CN102521826A - Image registration device and method - Google Patents
Image registration device and method Download PDFInfo
- Publication number
- CN102521826A CN102521826A CN2011103720320A CN201110372032A CN102521826A CN 102521826 A CN102521826 A CN 102521826A CN 2011103720320 A CN2011103720320 A CN 2011103720320A CN 201110372032 A CN201110372032 A CN 201110372032A CN 102521826 A CN102521826 A CN 102521826A
- Authority
- CN
- China
- Prior art keywords
- data
- template
- multiply accumulating
- registering images
- accumulating device
- 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.)
- Granted
Links
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses an image registration device and method and aims to provide an image registration device with simple structure and high calculating speed and an image registration method. The image registration device comprises a template storage device, a real-time storage device, a data dispatching component, a plurality of multiply accumulators and a comparer, wherein the template storage device is used for storing template data for a proportion operation; the real-time storage device is used for storing data of image to be registered for the proportion operation; the data dispatching component is used for selecting different conversion parameters according to different functions of the multiply accumulators and transmitting conversion parameters and a template data window extracted from the template storage device to corresponding multiply accumulators; each multiply accumulator regulates the template data window according to the conversion parameter and executes convolution algorithm on the data window of the image to be registered and the regulated template data window, wherein the data window of the image to be registered is obtained from the real-time storage device; and the comparer is used for comparing operation results of the multiply accumulators to find out the maximum. The image registration device and method disclosed by the invention are suitable for all occasions needing image registration.
Description
Technical field
The present invention relates to the image recognition processing field, especially relate to a kind of image registration device and correlation technique.
Background technology
Process of image registration is meant the process that two width of cloth to the Same Scene of taking from different time, different sensors or different visual angles mate with epigraph.The geometric parameter that space changes between each two field picture of video to obtain technical applications very extensive, such as remote Sensing Image Analysis, Panoramic Warping, Target Recognition and location etc.This technology becomes image registration.
Images match is an of paramount importance step in the image registration, and the quality of coupling will directly influence final registration accuracy and working time.Generally original graph is called template, the figure that treats registration is called real-time figure.
Realtime graphic is confirmed generally that in the position in space they are with respect to template by Several Parameters in scene registration field: directions X translation parameters, Y direction translation parameters, rotation parameter, tilt parameters, zooming parameter.In the actual registration process,, often to do the exhaustive of various parameters in order to find optimized parameter.Although there are many methods can reduce calculated amount, on principle the method for any reduction calculated amount cost is all arranged.
Generally do when registration calculates all be with a sub-graphs as calculating object, be called window.Typical method for registering is to judge the registration degree through calculation template window and the convolution of scheming window in real time.After having confirmed one group of supposition parameter, do the convolutional calculation of real-time figure and template corresponding window.Convolutional calculation is exactly that multiply accumulating calculates.If the result of convolutional calculation surpasses a setting, the registration success just is described, found registration parameter.
In order to improve computing velocity, hardware can adopt the mode of many multiply accumulatings device to improve concurrency.Here the hardware of saying is meant IC design here, but also can be general circuit design.The typical structure of the concurrent calculating of multichannel multiply accumulating device is as shown in Figure 1.
In the scheme of Fig. 1, in n template storage, there is the copy of a plurality of template storages.N template done registration with real-time figure simultaneously then and calculated simultaneously according to different parameter sense datas.Then relatively, find out the superior of registration result, the optimal registration result that its corresponding spatial parameter just finds through comparer.The difficulty of a maximum is exactly to need n template storage in this scheme.Cost is very high and the implementing process difficulty is very big when the n value is big.Because except storer increase can the scheme cost increase, the more important thing is need a large amount of from each template storage to the line the multiply accumulating device.Because n multiply accumulating device generally all is to be made in a slice integrated circuit, storer is made in outside the integrated circuit, and the quantity of integrated circuit extension line has just become the big difficult point that integrated circuit is realized so, even integrated circuit all can't have been realized when n is too big.
Summary of the invention
The present invention mainly is that solution existing in prior technology registration computing velocity is slow; Precision is low; The technical matters that IC design difficulty and operand and matching accuracy are difficult to take into account provides a kind of a small amount of leads of IC of need that the image registration device and method of higher matching accuracy, higher computing velocity is arranged again.
The present invention is directed to above-mentioned technical matters mainly is able to solve through following technical proposals: a kind of image registration device, be used for searching the relative tertiary location parameter of treating that registering images is corresponding from template, and it comprises:
Template storage is used to store the template data that carries out the proportioning operation;
The real-time storage device, be used to store carry out proportioning operation treat the registering images data;
Data are distributed parts, be used for selecting different transformation parameters according to the work allocation of each multiply accumulating device (MAC), and transformation parameter and the template data window that from template storage, takes out are sent to corresponding M AC;
Several multiply accumulating devices (MAC); Each multiply accumulating device is carried out the calculation task of one group of transformation parameter; Promptly according to transformation parameter adjustment template data window; And treat the registering images data window and make convolution algorithm with the later template data window of adjustment, treat that the registering images data window obtains from the real-time storage device;
Comparer is used for the operation result of each multiply accumulating device is compared, and finds out maximal value.
Template storage connects data and distributes parts, and data are distributed parts and are connected with the multiply accumulating device.
As preferably, the image registration device also comprises buffer register, and said buffer register is connected between real-time storage device and the multiply accumulating device, is used to store the data that will mate calculating, treats that registering images data window and template data window are synchronous so that make.
As preferably, said data are distributed parts and are comprised and select library, data register and the selector switch corresponding with each MAC that said selection library is connected with said selector switch respectively.Select library the control signal of transformation parameter and each selector switch to be provided, make selector switch select the required data of this MAC according to the task of each multiply accumulating device.Said data register is connected with said template storage with said selector switch respectively.
Comparer is used for seeking the maximum as a result of each multiply accumulating device computing.It is connected with all multiply accumulating devices with the mode of bus, can reduce between multiply accumulating device and the comparer line one to one like this, has reduced the complexity of integrated circuit.
A kind of method for registering images may further comprise the steps:
Step a, data are distributed parts and from template data, are extracted a template data window according to underlying parameter and send into data and distribute parts.So-called underlying parameter is exactly the estimates of parameters of before the registration of doing this programme calculates, being known, this programme is further done smart coupling on this basis;
Step b, data are distributed the division of labor of parts according to each multiply accumulating device; Select different transform datas; And template data window and transform data sent to corresponding multiply accumulating device, selection course is by selecting library to accomplish according to the selection signal controlling selector switch that designs and be stored in advance in this storehouse;
Step c, real-time storage device are chosen the treat registering images data window identical with the template data window shape and are sent to each multiply accumulating device from treat the registering images data;
Steps d, each multiply accumulating device are to the template data window that receives and treat that the registering images data window does convolution algorithm, and operation result sends to comparer successively;
Step e, comparer compare the operation result that each multiply accumulating device sends over; Select the maximum; And will try to achieve the pairing MAC numbering of peaked multiply accumulating device as registration result output, can confirm to treat the space geometry relation of registering images and template through the MAC numbering.
As preferably, the parameter of said conversion comprises rotation, translation, convergent-divergent and inclination.
As preferably, the said anglec of rotation comprises 0 °, 15 °, 30 ° ,-15 ° and-30 °.
As preferably, said translation comprises the unit that do not move, moves to left, the unit that moves to right, on move a unit and move down a unit.
As preferably, that said convergent-divergent comprises is constant, dwindle 25% and amplify 25%.
As preferably, said angle of inclination comprises-15 °, 0 ° and 15 °
The substantial effect that the present invention brings is, can obtain the image registration parameter at a high speed, uses minimum integrated circuit lead leg number to support a plurality of multiply accumulating devices (225) computing desired data, the simplifying integrated circuit design, and matching accuracy is high.
Description of drawings
Fig. 1 is a kind of circuit module figure of existing image registration circuit;
Fig. 2 is a kind of circuit module figure of the present invention;
Fig. 3 is a kind of hardware structure diagram that data of the present invention are distributed parts;
Among the figure: 1, template storage, 2, the real-time storage device, 3, the multiply accumulating device, 4, comparer, 5, data distribute parts, 6, buffer register, 51, select library, 53, selector switch, 54, data register.
Embodiment
Pass through embodiment below, and combine accompanying drawing, do further bright specifically technical scheme of the present invention.
Embodiment: a kind of image registration device, as shown in Figure 2, comprise that template storage 1, real-time storage device 2, data distribute parts 5, buffer register 6, comparer 4 and 225 multiply accumulating devices 3.Template storage is distributed parts through data and is connected each multiply accumulating device, and the real-time storage device connects each multiply accumulating device through buffer register, and comparer connects each multiply accumulating device through bus.
As shown in Figure 2, data are distributed parts and are comprised selection library 51, data register 54 and several selector switchs 53, select library to be connected with selector switch with parameter register respectively, and data register is connected with template storage with selector switch respectively.
The method for registering images of said apparatus may further comprise the steps: step a, data are distributed parts and from template data, are extracted a template data window according to underlying parameter and send into data and distribute parts.So-called underlying parameter is exactly the estimates of parameters of before the registration of doing this programme calculates, being known, this programme is further done smart coupling on this basis;
Step b, data are distributed the division of labor of parts according to each multiply accumulating device; Select different transform datas; And template data window and transform data sent to corresponding multiply accumulating device, selection course is by selecting library to accomplish according to the selection signal controlling selector switch that designs and be stored in advance in this storehouse;
Step c, real-time storage device are chosen the treat registering images data window identical with the template data window shape and are sent to each multiply accumulating device from treat the registering images data;
Steps d, each multiply accumulating device are to the template data window that receives and treat that the registering images data window does convolution algorithm, and operation result sends to comparer successively;
Step e, comparer compare the operation result that each multiply accumulating device sends over; Select the maximum; And will try to achieve the pairing MAC numbering of peaked multiply accumulating device as registration result output, can confirm to treat the space geometry relation of registering images and template through the MAC numbering.
The parameter of conversion comprises rotation, translation, convergent-divergent and inclination.The anglec of rotation comprises 0 °, 15 °, 30 ° ,-15 ° and-30 °.Translation comprises the unit that do not move, moves to left, the unit that moves to right, on move a unit and move down a unit.That convergent-divergent comprises is constant, dwindle 25% and amplify 25%.The angle of inclination comprises-15 °, 0 ° and 15 °
The process of seeking optimized parameter is such:
At first from template storage, read the comparison window data according to basic parameter; Data are distributed parts are done various parameter fine settings on the basis of this basic parameter combination, from data register, extract data and send to corresponding multiply accumulating device according to the rule of parameter fine setting according to the division of labor of each multiply accumulating device then.
Each MAC (multiply accumulating device) accomplishes the calculating of all parameter combinations jointly.For each parameter combinations, each MAC has a kind of pixel data selection scheme, and this scheme is pre-existing in to be selected in the library.Select library to preserve the selector control signal of the MUX of each MAC.
For every group of parameter combinations, order executes the convolution of these 225 groups of data, then the result of calculation of each MAC is all delivered to comparer in proper order.Maximal value and corresponding with it multiply accumulating device that comparer is always preserved in the result of calculation are numbered, and can know parameter combinations through multiply accumulating device numbering.
225 MAC handle all parameter combinations jointly owing to adopt nearly, so speed is fast doubly more a lot of than single MAC.
Specific embodiment described herein only is that the present invention's spirit is illustrated.Person of ordinary skill in the field of the present invention can make various modifications or replenishes or adopt similar mode to substitute described specific embodiment, but can't depart from spirit of the present invention or surmount the defined scope of appended claims.
Although this paper has used data to distribute terms such as parts, template storage morely, do not get rid of the possibility of using other term.Using these terms only is in order to describe and explain essence of the present invention more easily; It all is contrary with spirit of the present invention being construed to any additional restriction to them.
Claims (9)
1. an image registration device is used for searching the position of treating that registering images is corresponding from template, it is characterized in that it comprises:
Template storage is used to store the template data that carries out the proportioning operation;
The real-time storage device, be used to store carry out proportioning operation treat the registering images data;
Data are distributed parts, be used for selecting different transformation parameters according to the work allocation of each multiply accumulating device, and transformation parameter and the template data that from template storage, takes out are sent to corresponding multiply accumulating device;
Several multiply accumulating devices; Each multiply accumulating device is carried out the calculation task of one group of transformation parameter; Promptly adjust template data, and treat the registering images data and make convolution algorithm, treat that the registering images data obtain from the real-time storage device with the later template data of adjustment according to transformation parameter;
Comparer is used for the operation result of each multiply accumulating device is compared, and finds out the maximum.
2. image registration device according to claim 1 is characterized in that, also comprises buffer register, and said buffer register is connected between real-time storage device and the multiply accumulating device, is used to make real time data and template data synchronous.
3. image registration device according to claim 1 and 2; It is characterized in that; Said data are distributed parts and are comprised selection library, data register and the selector switch corresponding with each multiply accumulating device; Said selection library is connected with said selector switch respectively, and said data register is connected with said template storage with said selector switch respectively.
4. a method for registering images is characterized in that, may further comprise the steps:
Step a, data are distributed parts and from template data, are extracted a template data window according to underlying parameter and send into data and distribute parts;
Step b, data are distributed the division of labor of parts according to each multiply accumulating device; Select different transform datas; And template data window and transform data sent to corresponding multiply accumulating device, selection course is by selecting library to accomplish according to the selection signal controlling selector switch that designs and be stored in advance in this storehouse;
Step c, real-time storage device are chosen the treat registering images data window identical with the template data window shape and are sent to each multiply accumulating device from treat the registering images data;
Steps d, each multiply accumulating device are to the template data window that receives and treat that the registering images data window does convolution algorithm, and operation result sends to comparer successively;
Step e, comparer compare the operation result that each multiply accumulating device sends over; Select the maximum; And will try to achieve the pairing MAC numbering of peaked multiply accumulating device as registration result output, can confirm to treat the space geometry relation of registering images and template through the MAC numbering.
5. method for registering images according to claim 4 is characterized in that the parameter of said conversion comprises rotation, translation, convergent-divergent and inclination.
6. method for registering images according to claim 5 is characterized in that, the said anglec of rotation comprises 0 °, 15 °, 30 ° ,-15 ° and-30 °.
7. method for registering images according to claim 5 is characterized in that, said translation comprises the unit that do not move, moves to left, the unit that moves to right, on move a unit and move down a unit.
8. method for registering images according to claim 5 is characterized in that, that said convergent-divergent comprises is constant, dwindle 25% and amplify 25%.
9. method for registering images according to claim 5 is characterized in that, said angle of inclination comprises-15 °, 0 ° and 15 °.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201110372032.0A CN102521826B (en) | 2011-11-22 | 2011-11-22 | Image registration device and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201110372032.0A CN102521826B (en) | 2011-11-22 | 2011-11-22 | Image registration device and method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102521826A true CN102521826A (en) | 2012-06-27 |
| CN102521826B CN102521826B (en) | 2015-05-13 |
Family
ID=46292730
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201110372032.0A Expired - Fee Related CN102521826B (en) | 2011-11-22 | 2011-11-22 | Image registration device and method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102521826B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106845550A (en) * | 2017-01-22 | 2017-06-13 | 阿依瓦（北京）技术有限公司 | A kind of image-recognizing method based on multi-template |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080193001A1 (en) * | 2007-02-09 | 2008-08-14 | Agfa Healthcare Nv | Visual Enhancement of Interval Changes Using Temporal Subtraction |
| CN101576961A (en) * | 2009-06-16 | 2009-11-11 | 天津大学 | High-speed image matching method and device thereof |
| CN202339580U (en) * | 2011-11-22 | 2012-07-18 | 苏州科雷芯电子科技有限公司 | Image registration device |
-
2011
- 2011-11-22 CN CN201110372032.0A patent/CN102521826B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080193001A1 (en) * | 2007-02-09 | 2008-08-14 | Agfa Healthcare Nv | Visual Enhancement of Interval Changes Using Temporal Subtraction |
| CN101576961A (en) * | 2009-06-16 | 2009-11-11 | 天津大学 | High-speed image matching method and device thereof |
| CN202339580U (en) * | 2011-11-22 | 2012-07-18 | 苏州科雷芯电子科技有限公司 | Image registration device |
Non-Patent Citations (2)
| Title |
|---|
| 卢丽君: "InSAR影像配准及其并行化算法研究", 《中国优秀硕博士学位论文全文数据库(硕士) 基础科学辑》, no. 05, 15 May 2006 (2006-05-15), pages 19 - 20 * |
| 杨靓等: "最大互相关图象匹配算法的硬件实现", 《微电子学与计算机》, no. 5, 31 December 1999 (1999-12-31), pages 28 - 30 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106845550A (en) * | 2017-01-22 | 2017-06-13 | 阿依瓦（北京）技术有限公司 | A kind of image-recognizing method based on multi-template |
| CN106845550B (en) * | 2017-01-22 | 2020-03-17 | 阿依瓦（北京）技术有限公司 | Image identification method based on multiple templates |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102521826B (en) | 2015-05-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110310326B (en) | Visual positioning data processing method and device, terminal and computer readable storage medium | |
| DE112018001050T5 (en) | SYSTEM AND METHOD FOR VIRTUALLY ENHANCED VISUAL SIMULTANEOUS LOCALIZATION AND CARTOGRAPHY | |
| US8004557B2 (en) | Advanced dynamic stitching method for multi-lens camera system | |
| EP3296953B1 (en) | Method and device for processing depth images | |
| CN114527872B (en) | Virtual reality interaction system, method and computer storage medium | |
| US20160286138A1 (en) | Apparatus and method for stitching panoramaic video | |
| US11107246B2 (en) | Method and device for capturing target object and video monitoring device | |
| CN110631554A (en) | Robot posture determining method and device, robot and readable storage medium | |
| US9781340B2 (en) | Image processing system, information processing device, information storage device, and image processing method | |
| US9390330B2 (en) | Apparatus and method for extracting correspondences between aerial images | |
| KR20110042971A (en) | Marker-less augmented reality system using projective invariant and method the same | |
| CN104457758B (en) | Video-acquisition-based Visual Map database establishing method and indoor visual positioning method using database | |
| CN104182974A (en) | A speeded up method of executing image matching based on feature points | |
| GB2605727A (en) | Stream object tracking with delayed object detection | |
| CN103595915A (en) | Method for controlling video location positioning of high-definition speed dome camera | |
| GB2603715A (en) | Depth estimation using a neural network | |
| CN110827321A (en) | Multi-camera cooperative active target tracking method based on three-dimensional information | |
| CN108257153B (en) | Target tracking method based on direction gradient statistical characteristics | |
| CN113532420A (en) | Visual inertial odometer method integrating point-line characteristics | |
| CN111179309B (en) | Tracking method and device | |
| US20200234465A1 (en) | Supermarket shopping cart positioning method, supermarket shopping cart positioning system, and supermarket shopping cart | |
| CN202339580U (en) | Image registration device | |
| CN111684462B (en) | Image matching method and vision system | |
| CN111508072A (en) | Map construction method and device, electronic equipment and storage medium | |
| CN102521826A (en) | Image registration device and method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20150513 Termination date: 20191122 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |