095] For example, in the acquisition module 11, an infrared image is acquired. In the image decomposition module 12, the infrared image is decomposed into a vertical sub-band, adjacent high-frequency sub-bands, and a low-frequency sub-band. In the edge binary mask acquisition module 13, the infrared image is subjected to feature extraction processing at least two specific scales to obtain initial images at each scale; the initial images at each scale are convolved using preset horizontal and vertical convolution kernels to obtain the horizontal and vertical gradients of the infrared image at each scale; the gradient magnitude at each scale is determined based on the horizontal and vertical gradients of the infrared image at each scale; and the gradient magnitude maps at at least two different scales are fused to obtain a fused gradient map. The edge regions of the fused gradient map are compressed to obtain an edge response map. Pixels in the edge response map whose pixel values are greater than a preset high threshold are designated as strong edge pixels, which are then the first effective edge pixels. Pixels in the edge response map whose pixel values are less than or equal to a preset high threshold and greater than a preset low threshold are designated as candidate edge pixels. Second effective edge pixels are determined based on the connectivity between strong edge pixels and candidate edge pixels. The first and second effective edge pixels are then converted into a binary edge mask. In the noise-dominant region mask module 14, a threshold function is constructed based on the local energy of the vertical sub-band and adjacent high-frequency sub-bands, and the noise-dominant region mask is determined based on the local energy of the vertical sub-band. In the infrared image correction module 15, the binary edge mask is combined with the noise-dominant region mask, and the threshold function is used to process the vertical sub-band. The processed vertical sub-band, adjacent high-frequency sub-band, and low-frequency sub-band are then inversely transformed to obtain the corrected infrared image.
[0096] like Figure 4 As shown, Figure 4 This is a schematic diagram of the structure of an embodiment of the medium provided by the present invention. The medium 30 stores at least one computer program 21, which is executed by a processor to implement... Figure 1 and Figure 2 The method shown is detailed above and will not be repeated here. In one embodiment, the medium 20 can be a storage chip, hard disk, portable hard disk, USB flash drive, optical disk, or other read / write storage device, or even a server, etc.
[0097] Furthermore, the processes depicted in the accompanying drawings do not necessarily have to be performed in the specific or sequential order shown to achieve the desired result. In some implementations, multitasking and parallel processing are possible or may be advantageous.
[0098] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments of apparatus, devices, and non-volatile computer-readable storage media are basically similar to the method embodiments, and therefore described more simply; relevant parts can be referred to the descriptions of the method embodiments.
[0099] The apparatus, device, non-volatile computer-readable storage medium and method provided in the embodiments of this specification are corresponding. Therefore, the apparatus, device and non-volatile computer storage medium also have similar beneficial technical effects as the corresponding method. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the corresponding apparatus, device and non-volatile computer storage medium will not be repeated here.
[0100] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.
[0101] For ease of description, the above apparatus is described by dividing it into various functional units. Of course, in implementing this specification, the functions of each unit can be implemented in one or more software and / or hardware components. Those skilled in the art will understand that the embodiments of this specification can be provided as methods, systems, or computer program products. Therefore, the embodiments of this specification can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, the embodiments of this specification can take the form of computer program products implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0102] This specification is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this specification. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations and / or block diagrams. Figure 1One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0103] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0104] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0105] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.
[0106] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.
[0107] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.
[0108] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0109] This specification can be described in the general context of computer-executable instructions that are executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. This specification can also be practiced in distributed computing environments, where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.
[0110] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.
[0111] The above description discloses only preferred embodiments of the present invention and should not be construed as limiting the scope of the present invention. Therefore, equivalent variations made in accordance with the claims of the present invention are still within the scope of the present invention.
Claims
1. A method for non-uniformity correction of a single image based on spatial-frequency joint method, characterized in that, The method includes: Acquire infrared images; The infrared image is decomposed into vertical sub-bands, adjacent high-frequency sub-bands, and low-frequency sub-bands; The infrared image is subjected to feature extraction processing at at least two specific scales to determine the gradient magnitude at each scale, and the gradient magnitude maps at at least two different scales are fused to obtain a fused gradient map. The edge regions of the fused gradient map are compressed to obtain an edge response map. Based on the comparison between the pixel value of each pixel in the edge response map and the preset high threshold and preset low threshold, the effective edge pixels are determined and the effective edge pixels are converted into an edge binary mask. A threshold function is constructed based on the local energy of the vertical sub-band and the adjacent high-frequency sub-band, and a noise-dominant region mask is determined based on the local energy of the vertical sub-band. The edge binary mask is combined with the noise-dominant region mask, and the vertical sub-band is processed using the threshold function. The processed vertical sub-band, adjacent high-frequency sub-band, and low-frequency sub-band are then inversely transformed to obtain the corrected infrared image.
2. The method for non-uniformity correction of a single image based on spatial-frequency joint analysis according to claim 1, characterized in that, The process of processing the infrared image at at least two specific scales to determine the gradient magnitude at each scale specifically includes: The infrared image is subjected to feature extraction processing at at least two specific scales to obtain initial images at each scale; The initial image at each scale is convolved using preset horizontal and vertical convolution kernels to obtain the horizontal and vertical gradients of the infrared image at each scale. The gradient magnitude at each scale is determined based on the horizontal and vertical gradients of the infrared image at each scale.
3. The method for non-uniformity correction of a single image based on spatial-frequency joint method according to claim 1, characterized in that, The step of fusing at least two gradient magnitude maps of different scales to obtain a fused gradient map specifically includes: according to The gradient magnitude maps at least two different scales are fused to obtain a fused gradient map, wherein... This is a gradient magnitude plot on an s-scale. These are parameters used to control the weight decay rate. To fuse gradient maps.
4. The method for non-uniformity correction of a single image based on spatial-frequency joint method according to claim 1, characterized in that, The step of determining effective edge pixels based on a comparison between the pixel value of each pixel in the edge response map and a preset high threshold and a preset low threshold, and then converting the effective edge pixels into an edge binary mask, specifically includes: The pixels whose pixel values in the edge response map are greater than a preset high threshold are defined as strong edge pixels, and the strong edge pixels are the first valid edge pixels. The pixels whose pixel values in the edge response map are less than or equal to the preset high threshold and greater than the preset low threshold are selected as candidate edge pixels. The second effective edge pixel is determined based on the connectivity between the strong edge pixel and the candidate edge pixel; The first and second effective edge pixels are converted into a binary edge mask.
5. The method for non-uniformity correction of a single image based on spatial-frequency joint analysis according to claim 4, characterized in that, The step of determining effective edge pixels based on the connectivity between the strong edge pixels and the candidate edge pixels specifically includes: If the strong edge pixel and the candidate edge pixel are connected, then the candidate edge pixel is the second valid edge pixel; If the strong edge pixel and the candidate edge pixel are not connected, then the candidate edge pixel is a pseudo edge pixel, and the pseudo edge pixel is discarded.
6. The method for non-uniformity correction of a single image based on spatial-frequency joint method according to claim 1, characterized in that, The step of constructing a threshold function based on the local energy of the vertical sub-band and the adjacent high-frequency sub-band specifically includes: according to Determine the local energy of the vertical sub-band, where, Let K be the local energy of the vertical subband, K be the half-size of the window in the horizontal direction, L be the half-size of the window in the vertical direction, k be the downsampling scale level, and l be the weight index during gradient fusion. Vertical sub-band; according to Construct a threshold function, where T(x,y) is the threshold function, and α and β are empirical weights. The local energy of the vertical subband. for Local energy of adjacent high-frequency subbands in the direction, for Local energy of adjacent high-frequency subbands in the direction.
7. The method for non-uniformity correction of a single image based on spatial-frequency joint analysis according to claim 6, characterized in that, The method for determining the noise-dominant region mask based on the local energy of the vertical sub-bands specifically includes: according to Determine the mask for the noise-dominant region, where, For noise-dominant regions, The local energy of the vertical subband. This is the energy threshold.
8. The method for non-uniformity correction of a single image based on spatial-frequency joint analysis according to claim 7, characterized in that, The step of combining the edge binary mask image with the noise-dominant region mask and processing the vertical sub-band using the threshold function specifically includes: according to The vertical sub-strips are processed; in, The vertical sub-band coefficient after processing. For edge binary mask, Let T(x,y) be the mask for the noise-dominant region, and T(x,y) be the threshold function. This represents the sub-band coefficient in the vertical direction.
9. A single-image non-uniformity correction system based on space-frequency joint method, characterized in that, The system includes: The image acquisition module is used to acquire infrared images; An image decomposition module is used to decompose the infrared image into vertical sub-bands, adjacent high-frequency sub-bands, and low-frequency sub-bands; An edge binary mask acquisition module is used to perform feature extraction processing on the infrared image at least two specific scales, determine the gradient magnitude at each scale, fuse the gradient magnitude maps at at least two different scales to obtain a fused gradient map, compress the edge region of the fused gradient map to obtain an edge response map, determine the effective edge pixels based on the comparison between the pixel value of each pixel in the edge response map and the preset high threshold and preset low threshold, and convert the effective edge pixels into an edge binary mask. The noise-dominant region mask module is used to construct a threshold function based on the local energy of the vertical sub-band and the adjacent high-frequency sub-band, and to determine the noise-dominant region mask based on the local energy of the vertical sub-band. The infrared image correction module is used to combine the edge binary mask image with the noise-dominant region mask, and use the threshold function to process the vertical sub-band. The processed vertical sub-band, adjacent high-frequency sub-band and low-frequency sub-band are inversely transformed to obtain the corrected infrared image.
10. A computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to perform the steps of the method as claimed in any one of claims 1 to 8.