CN114463391A - SAR image registration method using image block matching - Google Patents
SAR image registration method using image block matching Download PDFInfo
- Publication number
- CN114463391A CN114463391A CN202210091148.5A CN202210091148A CN114463391A CN 114463391 A CN114463391 A CN 114463391A CN 202210091148 A CN202210091148 A CN 202210091148A CN 114463391 A CN114463391 A CN 114463391A
- Authority
- CN
- China
- Prior art keywords
- matrix
- image
- position deviation
- amplitude
- column
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
- G06T2207/10044—Radar image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20021—Dividing image into blocks, subimages or windows
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20048—Transform domain processing
- G06T2207/20056—Discrete and fast Fourier transform, [DFT, FFT]
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses an SAR image registration method by utilizing image block matching, which comprises the steps of respectively matching two images to be matched in a block mode to obtain a position deviation matrix and an amplitude mean ratio matrix, setting a threshold, removing singular values of the two matrixes, reducing matching errors, multiplying the SAR images according to coefficients of the amplitude mean ratio, and interpolating the SAR images according to parameters of position deviation.
Description
Technical Field
The invention belongs to the technical field of signal processing, and particularly relates to an image registration technology.
Background
The main function of the video SAR is to detect and track a ground moving target, and the premise of realizing the detection of the moving target by using change detection is high-precision image registration. After the SAR image is subjected to high-precision registration, the moving target is detected through change detection, and then the moving target tracking is realized on the basis of the detection result.
The multi-frame SAR image after registration can realize high-precision estimation of the speed of a moving target and can also realize image splicing and multi-view image fusion, and the image registration is one of important research contents of SAR image processing.
Existing registration methods for SAR images include:
in synthetic aperture radar image registration method using SAR-FAST corner detection, a method for SAR image matching using corners is disclosed. The method firstly inhibits speckle noise, and then extracts and screens corner regions to obtain a stable matching technology with good repeatability.
In the SAR image registration method based on point characteristics under affine-all deformation, an image registration method based on a radiation deformation model is disclosed. The method comprises the steps of firstly estimating scale transformation parameters of an image by using a particle swarm optimization algorithm, and then converting an operator SIFT registration image by using scale invariant features.
An image registration method is disclosed in 'a new SAR image registration algorithm based on local invariant features'. The method comprises the steps of firstly detecting image corners by using an accelerated segmentation detection characteristic FAST, secondly performing non-deformation description on the FAST characteristic by using a DAISY descriptor, then realizing image registration by using a KD tree, Euclidean distance and RANSAC algorithm, and finally realizing image interpolation and image transformation by using affine transformation.
In the existing SAR image registration method, feature point extraction algorithms such as SIFT and SURF are mostly used, a complex feature screening method is used for screening matching points, the calculation efficiency is low, more mismatch conditions exist, and an efficient SAR image registration method is needed on the basis of ensuring the image registration accuracy.
Disclosure of Invention
The invention provides an SAR image registration method by utilizing image block matching in order to solve the problems in the prior art, the position deviation of different areas of an SAR image is obtained through the block registration of the SAR image, the high-precision registration of the SAR image is realized through interpolation, and the detection and tracking of a moving target are guaranteed.
The method comprises the steps of respectively matching two images to be matched in a blocking mode to obtain a position deviation matrix and an amplitude mean value ratio matrix, setting a threshold value, eliminating singular values of the two matrices, reducing matching errors, multiplying SAR images according to coefficients of the amplitude mean value ratio, and interpolating the SAR images according to parameters of position deviation.
Further, the two images are partitioned according to a certain size, the correlation coefficient between the partitions is calculated by adopting fast Fourier transform, the position corresponding to the maximum value in the correlation coefficient matrix is selected, the position deviation of the partitions is calculated by combining the partition sizes, the corresponding partitions of the two images are matched, the position deviation of all the partitions of the two images is calculated, a row position deviation matrix and a column position deviation matrix are generated, the amplitude-to-mean ratio between the two images is calculated, and an amplitude-to-mean ratio matrix is generated.
The purpose of the matching of the two image blocks is to calculate the position deviation and the amplitude deviation between all the blocks of the two images, and the position deviation comprises the line position deviation and the column position deviation.
Setting the line and column size of the main SAR image and the auxiliary SAR image as NrAnd NcThe row and column size of the partitioned image block is NbkThe number of matching steps is NstepMaster image I1The ith row and j column of image blocks are I1 (i,j)Auxiliary image I2The ith row and j column of image blocks are I2 (i,j)Image block I1 (i,j)And I2 (i,j)Is ρ, FFT2 is a two-dimensional fourier transform, IFFT2 is a two-dimensional inverse fourier transform, and conj (·) is a conjugate calculation, then ρ ═ IFFT2[ FFT2 (I)1 (i,j))·conj(FFT2(I2 (i,j)))]The maximum value of ρ is [ a, b ]]Setting the matrix M as arcmax (rho)rAnd McFor the line position deviation and the column position deviation of each image block of the main and auxiliary images, the position deviation of the ith line and the jth column of image blocks isLet a matrix etaAThe amplitude mean value ratio of each image block of the main and auxiliary images is etaA(i,j)=mean(I1 (i,j))/mean(I2 (i,j))。
And calculating the position deviation and the amplitude-to-average ratio of each image block, realizing the matching between the image blocks, and obtaining a row position deviation matrix, a column position deviation matrix and an amplitude-to-average ratio matrix between the main image and the auxiliary image.
And further, comparing the median of each element in the row and column position deviation matrix and the amplitude-average ratio matrix with the median of eight surrounding connected elements, if the absolute value of the difference exceeds a threshold value, re-assigning the median, otherwise, repeating assignment for three times to obtain the row and column position deviation matrix and the amplitude-average ratio matrix without singular values.
Errors with certain probability exist in image block matching calculation, so that large errors exist in row-column position deviation, namely position deviation singular values, the singular values need to be removed in matrix matching, and the image matching precision is guaranteed.
Let the position deviation threshold be TdifThe median of eight connected elements of the ith row and the j column elements of the row and column position deviation matrix is Vr(i, j) and Vc(i, j) usingAndeliminating singular values of a row-column position deviation matrix, and setting an amplitude-to-mean ratio deviation threshold value as TηSetting the median of eight connected elements of the ith row and j column elements of the amplitude-to-mean ratio matrix as Vη(i, j) usingAnd removing singular values of the amplitude-to-average ratio matrix.
The calculation is repeated three times enough to eliminate the singular values of the position deviation matrix and the amplitude-to-average ratio matrix.
Further, Gaussian smoothing is carried out on the position deviation matrix and the amplitude mean value ratio matrix, the position deviation matrix of the row and the column and the amplitude mean value ratio matrix are interpolated to enable the size to be the same as the size of the SAR image, the amplitude mean value ratio matrix is multiplied by corresponding elements of the SAR image, the amplitude of the SAR image is corrected, the SAR image is interpolated according to parameters in the position deviation matrix of the row and the column, and a cubic spline interpolation method is adopted to obtain the matched SAR image.
Position deviation matrix MrAnd McGaussian smooth filtering is carried out, position registration error is reduced, and interpolation is carried out to obtain M'rAnd M'cThe size of the matrix is the same as that of the original image, and the amplitude-to-mean ratio matrix eta is obtainedAGaussian smooth filtering is carried out, amplitude registration errors are reduced, and eta 'is obtained through interpolation'AIs the same as the original image in line eta'AThe amplitude ratio of (1) is corrected by corresponding multiplication of two matrix elements2Amplitude of (I)2_amp=I2*η′AAccording to M'rAnd M'cBy dimensional two-dimensional interpolation of1And I2Image registration of (2).
The invention has the beneficial effects that: the method is characterized in that the image is subjected to block matching to obtain the position deviation and the amplitude-to-mean ratio of each region of the image, numerical values with larger matching errors are removed through singular value elimination and numerical value smoothing to obtain high-precision position deviation parameters and amplitude-to-mean ratios, and high-precision image registration is realized through image interpolation.
Drawings
Fig. 1 is a SAR main image, fig. 2 is a SAR auxiliary image, fig. 3 is a difference image of the main and auxiliary images, fig. 4 is a magnitude map of a correlation coefficient matrix, fig. 5 is a magnitude map of a row position deviation matrix, fig. 6 is a magnitude map of a column position deviation matrix, fig. 7 is a magnitude map of a magnitude-to-average ratio matrix, fig. 8 is a magnitude map of a row position deviation matrix after singular value removal, fig. 9 is a magnitude map of a column position deviation matrix after singular value removal, fig. 10 is a registered SAR auxiliary image, and fig. 11 is a difference image of the registered main and auxiliary images.
Detailed Description
The technical scheme of the invention is specifically explained in the following by combining the attached drawings.
Selecting 2 SAR radar images at different time, wherein the size of the image is Nr2000 and Nc2000, the size of each row and column of image blocks is Nbk300, the number of steps when the image blocks match is Nstep=100。
The main image and the auxiliary image are respectively shown in fig. 1 and fig. 2, the absolute value of the difference between the two images has more residuals, as shown in fig. 3, the amplitude of the matrix of matched correlation coefficients obtained by block matching is shown in fig. 4, the amplitude of the matrix of row position deviation and the matrix of column position deviation is calculated as shown in fig. 5 and fig. 6, and the average value of the image block amplitudes is calculated as shown in fig. 7.
