无序图像自动匹配与识别
|
摘要
设计一个多幅无序图像的自动匹配和识别系统,能够根据用户输入的多幅无序图像进行自动匹配和识别,并对具有重叠的图像进行自动拼接。系统首先对输入的每一幅图像进行MOPS特征检测,然后通过k-d树的最近邻搜索完成不同图像特征之间的快速匹配。其次基于图像特征之间的对应关系使用RANSAC算法建立任意两幅图像之间的匹配模型,并用概率算法进行鲁棒校验。通过构建与图像匹配关系对应的无向连通图结构,实现多幅无序图像的自动识别。最后使用递归算法对无向连通图进行深度优先遍历,并用多频带融合算法消除拼接痕迹,合成相应的全景图像。实验结果表明该系统能够自动对多幅无序图像进行自动匹配与识别,验证了算法的可行性和有效性。
We design an automatic matching and recognition system for multiple unordered images,it is able to automatically match and recognise multiple unordered images inputted by users,and automatically stitch the images with overlapping portions. First,the system detects MOPS feature on every inputting image,and completes the fast matching between different image features through nearest neighbour search of k-d tree. The next,based on corresponding relationship between image features,it uses RANSAC algorithm to build the matching model between any two images,and verifies the robustness with probabilistic algorithm. The automatic recognition on the multiple unordered images can be achieved by constructing the structure of undirected connected graph corresponding to image matching relationship. Finally,the recursive algorithm is used to do depth-first traversal across the undirected connected graph,and the multi-band fusion algorithm is employed to eliminate stitching seam,as well as to compose corresponding panoramas image. Experimental results show that the system can automatically match and recognise the multiple unordered images,the feasibility and effectiveness of the algorithm is also verified.
We design an automatic matching and recognition system for multiple unordered images,it is able to automatically match and recognise multiple unordered images inputted by users,and automatically stitch the images with overlapping portions. First,the system detects MOPS feature on every inputting image,and completes the fast matching between different image features through nearest neighbour search of k-d tree. The next,based on corresponding relationship between image features,it uses RANSAC algorithm to build the matching model between any two images,and verifies the robustness with probabilistic algorithm. The automatic recognition on the multiple unordered images can be achieved by constructing the structure of undirected connected graph corresponding to image matching relationship. Finally,the recursive algorithm is used to do depth-first traversal across the undirected connected graph,and the multi-band fusion algorithm is employed to eliminate stitching seam,as well as to compose corresponding panoramas image. Experimental results show that the system can automatically match and recognise the multiple unordered images,the feasibility and effectiveness of the algorithm is also verified.
引文
[1]Bartoli A.Groupwise geometric and photometric direct image registration[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2008,30(2):2098-2108.
[2]Gay V.Direct Estimation of Nonrigid Registrations with Image-Based Self-Occlusion Reasoning[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2010,32(1):87-104.
[3]Sandip R,Satish K.A Survey:Methods of Feature Based Image Registration[J].International Journal of Electronics Communication and Computer Engineering,2012,3(4):787-791.
[4]Jose S,Oscar C.GRASP and path relinking hybridizations for the point matching-based image registration problem[J].Journal of Heuristics,2012,18(1):169-192.
[5]Lowe D.Distinctive image features from scale-invariant keypoints[J].International Journal of Computer Vision,2004,60(2):91-110.
[6]Calonder M,Lepetit V.BRIEF:Binary Robust Independent Elementary Features[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2012,34(7):1281-1298.
[7]Ethan R,Vincent R,Kurt K,et al.ORB:An efficient alternative to SIFT or SURF[C]//13th International Conference on Computer Vision.Barcelona,2011:2564-2571.
[8]Pierre M,Pietro P.Evaluation of Features Detectors and Descriptors based on 3D Objects[J].International Journal of Computer Vision,2007,73(3):263-284.
[9]Steffen G,Tobias H,Matthew T.Evaluation of Interest Point Detectors and Feature Descriptors for Visual Tracking[J].International Journal of Computer Vision,2011,94(3):335-360.
[10]Mikolajczyk K,Schmid C.A Performance Evaluation of Local Descriptors[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2005,27(10):1615-1630.
[11]Noah S,Rahul G,Steven M,et al.Finding Paths through the World's Photos[J].ACM Transactions on Graphics,2008,27(3):11-21.
[12]Furukawa Y,Snavely N,Curless B,et al.Reconstructing Rome[J].IEEE Computer,2010,43(6):40-47.
[13]Byrod M,Brown M,Astrom A.Minimal Solutions for Panoramic Stitching with Radial Distortion[C]//Proceedings of the British Machine Vision Conference.London,2009:1-7.
[14]Brown M,Szeliski R,Winder S.Multi-Image Matching using MultiScale Oriented Patches[C]//International Conference on Computer Vision and Pattern Recognition.Anchorage:IEEE,2005:510-517.
[15]Jongmoo C,Medioni G.StaRSaC:Stable random sample consensus for parameter estimation[C]//IEEE Conference on Computer Vision and Pattern Recognition.Miami,2009:675-682.
[16]Allene C,Pons J,Keriven R.Seamless image-based texture atlases using multi-band blending[C]//19th International Conference on Pattern Recognition.Tampa,2008:1-4.
[2]Gay V.Direct Estimation of Nonrigid Registrations with Image-Based Self-Occlusion Reasoning[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2010,32(1):87-104.
[3]Sandip R,Satish K.A Survey:Methods of Feature Based Image Registration[J].International Journal of Electronics Communication and Computer Engineering,2012,3(4):787-791.
[4]Jose S,Oscar C.GRASP and path relinking hybridizations for the point matching-based image registration problem[J].Journal of Heuristics,2012,18(1):169-192.
[5]Lowe D.Distinctive image features from scale-invariant keypoints[J].International Journal of Computer Vision,2004,60(2):91-110.
[6]Calonder M,Lepetit V.BRIEF:Binary Robust Independent Elementary Features[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2012,34(7):1281-1298.
[7]Ethan R,Vincent R,Kurt K,et al.ORB:An efficient alternative to SIFT or SURF[C]//13th International Conference on Computer Vision.Barcelona,2011:2564-2571.
[8]Pierre M,Pietro P.Evaluation of Features Detectors and Descriptors based on 3D Objects[J].International Journal of Computer Vision,2007,73(3):263-284.
[9]Steffen G,Tobias H,Matthew T.Evaluation of Interest Point Detectors and Feature Descriptors for Visual Tracking[J].International Journal of Computer Vision,2011,94(3):335-360.
[10]Mikolajczyk K,Schmid C.A Performance Evaluation of Local Descriptors[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2005,27(10):1615-1630.
[11]Noah S,Rahul G,Steven M,et al.Finding Paths through the World's Photos[J].ACM Transactions on Graphics,2008,27(3):11-21.
[12]Furukawa Y,Snavely N,Curless B,et al.Reconstructing Rome[J].IEEE Computer,2010,43(6):40-47.
[13]Byrod M,Brown M,Astrom A.Minimal Solutions for Panoramic Stitching with Radial Distortion[C]//Proceedings of the British Machine Vision Conference.London,2009:1-7.
[14]Brown M,Szeliski R,Winder S.Multi-Image Matching using MultiScale Oriented Patches[C]//International Conference on Computer Vision and Pattern Recognition.Anchorage:IEEE,2005:510-517.
[15]Jongmoo C,Medioni G.StaRSaC:Stable random sample consensus for parameter estimation[C]//IEEE Conference on Computer Vision and Pattern Recognition.Miami,2009:675-682.
[16]Allene C,Pons J,Keriven R.Seamless image-based texture atlases using multi-band blending[C]//19th International Conference on Pattern Recognition.Tampa,2008:1-4.