CN105261014B - A kind of multisensor Remote Sensing Images Matching Method - Google Patents
A kind of multisensor Remote Sensing Images Matching Method Download PDFInfo
- Publication number
- CN105261014B CN105261014B CN201510638280.3A CN201510638280A CN105261014B CN 105261014 B CN105261014 B CN 105261014B CN 201510638280 A CN201510638280 A CN 201510638280A CN 105261014 B CN105261014 B CN 105261014B
- Authority
- CN
- China
- Prior art keywords
- pointi
- image
- characteristic
- matching
- phase equalization
- 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.)
- Active
Links
Classifications
-
- 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
-
- 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
- G06T2207/20212—Image combination
Abstract
The invention discloses a kind of multisensor Remote Sensing Images Matching Method, belong to satellite image processing technology field, it can efficiently solve the geometric deformation and radiation difference problem between multisensor image.The Gaussian difference scale space of image is initially set up, and feature point detection is carried out in this space.Then the phase equalization model with illumination and contrast consistency is introduced, and structure phase equalization directional information is extended to it, simultaneously by the thought of SIFT descriptors, a kind of local feature description's symbol-local phase uniformity direction histogram is built using phase equalization characteristic value and characteristic direction.Same place identification finally is carried out as similarity measure using the Euclidean distance between descriptor, and rejects Mismatching point, so as to realize the accurate matching of remote sensing image.Compared to traditional matching process, the present invention can more effectively overcome the geometric deformation and radiation difference between remote sensing image, improve the accuracy of Image Matching.It is mainly used in satellite image processing.
Description
Technical field
The invention belongs to the matching technique of satellite image processing technology field, more particularly to remote sensing image.
Background technology
Image Matching is substantially the process that same place is identified between two width or several images, is widely used in remote sensing image
Registration, image joint and change detection etc..It is past between more sensing remote sensing images due to imaging mechanism and the difference of spectral characteristic
Toward notable geometry and radiation difference be present, so as to cause the difficulty for carrying out same place automatic identification between image larger.
Recently in computer vision field, local feature description accords with having obtained quick development, and is widely used in shadow
As matching.Wherein foremost local feature description is Scale Invariant Feature Transform (SIFT) operator,
Due to being widely used with rotation and scale invariability, SIFT in Remote Sensing Images Matching.Nevertheless,
SIFT descriptors are more sensitive the radiation difference image, therefore SIFT is difficult to preferably be applied to multisensor remote sensing image
Matching.On the basis of SIFT, have scholar propose in succession Speeded Up Robust Features (SURF),
The descriptors such as Oriented FAST and Rotated BRIEF (ORB).They have obtained larger carry in terms of computational efficiency
Rise, but these descriptions are still very sensitive for the radiation difference between image.In view of not only having between multisensor remote sensing image
There is geometric deformation, but also larger radiation difference be present, will utilize phase equalization (phase congruency) right here
Remote sensing image is matched.Phase equalization is a kind of feature extraction algorithm, has illumination and contrast consistency, in remote sensing shadow
As matching field has obtained relatively broad application, but the method for phase equalization is currently based on for yardstick between image and rotation
The different applicability of slip is weaker, and has only used the characteristic value of phase equalization, without the direction using phase equalization
Information, fail fully to excavate potentiality of the phase equalization in terms of feature extraction and description, performance is sane not enough, therefore
The present invention will be extended to phase equalization computation model, and phase equalization feature side is built using its odd symmetric filter
To, then using its characteristic value and characteristic direction, by means of the thought of SIFT descriptors, build a kind of local feature description's symbol-
Local phase uniformity direction histogram (local histogram of orientated phase congruency,
LHOPC), yardstick, rotation and the radiation difference between image can be preferably resisted, so as to realize multisensor remote sensing image
Accurate matching.