Let Tdif=3,TηRepeat the calculation of matrix M as 0.1r、McAnd ηAAnd 3 times in total, all singular values in the matrix are eliminated, and a row-column position deviation matrix and an amplitude-mean ratio matrix after elimination are obtained, wherein as shown in fig. 8 and 9, the amplitude-mean ratio matrix has no singular value and is not changed.
Respectively combining the matrix Mr、McAnd ηAAnd performing interpolation to make the size the same as that of the main image and the auxiliary image to obtain the auxiliary image after amplitude correction, and performing cubic spline interpolation on the auxiliary image to obtain the SAR image after image registration, as shown in FIG. 10.
The difference between the main image and the auxiliary image is obtained by taking the difference and then taking the absolute value, as shown in fig. 11, as can be seen from the comparison between fig. 11 and fig. 3, the difference between the images is obviously reduced after the images are registered.
The above-described embodiments are not intended to limit the present invention, and any modifications, equivalents, improvements, etc. made within the spirit and principle of the present invention are included in the scope of the present invention.
Claims (7)
1. A SAR image registration method using image block matching is characterized by comprising the following steps: the method comprises the steps of respectively matching two images to be matched in a blocking mode to obtain a position deviation matrix and an amplitude mean value ratio matrix, setting a threshold value, eliminating singular values of the two matrices, multiplying SAR images according to coefficients of the amplitude mean value ratio, and interpolating the SAR images according to parameters of position deviation.
2. The SAR image registration method by image block matching according to claim 1, wherein the two image blocks are matched respectively, comprising: the method comprises the steps of partitioning two images according to a certain size, calculating correlation coefficients between the partitions by adopting fast Fourier transform, selecting a position corresponding to a maximum value in a correlation coefficient matrix, calculating position deviation of the partitions by combining the sizes of the partitions, matching the corresponding partitions of the two images, calculating position deviation of all the partitions of the two images, generating a row position deviation matrix and a column position deviation matrix, calculating an amplitude-to-mean ratio between the two images, and generating an amplitude-to-mean ratio matrix.
3. The method of claim 2, wherein the two image patches are matched separately, further comprising: setting the line and column size of the main SAR image and the auxiliary SAR image as NrAnd NcThe row and column size of the partitioned image block is NbkThe number of matching steps is NstepMaster image I1The ith row and j column of image blocks are I1 (i,j)Auxiliary image I2The ith row and j column of image blocks are I2 (i,j)Image block I1 (i,j)And I2 (i,j)Is ρ, FFT2 is a two-dimensional fourier transform, IFFT2 is a two-dimensional inverse fourier transform, and conj (·) is a conjugate calculation, then ρ ═ IFFT2[ FFT2 (I)1 (i,j))·conj(FFT2(I2 (i ,j)))]The maximum value of ρ is [ a, b ]]Setting the matrix M as arcmax (rho)rAnd McFor the line position deviation and the column position deviation of each image block of the main and auxiliary images, the position deviation of the ith line and the jth column of image blocks isLet a matrix etaAThe amplitude mean value ratio of each image block of the main and auxiliary images is etaA(i,j)=mean(I1 (i,j))/mean(I2 (i,j))。
4. The SAR image registration method using image block matching according to claim 1, wherein the removing singular values of two matrices comprises: comparing the median of each element in the row and column position deviation matrix and the amplitude-average ratio matrix with the median of the eight surrounding communicated elements, if the absolute value of the difference exceeds a threshold value, re-assigning the median, otherwise, repeating the assignment for three times to obtain the row and column position deviation matrix and the amplitude-average ratio matrix without singular values.
5. The SAR image registration method using image block matching according to claim 4, wherein the removing the singular values of the two matrices further comprises: let the position deviation threshold be TdifThe median of eight connected elements of the ith row and the j column elements of the row and column position deviation matrix is Vr(i, j) and Vc(i, j) usingAndeliminating singular values of a row-column position deviation matrix, and setting an amplitude-to-mean ratio deviation threshold value as TηSetting the median of eight connected elements of the ith row and j column elements of the amplitude-to-mean ratio matrix as Vη(i, j) usingAnd removing singular values of the amplitude-to-average ratio matrix.
6. The method of claim 1, wherein the interpolating the SAR image comprises: and performing Gaussian smoothing on the position deviation matrix and the amplitude mean value ratio matrix, interpolating the position deviation matrix of the row and the column and the amplitude mean value ratio matrix to ensure that the size of the position deviation matrix and the amplitude mean value ratio matrix is the same as the size of the SAR image, multiplying the amplitude mean value ratio matrix with corresponding elements of the SAR image, correcting the amplitude of the SAR image, interpolating the SAR image according to parameters in the position deviation matrix of the row and the column, and obtaining the matched SAR image by adopting a cubic spline interpolation method.
7. The method of SAR image registration with image block matching according to claim 6, wherein the interpolating SAR image further comprises: position deviation matrix MrAnd McGaussian smooth filtering is carried out, position registration error is reduced, and interpolation is carried out to obtain M'rAnd M'cThe size of the matrix is the same as that of the original image, and the amplitude-to-mean ratio matrix eta is obtainedAGaussian smooth filtering is carried out, amplitude registration errors are reduced, and eta 'is obtained through interpolation'AIs the same as the original image in line eta'AThe amplitude ratio of (1) is corrected by corresponding multiplication of two matrix elements2Amplitude of (I)2_amp=I2*η′AAccording to M'rAnd M'cTwo-dimensional interpolation of the size of (2).
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210091148.5A CN114463391A (en) | 2022-01-26 | 2022-01-26 | SAR image registration method using image block matching |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210091148.5A CN114463391A (en) | 2022-01-26 | 2022-01-26 | SAR image registration method using image block matching |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114463391A true CN114463391A (en) | 2022-05-10 |
Family
ID=81412246
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210091148.5A Pending CN114463391A (en) | 2022-01-26 | 2022-01-26 | SAR image registration method using image block matching |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114463391A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116051435A (en) * | 2022-08-23 | 2023-05-02 | 荣耀终端有限公司 | Image fusion method and electronic equipment |
CN116563357A (en) * | 2023-07-10 | 2023-08-08 | 深圳思谋信息科技有限公司 | Image matching method, device, computer equipment and computer readable storage medium |
-
2022
- 2022-01-26 CN CN202210091148.5A patent/CN114463391A/en active Pending
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116051435A (en) * | 2022-08-23 | 2023-05-02 | 荣耀终端有限公司 | Image fusion method and electronic equipment |
CN116051435B (en) * | 2022-08-23 | 2023-11-07 | 荣耀终端有限公司 | Image fusion method and electronic equipment |
CN116563357A (en) * | 2023-07-10 | 2023-08-08 | 深圳思谋信息科技有限公司 | Image matching method, device, computer equipment and computer readable storage medium |
CN116563357B (en) * | 2023-07-10 | 2023-11-03 | 深圳思谋信息科技有限公司 | Image matching method, device, computer equipment and computer readable storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107301661B (en) | High-resolution remote sensing image registration method based on edge point features | |
CN114463391A (en) | SAR image registration method using image block matching | |
CN109993800A (en) | A kind of detection method of workpiece size, device and storage medium | |
CN109523585B (en) | Multisource remote sensing image feature matching method based on direction phase consistency | |
CN108765476B (en) | Polarized image registration method | |
CN112686935A (en) | Airborne depth sounding radar and multispectral satellite image registration method based on feature fusion | |
CN108460792B (en) | Efficient focusing stereo matching method based on image segmentation | |
CN110211169B (en) | Reconstruction method of narrow baseline parallax based on multi-scale super-pixel and phase correlation | |
EP3293700A1 (en) | 3d reconstruction for vehicle | |
CN113223066B (en) | Multi-source remote sensing image matching method and device based on characteristic point fine tuning | |
CN113935984A (en) | Multi-feature fusion method and system for detecting infrared dim small target in complex background | |
CN103824287A (en) | Robust plane fitting-based phase correlation sub-pixel matching method | |
CN110942102B (en) | Probability relaxation epipolar matching method and system | |
CN110688440B (en) | Map fusion method suitable for less sub-map overlapping parts | |
CN117152330B (en) | Point cloud 3D model mapping method and device based on deep learning | |
CN110969650B (en) | Intensity image and texture sequence registration method based on central projection | |
CN103824286A (en) | Singular value decomposition-random sample consensus (SVD-RANSAC) sub-pixel phase correlation matching method | |
CN117030620A (en) | Method and device for adjusting regional network based on multisource optical remote sensing satellite image | |
CN116958416A (en) | Three-dimensional modeling method, device, system and storage medium | |
CN107103579A (en) | A kind of RANSAC improved methods towards image mosaic | |
CN116563354A (en) | Laser point cloud registration method combining feature extraction and clustering algorithm | |
CN112991408B (en) | Large-scene high-resolution remote sensing image self-adaptive area multi-feature registration method and system | |
CN109886988A (en) | A kind of measure, system, device and the medium of Microwave Imager position error | |
CN113592744B (en) | Geometric fine correction method suitable for high-resolution remote sensing image | |
CN115601278A (en) | High-precision motion error compensation method based on sub-image registration |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |