CN104966283A - Imaging layered registering method - Google Patents
Imaging layered registering method Download PDFInfo
- Publication number
- CN104966283A CN104966283A CN201510266394.XA CN201510266394A CN104966283A CN 104966283 A CN104966283 A CN 104966283A CN 201510266394 A CN201510266394 A CN 201510266394A CN 104966283 A CN104966283 A CN 104966283A
- Authority
- CN
- China
- Prior art keywords
- image
- multi resolution
- resolution image
- point matching
- layer
- 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
Landscapes
- Image Analysis (AREA)
- Apparatus For Radiation Diagnosis (AREA)
Abstract
The invention provides an imaging layered registering method comprising: acquiring a first multilayer resolution image corresponding to a reference image and a second multilayer resolution image corresponding to a test image to be registered; computing an optimal image transformation relation of a first layer of image by using a RANSAC (random sample consensus) algorithm; removing erroneous characteristic point matched pairs in the initial characteristic point matched pairs of the second layer of image in the first multilayer resolution image and the second multilayer resolution image by using the optimal image transformation relation of the first layer image, and computing an optimal image transformation relation of a second layer of image by using the RANSAC algorithm; and performing operation in the same way until the last layer of image of the first multilayer resolution image and the last layer of image of the second multilayer resolution image. The imaging layering registering method improves the realtimeness of image registration technology while guaranteeing the accuracy of the image registration technology.
Description
Technical field
The present invention relates to image processing field, particularly relate to a kind of image layered method for registering.
Background technology
Image registration is the process that two width that (weather, illumination, camera position and angle etc.) under different time, different sensors (imaging device) or different condition obtained or multiple image carry out mating, superposing, and image registration techniques is widely used in the fields such as remotely-sensed data analysis, computer vision, image procossing.Especially in recent years, image registration techniques is applied to the fields such as image mosaic, target identification and safety monitoring more and more widely, is proposed higher requirement to the accuracy of image registration and real-time.
Current method for registering images flow process is as follows: first, carrying out feature extraction and obtains unique point, finding the feature point pairs of coupling by carrying out similarity measurement to two width images; Then, image space coordinate conversion parameter is obtained by the feature point pairs of coupling; Finally, image registration is carried out by coordinate conversion parameter.In actual moving process, in order to accurate extract minutiae and matching characteristic point pair, the computation complexity of method for registering images is higher, and calculated amount is comparatively large, and therefore, the computing time of needs is very long.And calculated amount is also relevant to picture size and resolution, and picture size is larger, resolution is higher, and calculated amount is larger.
Because current method for registering images has paid larger time cost while guarantee accuracy, cause the real-time of traditional images registration technology poor, hindered the application of image registration techniques in a lot of field.
Summary of the invention
The invention provides a kind of image layered method for registering, while guarantee image registration techniques accuracy, improve the real-time of image registration techniques.
Image layered method for registering provided by the invention, comprise: obtain the first Multi resolution image corresponding to reference picture respectively, the second Multi resolution image corresponding with test pattern subject to registration, described first Multi resolution image and described second Multi resolution image include the image that multiple resolution increases gradually;
Obtain unique point and the descriptor of each tomographic image in described first Multi resolution image successively, and obtain unique point and the descriptor of each tomographic image in described second Multi resolution image successively;
Carry out feature point pairs according to described descriptor to the described unique point of respective layer in described first Multi resolution image and described second Multi resolution image slightly to mate, obtain the initial characteristics Point matching pair of respective layer;
Adopt random sampling consistent RANSAC algorithm to the described initial characteristics Point matching of ground floor image in described first Multi resolution image to the described initial characteristics Point matching with ground floor image in described second Multi resolution image to screening, obtain the optimal characteristics Point matching pair of ground floor image between described first Multi resolution image and described second Multi resolution image; According to the optimal characteristics Point matching of described ground floor image to the optimum image transformation relation calculating ground floor image;
The optimum image transformation relation of described ground floor image is adopted to reject the error characteristic Point matching pair of the described initial characteristics Point matching centering of second layer image in described first Multi resolution image, obtain the residue character Point matching pair of second layer image in described first Multi resolution image, and adopt the optimum image transformation relation of described ground floor image to reject the error characteristic Point matching pair of the described initial characteristics Point matching centering of second layer image in described second Multi resolution image, obtain the residue character Point matching pair of second layer image in described second Multi resolution image, adopt RANSAC algorithm to screen with the described residue character Point matching of second layer image in described second Multi resolution image the described residue character Point matching of second layer image in described first Multi resolution image, obtain the optimal characteristics Point matching pair of second layer image between described first Multi resolution image and described second Multi resolution image, according to the optimal characteristics Point matching of described second layer image to the optimum image transformation relation calculating second layer image, by that analogy, until last tomographic image of described first Multi resolution image and described second Multi resolution image.
Optionally, the first Multi resolution image that the described reference picture of acquisition is respectively corresponding, and the second Multi resolution image that test pattern subject to registration is corresponding, comprising:
According to default sampling rate, multiple first subimage of down-sampled acquisition is carried out to described reference picture, described multiple first subimage and the described first Multi resolution image of described reference picture composition; With
According to described default sampling rate, multiple second subimage of down-sampled acquisition is carried out to described test pattern, described multiple second subimage and the described second Multi resolution image of described test pattern composition.
Optionally, described unique point and the descriptor obtaining each tomographic image in described first Multi resolution image successively, and obtain unique point and the descriptor of each tomographic image in described second Multi resolution image successively, comprising:
Adopt scale invariant feature conversion SIFT algorithm to obtain unique point and the descriptor of each tomographic image in described first Multi resolution image successively, and obtain unique point and the descriptor of each tomographic image in described second Multi resolution image successively.
Optionally, described unique point and the descriptor obtaining each tomographic image in described first Multi resolution image successively, and obtain unique point and the descriptor of each tomographic image in described second Multi resolution image successively, comprising:
Adopt and accelerate unique point and the descriptor that robust feature SURF algorithm obtains each tomographic image in described first Multi resolution image successively, and obtain unique point and the descriptor of each tomographic image in described second Multi resolution image successively.
Optionally, the optimum image transformation relation of described employing described ground floor image rejects the error characteristic Point matching pair of the described initial characteristics Point matching centering of second layer image in described first Multi resolution image, obtain the residue character Point matching pair of second layer image in described first Multi resolution image, and adopt the optimum image transformation relation of described ground floor image to reject the error characteristic Point matching pair of the described initial characteristics Point matching centering of second layer image in described second Multi resolution image, obtain the residue character Point matching pair of second layer image in described second Multi resolution image, adopt RANSAC algorithm to screen with the described residue character Point matching of second layer image in described second Multi resolution image the described residue character Point matching of second layer image in described first Multi resolution image, obtain the optimal characteristics Point matching pair of second layer image between described first Multi resolution image and described second Multi resolution image, according to the optimal characteristics Point matching of described second layer image to the optimum image transformation relation calculating second layer image, by that analogy, until after last tomographic image of described first Multi resolution image and described second Multi resolution image, also comprise:
According to the described optimum image transformation relation of last tomographic image of described first Multi resolution image and described second Multi resolution image, image mosaic and image co-registration are carried out to described reference picture and described test pattern, obtain panoramic picture.
Optionally, described reference picture is the frame reference picture in reference video, and described test pattern subject to registration is the frame test pattern in video to be matched, and the frame of described reference picture is identical with the frame of described test pattern.
Optionally, the described optimum image transformation relation of described last tomographic image according to described first Multi resolution image and described second Multi resolution image, image mosaic and image co-registration are carried out to described reference picture and described test pattern, after obtaining panoramic picture, also comprise:
By panoramic picture generating video stream described in each frame of obtaining.
The invention provides a kind of image layered method for registering, by obtaining the first Multi resolution image corresponding to reference picture, the second Multi resolution image corresponding with test pattern subject to registration, image pyramid structure is generated to unique point and the descriptor of every tomographic image from top layer to bottom, the initial characteristics Point matching pair of every tomographic image is generated according to descriptor, from the second layer image of image pyramid structure, the error characteristic Point matching pair of this tomographic image is rejected by the optimum image transformation relation of last layer image, RANSAC algorithm is adopted to obtain the optimal characteristics Point matching pair of this tomographic image, calculate the optimum image transformation relation obtaining this tomographic image, owing to successively rejecting the Feature Points Matching pair of mistake, the calculated amount of algorithm is greatly reduced by the mode of iteration, while guarantee image registration techniques accuracy, shorten computing time, further increase the real-time of image registration techniques.
Accompanying drawing explanation
In order to be illustrated more clearly in the embodiment of the present invention or technical scheme of the prior art, be briefly described to the accompanying drawing used required in embodiment or description of the prior art below, apparently, accompanying drawing in the following describes is some embodiments of the present invention, for those of ordinary skill in the art, under the prerequisite not paying creative work, other accompanying drawing can also be obtained according to these accompanying drawings.
The process flow diagram of the image layered method for registering that Fig. 1 provides for the embodiment of the present invention one;
The process flow diagram of the image layered method for registering that Fig. 2 provides for the embodiment of the present invention two.
Embodiment
For making the object of the embodiment of the present invention, technical scheme and advantage clearly, below in conjunction with the accompanying drawing in the embodiment of the present invention, technical scheme in the embodiment of the present invention is clearly and completely described, obviously, described embodiment is the present invention's part embodiment, instead of whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art, not making the every other embodiment obtained under creative work prerequisite, belong to the scope of protection of the invention.
The process flow diagram of the image layered method for registering that Fig. 1 provides for the embodiment of the present invention one.As shown in Figure 1, the image layered method for registering that the present embodiment provides, can comprise:
Step 101, the first Multi resolution image that acquisition reference picture is corresponding respectively, the second Multi resolution image corresponding with test pattern subject to registration, the first Multi resolution image and the second Multi resolution image include the image that multiple resolution increases gradually.
Wherein, first Multi resolution image and the second Multi resolution image are image pyramid structure, so-called image pyramid structure, refer to the image collection progressively increased with the resolution that Pyramid arranges with the subimage of same image under different resolution, pyramidal bottom is the high resolving power subimage of image, and pyramidal top is the low resolution subimage of image.
Optionally, obtain the first Multi resolution image corresponding to reference picture respectively, and the second Multi resolution image that test pattern subject to registration is corresponding, a kind of implementation can be:
Carry out multiple first subimage of down-sampled acquisition according to default sampling rate to reference picture, multiple first subimage and reference picture form the first Multi resolution image; With
Carry out multiple second subimage of down-sampled acquisition according to default sampling rate to test pattern, multiple second subimage and test pattern form the second Multi resolution image.
Wherein, last tomographic image of the first Multi resolution image is the reference picture of original resolution, and last tomographic image of the second Multi resolution image is the test pattern of original resolution.Preset sampling rate to arrange according to actual needs, the present embodiment does not limit concrete numerical value.
Step 103, the unique point obtaining each tomographic image in the first Multi resolution image successively and descriptor, and the unique point and the descriptor that obtain each tomographic image in the second Multi resolution image successively.
So-called unique point, refers to not labile pixel or the regional area with certain pattern feature in image, also referred to as point of interest, significant point or key point.Unique point is the description that picture material is the most abstract, has good adaptive faculty to picture noise, grey scale change, image deformation and blocking etc.
So-called descriptor, refers to the feature interpretation set up unique point, also referred to as descriptor.Desirable descriptor will meet and has certain unchangeability to yardstick, rotation, the conversion such as even affine, and descriptor correlativity corresponding to different characteristic point is little, could effectively distinguish different unique points like this.
In this step, for each tomographic image in the first Multi resolution image and the second Multi resolution image, all need the unique point and the descriptor that obtain image.The size of image is less, resolution is lower, then the unique point of image is fewer, and that is, the unique point of the first Multi resolution image and the second Multi resolution image every tomographic image from top to bottom gets more and more.
The unique point and the descriptor that obtain image can have various ways, and the present embodiment is not limited.
Optionally, obtain unique point and the descriptor of each tomographic image in the first Multi resolution image successively, and obtain unique point and the descriptor of each tomographic image in the second Multi resolution image successively, can comprise:
Adopt scale invariant feature conversion (Scale-invariant feature transform, being called for short SIFT) algorithm obtains unique point and the descriptor of each tomographic image in the first Multi resolution image successively, and obtains unique point and the descriptor of each tomographic image in the second Multi resolution image successively.
SIFT algorithm is a kind of algorithm detecting local feature, and this algorithm is by the unique point of synthetic image and descriptor thereof thus obtain feature and carry out Image Feature Point Matching.In the algorithm, SIFT feature point is the unique point based on local appearance, and the tolerance changed scaling, rotation, brightness change, noise, visual angle is quite high, unique point highly significant and easily obtaining; SIFT descriptor is a kind of local feature description symbol, has scale invariability.But, it is relevant to the size of image and resolution that SIFT algorithm carries out the calculated amount of unique point exact matching, and the size of image is larger, the resolution of image is higher, then the unique point of image is more, the calculated amount that SIFT algorithm carries out unique point exact matching is larger, even increases by power.
Optionally, obtain unique point and the descriptor of each tomographic image in the first Multi resolution image successively, and obtain unique point and the descriptor of each tomographic image in the second Multi resolution image successively, can comprise:
Adopt and accelerate robust feature (Speeded Up Robust Features, being called for short SURF) algorithm obtains unique point and the descriptor of each tomographic image in the first Multi resolution image successively, and obtains unique point and the descriptor of each tomographic image in the second Multi resolution image successively.
SURF algorithm is a kind of sane image invariant features detection algorithm, by SIFT algorithm improvement.
Step 105, according to descriptor, feature point pairs is carried out to the unique point of respective layer in the first Multi resolution image and the second Multi resolution image and slightly mate, obtain the initial characteristics Point matching pair of respective layer.
In this step, need to carry out feature point pairs according to descriptor to the unique point of each identical layer image in the first Multi resolution image and the second Multi resolution image slightly to mate and obtain initial characteristics Point matching pair, initial characteristics Point matching is to the corresponding relation that may be one-to-many, for example, in first Multi resolution image, a unique point of ground floor image all can be mated for initial characteristics Point matching pair with four unique points of ground floor image in the second Multi resolution image, so, initial characteristics Point matching to and out of true.
Concrete, when the feature point pairs carrying out ground floor image in the first Multi resolution image and the second Multi resolution image slightly mates, carry out feature point pairs according to the descriptor of ground floor image in the descriptor of ground floor image in the first Multi resolution image and the second Multi resolution image slightly to mate, obtain the initial characteristics Point matching pair of ground floor image in the first Multi resolution image, and the initial characteristics Point matching pair of ground floor image in the second Multi resolution image.By that analogy, until obtain the initial characteristics Point matching pair of last tomographic image in the first Multi resolution image, and the initial characteristics Point matching pair of last tomographic image in the second Multi resolution image.
In this step, slightly mate owing to only carrying out feature point pairs to the unique point of respective layer in the first Multi resolution image and the second Multi resolution image, do not need exact matching, so greatly reduce the calculated amount of algorithm, shorten computing time.
Step 107, consistent (the Random Sample Consensus of employing random sampling, be called for short RANSAC) algorithm to the initial characteristics Point matching of ground floor image in the first Multi resolution image to the initial characteristics Point matching with ground floor image in the second Multi resolution image to screening, obtain the optimal characteristics Point matching pair of ground floor image between the first Multi resolution image and the second Multi resolution image; According to the optimal characteristics Point matching of ground floor image to the optimum image transformation relation calculating ground floor image.
RANSAC algorithm is the sample data collection comprising abnormal data according to a group, calculates the mathematical model parameter of data, obtains the algorithm of effective sample data.The calculated amount of RANSAC algorithm is relevant to the number of samples that sample data is concentrated, and number of samples is larger, and the calculated amount of RANSAC algorithm is larger.
Because the initial characteristics Point matching obtained in step 105 is to the corresponding relation that may there is one-to-many, and out of true, so, in this step, obtained the optimal characteristics Point matching pair of ground floor image between the first Multi resolution image and the second Multi resolution image by RANSAC algorithm.Optimal characteristics Point matching is to being man-to-man corresponding relation, for example, in first Multi resolution image a unique point of ground floor image only with a Feature Points Matching of ground floor image in the second Multi resolution image, so optimal characteristics Point matching is to being accurate Feature Points Matching result.Then according to the optimal characteristics Point matching of ground floor image to the optimum image transformation relation calculating ground floor image.
The optimum image transformation relation of step 109, employing ground floor image rejects the error characteristic Point matching pair of the initial characteristics Point matching centering of second layer image in the first Multi resolution image, obtain the residue character Point matching pair of second layer image in the first Multi resolution image, and adopt the optimum image transformation relation of ground floor image to reject the error characteristic Point matching pair of the initial characteristics Point matching centering of second layer image in the second Multi resolution image, obtain the residue character Point matching pair of second layer image in the second Multi resolution image; Adopt RANSAC algorithm to screen with the residue character Point matching of second layer image in the second Multi resolution image the residue character Point matching of second layer image in the first Multi resolution image, obtain the optimal characteristics Point matching pair of second layer image between the first Multi resolution image and the second Multi resolution image; According to the optimal characteristics Point matching of second layer image to the optimum image transformation relation calculating second layer image; By that analogy, until last tomographic image of the first Multi resolution image and the second Multi resolution image.
In this step, from the second layer image in the first Multi resolution image and the second Multi resolution image, first the optimum image transformation relation of last layer image is adopted to reject the error characteristic Point matching pair of the initial characteristics Point matching centering of this tomographic image, after such process, the total sample number that the residue character Point matching of this tomographic image obtained is right reduces, and, correct sample number ratio increases, error sample number ratio reduces, that is, optimal characteristics Point matching is obtained to before at employing RANSAC algorithm, greatly reduce the number that the invalid initial characteristics Point matching of this tomographic image is right, calculative Feature Points Matching is right in RANSAC algorithm number of samples is greatly reduced and sample more accurate, thus decrease the calculated amount of RANSAC algorithm, shorten computing time.That is, greatly reduced the calculated amount of algorithm by the mode of iteration, shorten computing time, while guarantee image registration techniques accuracy, improve the real-time of image registration techniques.
Optionally, after step 109, can also comprise:
Step 111, optimum image transformation relation according to last tomographic image of the first Multi resolution image and the second Multi resolution image, carry out image mosaic and image co-registration to reference picture and test pattern, obtain panoramic picture.
Optionally, reference picture is the frame reference picture in reference video, and test pattern subject to registration is the frame test pattern in video to be matched, and the frame of reference picture is identical with the frame of test pattern.
Present embodiments provide a kind of image layered method for registering, by obtaining the first Multi resolution image corresponding to reference picture, the second Multi resolution image corresponding with test pattern subject to registration, image pyramid structure is generated to unique point and the descriptor of every tomographic image from top layer to bottom, the initial characteristics Point matching pair of every tomographic image is generated according to descriptor, from the second layer image of image pyramid structure, the error characteristic Point matching pair of this tomographic image is rejected by the optimum image transformation relation of last layer image, RANSAC algorithm is adopted to obtain the optimal characteristics Point matching pair of this tomographic image, calculate the optimum image transformation relation obtaining this tomographic image, owing to successively rejecting the Feature Points Matching pair of mistake, the calculated amount of algorithm is greatly reduced by the mode of iteration, while guarantee image registration techniques accuracy, shorten computing time, further increase the real-time of image registration techniques.
The process flow diagram of the image layered method for registering that Fig. 2 provides for the embodiment of the present invention two, the present embodiment, on the basis of embodiment one, provides a kind of application scenarios of image layered method for registering, is namely applied to the real-time splicing of video flowing.As shown in Figure 2, the image layered method for registering that the present embodiment provides, can comprise:
Step 201, the two field picture captured in reference video are reference picture, and the two field picture captured in video to be matched is test pattern subject to registration, and the frame of reference picture is identical with the frame of test pattern.
Step 203, the first Multi resolution image that acquisition reference picture is corresponding respectively, the second Multi resolution image corresponding with test pattern subject to registration, the first Multi resolution image and the second Multi resolution image include the image that multiple resolution increases gradually.
Step 205, the unique point obtaining each tomographic image in the first Multi resolution image successively and descriptor, and the unique point and the descriptor that obtain each tomographic image in the second Multi resolution image successively.
Step 207, according to descriptor, feature point pairs is carried out to the unique point of respective layer in the first Multi resolution image and the second Multi resolution image and slightly mate, obtain the initial characteristics Point matching pair of respective layer.
Step 209, adopt RANSAC algorithm to the initial characteristics Point matching of ground floor image in the first Multi resolution image to the initial characteristics Point matching with ground floor image in the second Multi resolution image to screening, obtain the optimal characteristics Point matching pair of ground floor image between the first Multi resolution image and the second Multi resolution image; According to the optimal characteristics Point matching of ground floor image to the optimum image transformation relation calculating ground floor image.
The optimum image transformation relation of step 211, employing ground floor image rejects the error characteristic Point matching pair of the initial characteristics Point matching centering of second layer image in the first Multi resolution image, obtain the residue character Point matching pair of second layer image in the first Multi resolution image, and adopt the optimum image transformation relation of ground floor image to reject the error characteristic Point matching pair of the initial characteristics Point matching centering of second layer image in the second Multi resolution image, obtain the residue character Point matching pair of second layer image in the second Multi resolution image; Adopt RANSAC algorithm to screen with the residue character Point matching of second layer image in the second Multi resolution image the residue character Point matching of second layer image in the first Multi resolution image, obtain the optimal characteristics Point matching pair of second layer image between the first Multi resolution image and the second Multi resolution image; According to the optimal characteristics Point matching of second layer image to the optimum image transformation relation calculating second layer image; By that analogy, until last tomographic image of the first Multi resolution image and the second Multi resolution image.
Step 213, optimum image transformation relation according to last tomographic image of the first Multi resolution image and the second Multi resolution image, carry out image mosaic and image co-registration to reference picture and test pattern, obtain panoramic picture.
Step 215, each frame panoramic picture generating video stream that will obtain.
Present embodiments provide a kind of image layered method for registering, be applied in the real-time splicing scene of video flowing, the two field picture captured in reference video is reference picture, the two field picture captured in video to be matched is test pattern subject to registration, by obtaining the first Multi resolution image corresponding to reference picture, the second Multi resolution image corresponding with test pattern subject to registration, image pyramid structure is generated to unique point and the descriptor of every tomographic image from top layer to bottom, the initial characteristics Point matching pair of every tomographic image is generated according to descriptor, from the second layer image of image pyramid structure, the error characteristic Point matching pair of this tomographic image is rejected by the optimum image transformation relation of last layer image, RANSAC algorithm is adopted to obtain the optimal characteristics Point matching pair of this tomographic image, calculate the optimum image transformation relation obtaining this tomographic image, reference picture after registration and test pattern are carried out image mosaic and image co-registration, obtain panoramic picture, by each frame panoramic picture generating video stream obtained.Owing to successively rejecting the Feature Points Matching pair of mistake, greatly reduced the calculated amount of algorithm by the mode of iteration, while guarantee image registration techniques accuracy, shorten computing time, further increase the real-time of image registration techniques.
Last it is noted that above each embodiment is only in order to illustrate technical scheme of the present invention, be not intended to limit; Although with reference to foregoing embodiments to invention has been detailed description, those of ordinary skill in the art is to be understood that: it still can be modified to the technical scheme described in foregoing embodiments, or carries out equivalent replacement to wherein some or all of technical characteristic; And these amendments or replacement, do not make the essence of appropriate technical solution depart from the scope of various embodiments of the present invention technical scheme.
Claims (7)
1. an image layered method for registering, is characterized in that, comprising:
Obtain the first Multi resolution image corresponding to reference picture respectively, the second Multi resolution image corresponding with test pattern subject to registration, described first Multi resolution image and described second Multi resolution image include the image that multiple resolution increases gradually;
Obtain unique point and the descriptor of each tomographic image in described first Multi resolution image successively, and obtain unique point and the descriptor of each tomographic image in described second Multi resolution image successively;
Carry out feature point pairs according to described descriptor to the described unique point of respective layer in described first Multi resolution image and described second Multi resolution image slightly to mate, obtain the initial characteristics Point matching pair of respective layer;
Adopt random sampling consistent RANSAC algorithm to the described initial characteristics Point matching of ground floor image in described first Multi resolution image to the described initial characteristics Point matching with ground floor image in described second Multi resolution image to screening, obtain the optimal characteristics Point matching pair of ground floor image between described first Multi resolution image and described second Multi resolution image; According to the optimal characteristics Point matching of described ground floor image to the optimum image transformation relation calculating ground floor image;
The optimum image transformation relation of described ground floor image is adopted to reject the error characteristic Point matching pair of the described initial characteristics Point matching centering of second layer image in described first Multi resolution image, obtain the residue character Point matching pair of second layer image in described first Multi resolution image, and adopt the optimum image transformation relation of described ground floor image to reject the error characteristic Point matching pair of the described initial characteristics Point matching centering of second layer image in described second Multi resolution image, obtain the residue character Point matching pair of second layer image in described second Multi resolution image, adopt RANSAC algorithm to screen with the described residue character Point matching of second layer image in described second Multi resolution image the described residue character Point matching of second layer image in described first Multi resolution image, obtain the optimal characteristics Point matching pair of second layer image between described first Multi resolution image and described second Multi resolution image, according to the optimal characteristics Point matching of described second layer image to the optimum image transformation relation calculating second layer image, by that analogy, until last tomographic image of described first Multi resolution image and described second Multi resolution image.
2. method according to claim 1, is characterized in that, the first Multi resolution image that the described reference picture of acquisition is respectively corresponding, and the second Multi resolution image that test pattern subject to registration is corresponding, comprising:
According to default sampling rate, multiple first subimage of down-sampled acquisition is carried out to described reference picture, described multiple first subimage and the described first Multi resolution image of described reference picture composition; With
According to described default sampling rate, multiple second subimage of down-sampled acquisition is carried out to described test pattern, described multiple second subimage and the described second Multi resolution image of described test pattern composition.
3. method according to claim 1, it is characterized in that, described unique point and the descriptor obtaining each tomographic image in described first Multi resolution image successively, and obtain unique point and the descriptor of each tomographic image in described second Multi resolution image successively, comprising:
Adopt scale invariant feature conversion SIFT algorithm to obtain unique point and the descriptor of each tomographic image in described first Multi resolution image successively, and obtain unique point and the descriptor of each tomographic image in described second Multi resolution image successively.
4. method according to claim 1, it is characterized in that, described unique point and the descriptor obtaining each tomographic image in described first Multi resolution image successively, and obtain unique point and the descriptor of each tomographic image in described second Multi resolution image successively, comprising:
Adopt and accelerate unique point and the descriptor that robust feature SURF algorithm obtains each tomographic image in described first Multi resolution image successively, and obtain unique point and the descriptor of each tomographic image in described second Multi resolution image successively.
5. the method according to any one of claim 1-4, it is characterized in that, the optimum image transformation relation of described employing described ground floor image rejects the error characteristic Point matching pair of the described initial characteristics Point matching centering of second layer image in described first Multi resolution image, obtain the residue character Point matching pair of second layer image in described first Multi resolution image, and adopt the optimum image transformation relation of described ground floor image to reject the error characteristic Point matching pair of the described initial characteristics Point matching centering of second layer image in described second Multi resolution image, obtain the residue character Point matching pair of second layer image in described second Multi resolution image, adopt RANSAC algorithm to screen with the described residue character Point matching of second layer image in described second Multi resolution image the described residue character Point matching of second layer image in described first Multi resolution image, obtain the optimal characteristics Point matching pair of second layer image between described first Multi resolution image and described second Multi resolution image, according to the optimal characteristics Point matching of described second layer image to the optimum image transformation relation calculating second layer image, by that analogy, until after last tomographic image of described first Multi resolution image and described second Multi resolution image, also comprise:
According to the described optimum image transformation relation of last tomographic image of described first Multi resolution image and described second Multi resolution image, image mosaic and image co-registration are carried out to described reference picture and described test pattern, obtain panoramic picture.
6. method according to claim 5, it is characterized in that, described reference picture is the frame reference picture in reference video, and described test pattern subject to registration is the frame test pattern in video to be matched, and the frame of described reference picture is identical with the frame of described test pattern.
7. method according to claim 6, it is characterized in that, the described optimum image transformation relation of described last tomographic image according to described first Multi resolution image and described second Multi resolution image, image mosaic and image co-registration are carried out to described reference picture and described test pattern, after obtaining panoramic picture, also comprise:
By panoramic picture generating video stream described in each frame of obtaining.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510266394.XA CN104966283A (en) | 2015-05-22 | 2015-05-22 | Imaging layered registering method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510266394.XA CN104966283A (en) | 2015-05-22 | 2015-05-22 | Imaging layered registering method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN104966283A true CN104966283A (en) | 2015-10-07 |
Family
ID=54220314
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510266394.XA Pending CN104966283A (en) | 2015-05-22 | 2015-05-22 | Imaging layered registering method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104966283A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105872496A (en) * | 2016-07-01 | 2016-08-17 | 黄岩 | Ultrahigh-definition video fusion method |
| CN106657789A (en) * | 2016-12-29 | 2017-05-10 | 核动力运行研究所 | Thread panoramic image synthesis method |
| CN108510533A (en) * | 2018-04-02 | 2018-09-07 | 北京理工大学 | Fourier plum forests registration based on FPGA and Laplce's blending image acceleration system |
| CN109313708A (en) * | 2017-12-22 | 2019-02-05 | 深圳配天智能技术研究院有限公司 | Image matching method and vision system |
| CN110443835A (en) * | 2019-07-16 | 2019-11-12 | 北京迈格威科技有限公司 | Method for registering images, device, equipment and storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6359617B1 (en) * | 1998-09-25 | 2002-03-19 | Apple Computer, Inc. | Blending arbitrary overlaying images into panoramas |
| CN102436652A (en) * | 2011-08-31 | 2012-05-02 | 航天恒星科技有限公司 | Automatic registering method of multisource remote sensing images |
| CN102800098A (en) * | 2012-07-19 | 2012-11-28 | 中国科学院自动化研究所 | Multi-characteristic multi-level visible light full-color and multi-spectrum high-precision registering method |
| CN103077527A (en) * | 2013-02-05 | 2013-05-01 | 湖北工业大学 | Robust multi-source satellite remote sensing image registration method |
| US20140267833A1 (en) * | 2013-03-12 | 2014-09-18 | Futurewei Technologies, Inc. | Image registration and focus stacking on mobile platforms |
| CN104463786A (en) * | 2014-12-03 | 2015-03-25 | 中国科学院自动化研究所 | Mobile robot figure stitching method and device |
-
2015
- 2015-05-22 CN CN201510266394.XA patent/CN104966283A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6359617B1 (en) * | 1998-09-25 | 2002-03-19 | Apple Computer, Inc. | Blending arbitrary overlaying images into panoramas |
| CN102436652A (en) * | 2011-08-31 | 2012-05-02 | 航天恒星科技有限公司 | Automatic registering method of multisource remote sensing images |
| CN102800098A (en) * | 2012-07-19 | 2012-11-28 | 中国科学院自动化研究所 | Multi-characteristic multi-level visible light full-color and multi-spectrum high-precision registering method |
| CN103077527A (en) * | 2013-02-05 | 2013-05-01 | 湖北工业大学 | Robust multi-source satellite remote sensing image registration method |
| US20140267833A1 (en) * | 2013-03-12 | 2014-09-18 | Futurewei Technologies, Inc. | Image registration and focus stacking on mobile platforms |
| CN104463786A (en) * | 2014-12-03 | 2015-03-25 | 中国科学院自动化研究所 | Mobile robot figure stitching method and device |
Non-Patent Citations (3)
| Title |
|---|
| WEI QIAO 等: "Speed-up keypoint mapping technique by multi-resolution and global information", 《PROC. SPIE 9410, VISUAL INFORMATION PROCESSING AND COMMUNICATION VI》 * |
| 周军太 等: "一种改进SURF算法的图像配准", 《湖南工业大学学报》 * |
| 谷宗运 等: "基于SURF和改进的RANSAC算法的医学图像配准", 《中国医学影像学杂志》 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105872496A (en) * | 2016-07-01 | 2016-08-17 | 黄岩 | Ultrahigh-definition video fusion method |
| CN105872496B (en) * | 2016-07-01 | 2019-07-09 | 黄岩 | A kind of ultra high-definition video fusion method |
| CN106657789A (en) * | 2016-12-29 | 2017-05-10 | 核动力运行研究所 | Thread panoramic image synthesis method |
| CN109313708A (en) * | 2017-12-22 | 2019-02-05 | 深圳配天智能技术研究院有限公司 | Image matching method and vision system |
| WO2019119453A1 (en) * | 2017-12-22 | 2019-06-27 | 深圳配天智能技术研究院有限公司 | Image matching method and visual system |
| CN108510533A (en) * | 2018-04-02 | 2018-09-07 | 北京理工大学 | Fourier plum forests registration based on FPGA and Laplce's blending image acceleration system |
| CN108510533B (en) * | 2018-04-02 | 2021-08-17 | 北京理工大学 | Fourier mellin registration and Laplace fusion image acceleration system based on FPGA |
| CN110443835A (en) * | 2019-07-16 | 2019-11-12 | 北京迈格威科技有限公司 | Method for registering images, device, equipment and storage medium |
| CN110443835B (en) * | 2019-07-16 | 2022-07-12 | 北京迈格威科技有限公司 | Image registration method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhang et al. | GAN-FM: Infrared and visible image fusion using GAN with full-scale skip connection and dual Markovian discriminators | |
| US11244197B2 (en) | Fast and robust multimodal remote sensing image matching method and system | |
| CN108921176B (en) | Pointer instrument positioning and identifying method based on machine vision | |
| CN107993258B (en) | Image registration method and device | |
| JP5699788B2 (en) | Screen area detection method and system | |
| CN106709950B (en) | Binocular vision-based inspection robot obstacle crossing wire positioning method | |
| US8699786B2 (en) | 3D model generating apparatus, method and CRM by line pattern imaging | |
| US20150262346A1 (en) | Image processing apparatus, image processing method, and image processing program | |
| CN110033431B (en) | Non-contact detection device and detection method for detecting corrosion area on surface of steel bridge | |
| CN108229475B (en) | Vehicle tracking method, system, computer device and readable storage medium | |
| CN104751465A (en) | ORB (oriented brief) image feature registration method based on LK (Lucas-Kanade) optical flow constraint | |
| CN110516517B (en) | Target identification method, device and equipment based on multi-frame image | |
| CN102722731A (en) | Efficient image matching method based on improved scale invariant feature transform (SIFT) algorithm | |
| CN104966283A (en) | Imaging layered registering method | |
| CN105718931B (en) | System and method for determining clutter in acquired images | |
| CN103902953B (en) | A kind of screen detecting system and method | |
| CN107092905B (en) | Method for positioning instrument to be identified of power inspection robot | |
| CN111563896A (en) | Image processing method for catenary anomaly detection | |
| CN111462198B (en) | Multi-mode image registration method with scale, rotation and radiation invariance | |
| CN105654479A (en) | Multispectral image registering method and multispectral image registering device | |
| CN110675442A (en) | Local stereo matching method and system combined with target identification technology | |
| CN106780613B (en) | Visualization method and system for multi-camera calibration result | |
| CN103337080A (en) | Registration technology of infrared image and visible image based on Hausdorff distance in gradient direction | |
| CN113095385A (en) | Multimode image matching method based on global and local feature description | |
| CN107886530A (en) | A kind of improved image registration algorithm based on SIFT feature |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20151007 |
|
| RJ01 | Rejection of invention patent application after publication |