The content of the invention
It is an object of the invention to provide a kind of multisensor Remote Sensing Images Matching Method, and it can efficiently solve multisensor
Geometric deformation between image and radiation difference problem, automatically can be obtained between image largely be evenly distributed, stably of the same name
Point pair, realize the Accuracy Matching of image.
The purpose of the present invention is achieved through the following technical solutions:A kind of multisensor Remote Sensing Images Matching Method.
Comprise the following steps:
Step 1: using the gaussian kernel function of different scale to referring to image I1With image I to be matched2, convolution is carried out respectively
Computing, the Gaussian scale-space of image is established, and generation difference of Gaussian is made the difference by the adjacent two layers image of Gaussian scale-space
Metric space, then, extremum extracting is carried out in this space, extract the characteristic point PointI with scale invariabilityi(i=
1,2,3,….,N);
Step 2: being extended to phase equalization computation model, phase equalization characteristic direction is built, is then calculated high
The phase equalization characteristic value and characteristic direction of each pixel in this metric space on each layer image.
Step 3: utilizing phase equalization characteristic value and characteristic direction structure local feature description symbol -- local phase is consistent
Property direction histogram, is designated as LHOPC, specifically includes two parts of principal direction and feature description vectors:
(1) characteristic point PointI is extractediA point PointI in (i=1,2,3 ..., N)i, size is taken centered on it
For the neighborhood of 5 × 5 pixels, the phase equalization characteristic value and characteristic direction in neighborhood are calculated, forms phase equalization direction
Histogram, and principal direction of the peak value direction of histogram as characteristic point is selected, make descriptor that there is rotational invariance.
(2) coordinate system is established according to principal direction, and with characteristic point PointIiCentered on take size as 20 × 20 pixels
Neighborhood.Then the neighborhood is divided into 4 × 4 sub-regions, 8 directions will be divided into every sub-regions, and count each side
To phase equalization characteristic value, form phase equalization direction histogram, finally by the phase equalization in every sub-regions
Direction histogram links together, and forms final feature description vectors-local phase uniformity direction histogram (local
histogram of orientated phase congruency,LHOPC)。
Image I is referred to Step 4: calculating respectively1Upper characteristic point PointI1i(i=1,2,3 ..., N) and image to be matched
I2Upper characteristic point PointI2iThe LHOPC characteristic vectors of (i=1,2,3 ..., N), and similitude is used as by Euclidean distance and surveyed
Degree, using the ratio between the arest neighbors between characteristic vector and time nearest neighbor distance dratioTo be matched, work as dratioLess than or equal to given
Threshold value when, PointI1iAnd PointI2iThen it is considered as match point;
Step 5: in order that matching is more sane, here using two-way matching strategy, i.e., for referring to image I1On
Some characteristic point PointI1i, matched using LHOPC, obtain it in image I to be matched2On same place PointI2i,
Then with same matching way, PointI is obtained2iReferring to image I1Same place PointI ' corresponding to upper1i.If PointI1i
And PointI '1iIt is same point, then it is assumed that PointI1iAnd PointI2iIt is a pair of same places;
Step 6: according to step 4, PointI is traveled through1iEach point of (i=1,2,3 ..., N), obtain corresponding same
Famous cake PointI2i(i=1,2,3 ..., N), that is, same place is obtained to { PointI1i, PointI2i(i=1,2,3 ..., N).
Step 7: using stochastical sampling consistency algorithm, error matching points are rejected using projective transformation as geometrical constraint, are obtained
Correct same place pair finally is obtained, is designated as { PointID1i, PointID2i(i=1,2,3 ..., N), realize the accurate of image
Matching.
The beneficial effect of the present invention compared with prior art is:
1st, for matching difficult problem caused by geometric deformation between multisensor remote sensing image and radiation difference, the present invention carries
Gone out a kind of matching process based on local phase feature, by using phase equalization structure it is a kind of have yardstick, rotation and
The local feature description's symbol for radiating consistency carries out homotopy mapping, and it is main to compensate for traditional matching process (such as SIFT, SURF)
It is designed for geometric deformation between image, and to radiating the deficiency of difference more sensitivity, improve the efficiency of Image Matching.
2nd, in whole matching process, without manual intervention, can fully automatically identify same between multisensor remote sensing image
Famous cake, improve the actual production efficiency of matching.It is substantial amounts of test result indicates that, the accuracy of Image Matching reach 90% with
On, precision can reach within 1 pixel.
Brief description of the drawings
Fig. 1 is the overall flow figure of the present invention
Fig. 2 is the odd symmetric filter shape of log Gabor functions of the present invention
Fig. 3 is the schematic diagram of phase equalization characteristic direction of the present invention
Fig. 4 is the schematic diagram of local phase uniformity feature description vectors of the present invention
Fig. 5 is local phase uniformity feature description vectors interpolation schematic diagram of the present invention
Fig. 6 is bi-directional matching schematic diagram of the present invention
Embodiment
The present invention is described further below in conjunction with the accompanying drawings:
A kind of multisensor Remote Sensing Images Matching Method, what it was realized concretely comprises the following steps:
Step 1: using different scale σ gaussian kernel function to referring to image I1With image I to be matched2It is filtered to be formed
Gaussian scale-space, generation DoG metric spaces are made the difference by the adjacent two layers image of Gaussian scale-space.
In DoG spaces, by adjacent 8 of each pixel and same layer of intermediate layer (except the bottom and top)
Pixel and bilevel 18 adjacent pixels, 26 pixels are compared altogether, if the DoG values of the point are maximum or most
It is small, then it is considered as the characteristic point of candidate, then removes contrast point that is relatively low, and being located on edge, obtain final spy
Sign point PointIi (i=1,2,3 ..., N).
Step 2: being extended to phase equalization computation model, the characteristic direction of phase equalization is built, is calculated simultaneously
In Gaussian scale-space on each layer image each pixel phase equalization characteristic value and characteristic direction, be specifically divided into two
Step:
(1) the phase equalization characteristic value of image is calculated using log Gabor wavelets.
Wherein, PC (x, y) be phase equalization characteristic value, (x, y) be image coordinate, Wo(x, y) is frequency expansion
Weight factor, Ano(x, y) is amplitude of the picture point (x, y) on log Gabor filter yardstick n and direction o,Symbol table
Indicating value is that timing takes itself, and it is noise threshold otherwise to take 0, T, and ε is a constant for avoiding removing zero, Ano(x,y)ΔΦno(x,y)
It is the even symmetric filter e by log Gabor waveletsno(x, y) and odd symmetric filter ono(x, y) is calculated:
Wherein,E (x, y) is local energy
Flow function, it is each yardstick all directions even symmetric filter eno(x, y) and odd symmetric filter onoThe vector of (x, y) response results
Sum, it is defined as:
(2) the phase equalization characteristic direction of image is calculated using log Gabor odd symmetric filters.
In traditional phase equalization calculating process, phase equalization characteristic value can only be obtained, and the description of characteristic value
Performance is not sane enough.In consideration of it, being extended here to the computation model of phase equalization, the strange of Log Gabor functions is utilized
Balanced-filter ono(x, y) builds the characteristic direction of phase equalization, and characteristic direction represents that image feature changes most violent side
It is extremely important (Fig. 2) for the structure of local feature description's symbol to (being similar to gradient direction).Log Gabor functions it is strange right
Claim wave filter onoConvolution results represent image some direction energy variation.Here by the odd symmetric filter of multiple directions
As a result horizontal direction (X-axis) and vertical direction (Y-axis) are projected to respectively, can obtain the energy a and Vertical Square of horizontal direction respectively
To energy b, then calculate the arc-tangent value between them, that is, obtain phase equalization characteristic direction Φ.As shown in figure 3, (a)
For raw video, (b) is horizontal direction energy, and (c) is vertical direction energy, and (d) is phase equalization characteristic direction.
Φ=arctan (b, a)
Wherein, ono(θ) represents the odd symmetric filter convolution results on the θ of direction.
Step 3: calculating PointIi (i=1,2,3 ..., N) LHOPC local feature descriptions symbol, LHOPC includes feature
Point two parts (Fig. 4) of principal direction and feature description vectors, it is specially
1st, LHOPC principal directions:Extract PointIiA point PointI in (i=1,2,3 ..., N)i, centered on it
A certain size neighborhood is taken, phase equalization characteristic value and characteristic direction in neighborhood is calculated, it is straight to form phase equalization direction
Fang Tu.This histogram is uniformly divided into 36 equal portions, each equal portions are 10 °, and the phase equalization for counting each equal portions is special
Value indicative, and it is distance weighted using the progress of Gauss weight oeil de boeuf mouth, make the proportion shared by the pixel near the centre of neighbourhood bigger.Then
Select principal direction of the peak value direction as characteristic point of histogram, if in the presence of another equivalent to main peak value 80% peak value when,
The direction is considered as the auxiliary direction of the point.Final 3 histograms neighbouring using peak value carry out quadratic polynomial fitting, interpolation
Go out accurate principal direction, as shown in Fig. 4 (a).
2nd, LHOPC characteristic vectors:After characteristic point principal direction is obtained, coordinate system is established according to principal direction, and with feature
The neighborhood of 20 × 20 pixels is taken centered on point.Then the neighborhood is divided into 4 × 4 sub-regions, interior per sub-regions will divided
For 8 directions, and the phase equalization characteristic value in each direction is counted, form phase equalization direction histogram, finally will be every
Phase equalization direction histogram in sub-regions links together, and forms the characteristic vector of 4 × 4 × 8=128 dimensions, such as Fig. 4
(b) shown in.
3rd, in order to avoid boundary effect, during LHOPC is calculated, using tri-linear interpolation by each sampled point
Phase equalization characteristic value is assigned in adjacent histogram, and each histogram component (bin) is multiplied by 1-d power
Weight, d be by bin take up space in units of the sampled point that is measured the distance between to bin centers.As shown in Figure 5.4 in figure
Individual black grid represents that LHOPC 4 plane bin, bin centers are respectively b (0,0), b (0,1), b (1,0) and b (1,1), often
Individual plane bin is divided into 8 angle bin again, and red lines represent bin center line, and black origin represents sampled point, dr and
Dc represents that sampled point then represents sampled point to angle bin centers to the line direction and column direction distance at b (0,0) center, do respectively
Distance.According to the principle of tri-linear interpolation, the weight that the sampled point distributes to b (0,0) is (1-dr) (1-dc) (1-
Do), b (0,1) weight is (1-dr) (dc) (1-do), and b (1,0) weight is (dr) (1-dc) (1-do), b
(1,1) weight is (dr) (dc) (1-do).Fig. 5 is LHOPC description vectors interpolation schematic diagrames, and (a) is plane bin, (b)
For angle bin.
4th, in order to further eliminate the influence of illumination variation, the composition that 0.2 is more than in LHOPC characteristic vectors is entered as
0.2, and be normalized using formula below, form final LHOPC feature descriptors.
Wherein, v is characterized vector, and ε is a number avoided except zero
Step 4: according to step 3, calculate refer to image I respectively1Upper characteristic point PointI1i(i=1,2,3 ..., N) and
Image I to be matched2Upper characteristic point PointI2iThe LHOPC characteristic vectors of (i=1,2,3 ..., N).
Step 5: extraction PointI1iOne point PointI of (i=1,2,3 ..., N)1iLHOPC characteristic vectors, meter
Calculate it and PointI2iThe Euclidean distance d for the LHOPC characteristic vectors each put in (i=1,2,3 ..., N)1i(i=1,2,
3 ... .., N), and wherein nearest distance d1f and secondary near distance d1s are obtained, calculate their ratio dratio=d1f/d1s,
If dratio< 0.8, then it is assumed that the point is PointI1iSame place, be designated as PointI2i。
Wherein, v1kFor PointI1iThe LHOPC characteristic vectors of certain point, v in (i=1,2,3 ..., N)1kFor PointI2i
The LHOPC characteristic vectors of the certain point of (i=1,2,3 ..., N)
Step 6: carrying out bi-directional matching, according to the method for step 7, point PointI is found2iSame place PointI '1iIf
Point PointI1i and point PointI ' 1i is same point, then (PointI1i, PointI2i) is considered as into final same place pair.
As shown in Figure 6, (a) is that (b) is image to be matched with reference to image.
According to above step, PointI is traveled through1iEach point of (i=1,2,3 ..., N), obtain corresponding same place pair
PointI2i(i=1,2,3 ..., N).
Step 7: using stochastical sampling consistency algorithm, using projective transformation as geometry constraint conditions, iteratively reject and miss
The larger same place of difference, final correctly same place pair is obtained, is designated as { PointID1i, PointID2i(i=1,2,3 ...,
N), the accurate matching of image is realized.
This method can effectively compensate for traditional matching process such as SIFT, SURF to radiating difference more sensitivity image not
Foot, same place is carried out by using with illumination and contrast consistency phase coincident characteristic structure local feature description symbol
Match somebody with somebody, efficiently solve the matching problem of the multisensor remote sensing image such as visible ray, infrared.
Claims (1)
1. a kind of multisensor Remote Sensing Images Matching Method, comprises the following steps:
Step 1: using the gaussian kernel function of different scale to referring to image I1With image I to be matched2, convolution fortune is carried out respectively
Calculate, establish the Gaussian scale-space of image, and generation difference of Gaussian chi is made the difference by the adjacent two layers image of Gaussian scale-space
Space is spent, then, extremum extracting is carried out in this space, extracts the characteristic point PointI with scale invariabilityi;
Step 2: being extended to phase equalization computation model, phase equalization characteristic direction is built, then calculates Gauss chi
Spend the phase equalization characteristic value and characteristic direction of each pixel in space on each layer image;
Step 3: utilize phase equalization characteristic value and characteristic direction structure local feature description symbol -- local phase uniformity side
To histogram, LHOPC is designated as, specifically includes two parts of principal direction and feature description vectors:
(1) characteristic point PointI is extractediIn a point, neighborhood of the size for 5 × 5 pixels is taken centered on it, calculates neighborhood
Interior phase equalization characteristic value and characteristic direction, phase equalization direction histogram is formed, and select the peak value side of histogram
To the principal direction as characteristic point, make descriptor that there is rotational invariance;
(2) coordinate system is established according to principal direction, and with characteristic point PointIiCentered on take neighborhood of the size for 20 × 20 pixels,
Then the neighborhood is divided into 4 × 4 sub-regions, 8 directions will be divided into every sub-regions, and count the phase in each direction
Bit integrity characteristic value, phase equalization direction histogram is formed, it is finally that the phase equalization direction in every sub-regions is straight
Square figure links together, and forms final feature description vectors-local phase uniformity direction histogram;
Image I is referred to Step 4: calculating respectively1Upper characteristic point PointI1iWith image I to be matched2Upper characteristic point PointI2i's
LHOPC characteristic vectors, and similarity measure is used as by Euclidean distance, using the arest neighbors between characteristic vector and time nearest neighbor distance
The ratio between dratioTo be matched, work as dratioDuring less than or equal to given threshold value, PointI1iAnd PointI2iThen it is considered as matching
Point;
Step 5: in order that matching is more sane, here using two-way matching strategy, i.e., for referring to image I1On it is a certain
Individual characteristic point PointI1i, matched using LHOPC, obtain it in image I to be matched2On same place PointI2i, then
With same matching way, PointI is obtained2iReferring to image I1Same place PointI ' corresponding to upper1i;If PointI1iWith
PointI’1iIt is same point, then it is assumed that PointI1iAnd PointI2iIt is a pair of same places;
Step 6: according to step 4, PointI is traveled through1iEach point, obtain corresponding to same place PointI2i, that is, obtain same
Famous cake pair, it is designated as { PointI1i, PointI2i};
Step 7: using stochastical sampling consistency algorithm, error matching points are rejected using projective transformation as geometrical constraint, are obtained most
Whole correct same place is to { PointID1i, PointID2i, realize the accurate matching of image.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510638280.3A CN105261014B (en) | 2015-09-30 | 2015-09-30 | A kind of multisensor Remote Sensing Images Matching Method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201510638280.3A CN105261014B (en) | 2015-09-30 | 2015-09-30 | A kind of multisensor Remote Sensing Images Matching Method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105261014A CN105261014A (en) | 2016-01-20 |
CN105261014B true CN105261014B (en) | 2018-01-12 |
Family
ID=55100686
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201510638280.3A Active CN105261014B (en) | 2015-09-30 | 2015-09-30 | A kind of multisensor Remote Sensing Images Matching Method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105261014B (en) |
Families Citing this family (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106651927B (en) * | 2016-12-30 | 2020-03-24 | 北京航天泰坦科技股份有限公司 | Homonymous point gross error eliminating method for matching embedded edges of orthophoto images |
CN107025449B (en) * | 2017-04-14 | 2020-04-07 | 西南交通大学 | Oblique image straight line feature matching method constrained by local area with unchanged visual angle |
CN107909018B (en) * | 2017-11-06 | 2019-12-06 | 西南交通大学 | Stable multi-mode remote sensing image matching method and system |
CN108764249B (en) * | 2018-04-23 | 2021-07-09 | 云南民族大学 | Rotation-invariant local feature description method, system and device for multi-source image |
CN110309694B (en) * | 2018-08-09 | 2021-03-26 | 中国人民解放军战略支援部队信息工程大学 | Method and device for determining main direction of remote sensing image |
CN109509216A (en) * | 2018-10-31 | 2019-03-22 | 中国科学院遥感与数字地球研究所 | A kind of Landsat and SPOT satellite image automatic registration method |
CN109523585B (en) * | 2018-11-19 | 2021-10-22 | 武汉大学 | Multisource remote sensing image feature matching method based on direction phase consistency |
CN111126508A (en) * | 2019-12-31 | 2020-05-08 | 湖南云顶智能科技有限公司 | Hopc-based improved heterogeneous image matching method |
CN111985502A (en) * | 2020-08-03 | 2020-11-24 | 武汉大学 | Multi-mode image feature matching method with scale invariance and rotation invariance |
CN112132006B (en) * | 2020-09-21 | 2022-08-26 | 西南交通大学 | Intelligent forest land and building extraction method for cultivated land protection |
CN112284287B (en) * | 2020-09-24 | 2022-02-11 | 哈尔滨工业大学 | Stereoscopic vision three-dimensional displacement measurement method based on structural surface gray scale characteristics |
CN112634335A (en) * | 2020-12-25 | 2021-04-09 | 清华大学 | Method for extracting characteristic point pairs of robust remote sensing image facing to nonlinear radiation distortion |
CN113077437B (en) * | 2021-03-31 | 2023-07-25 | 上海晨兴希姆通电子科技有限公司 | Workpiece quality detection method and system |
CN113191202B (en) * | 2021-04-07 | 2022-07-12 | 烟台艾睿光电科技有限公司 | Image processing method and device and electronic equipment |
CN113256653B (en) * | 2021-05-25 | 2023-05-09 | 南京信息工程大学 | Heterogeneous high-resolution remote sensing image registration method for high-rise ground object |
CN113344105B (en) * | 2021-06-24 | 2022-01-21 | 中国科学院地理科学与资源研究所 | Sampling point distribution method and device, electronic equipment and readable storage medium |
CN114004770B (en) * | 2022-01-04 | 2022-04-26 | 成都国星宇航科技有限公司 | Method and device for accurately correcting satellite space-time diagram and storage medium |
CN115205558B (en) * | 2022-08-16 | 2023-03-24 | 中国测绘科学研究院 | Multi-mode image matching method and device with rotation and scale invariance |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103914847A (en) * | 2014-04-10 | 2014-07-09 | 西安电子科技大学 | SAR image registration method based on phase congruency and SIFT |
CN104732532A (en) * | 2015-03-11 | 2015-06-24 | 中国空间技术研究院 | Remote sensing satellite multispectral image registration method |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP5152231B2 (en) * | 2010-03-12 | 2013-02-27 | オムロン株式会社 | Image processing method and image processing apparatus |
-
2015
- 2015-09-30 CN CN201510638280.3A patent/CN105261014B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103914847A (en) * | 2014-04-10 | 2014-07-09 | 西安电子科技大学 | SAR image registration method based on phase congruency and SIFT |
CN104732532A (en) * | 2015-03-11 | 2015-06-24 | 中国空间技术研究院 | Remote sensing satellite multispectral image registration method |
Non-Patent Citations (2)
Title |
---|
一种结合SIFT和边缘信息的多源遥感影像匹配方法;叶沅鑫 等;《武汉大学学报·信息科学版》;20131031;第38卷(第10期);第1148-1151页,第1260页 * |
相位一致性和区域互信息量的医学图像配准;张娟 等;《计算机工程与应用》;20100930;第46卷(第27期);第162-165页 * |
Also Published As
Publication number | Publication date |
---|---|
CN105261014A (en) | 2016-01-20 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105261014B (en) | A kind of multisensor Remote Sensing Images Matching Method | |
Sui et al. | Automatic optical-to-SAR image registration by iterative line extraction and Voronoi integrated spectral point matching | |
CN103020945B (en) | A kind of remote sensing image registration method of Multiple Source Sensor | |
CN104200461B (en) | The remote sensing image registration method of block and sift features is selected based on mutual information image | |
Chen et al. | Robust affine-invariant line matching for high resolution remote sensing images | |
CN103886611A (en) | Image matching method suitable for automatically detecting flight quality of aerial photography | |
CN102722731A (en) | Efficient image matching method based on improved scale invariant feature transform (SIFT) algorithm | |
CN112396643A (en) | Multi-mode high-resolution image registration method with scale-invariant features and geometric features fused | |
Daixian | SIFT algorithm analysis and optimization | |
CN101650784A (en) | Method for matching images by utilizing structural context characteristics | |
CN104992400A (en) | Method and device for multispectral image registration | |
CN102446356A (en) | Parallel and adaptive matching method for acquiring remote sensing images with homogeneously-distributed matched points | |
Gao et al. | Multi-scale PIIFD for registration of multi-source remote sensing images | |
CN110222661A (en) | It is a kind of for motion estimate and the feature extracting method of tracking | |
Xiong et al. | Robust SAR image registration using rank-based ratio self-similarity | |
CN112614167A (en) | Rock slice image alignment method combining single-polarization and orthogonal-polarization images | |
Xiang et al. | Automatic registration of optical and SAR images via improved phase congruency | |
CN103336964B (en) | SIFT image matching method based on module value difference mirror image invariant property | |
Chen et al. | Hierarchical line segment matching for wide-baseline images via exploiting viewpoint robust local structure and geometric constraints | |
CN110020659A (en) | A kind of extraction of remote sensing image multi-scale edge and matching process and system based on dyadic wavelet | |
Changjie et al. | Algorithm of remote sensing image matching based on corner-point | |
CN104036494A (en) | Fast matching computation method used for fruit picture | |
Zhang et al. | A line matching method based on local and global appearance | |
Li et al. | Reliable and fast mapping of keypoints on large-size remote sensing images by use of multiresolution and global information | |
CN111160433A (en) | High-speed matching method and system for high-resolution image feature points |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |