|Table of Contents|

Robust Image Mosaicing by Corner Matching


Research Field:
Publishing date:


Robust Image Mosaicing by Corner Matching
LI Zhong-xin1MAO Yao-bin2WANG Zhi-quan2
1.School of Mechanical Engineering;2.School of Automation,NUST,Nanjing 210094,China
imagemosaic ing corner detection cornermatch ing random sample consesus algorithm
A robust imagemosaicingmethod is put forw ard to construct panoram ic image. Themethod uses Harris corner detector to ex tract corner-po ints for each image. A norma lized cross correlat ion method is used to coarselymatch corner-pairs betw een image pairs. Due to the ex istence of the outl-i ers, direct use of least square a lgorithm w ou ld lead to the inaccurate parameter est imation o f the transformationmodel that is required in image reg istration. To so lve the problem, theRandom Sample Consensus ( RANSAC) algorithm is introduced to est imate the transformat ionmodel parameters, wh ich is able to elim inate the outliers and to obta in exact co rner-pa irs. A ccording to the aforement ioned steps, thew ho le process of the imagemosa icing is automatica lly accomplished. Some rea l images are subject to corner detection, reg istration and fina lmosaic ing, wh ich results in an excellent outcome


[ 1] Iran iM, Anandan P, Bergen J, et a.l Effic ient represen tations o f v ideo sequences and the ir applications [ J]. S igna lProcessing: Im age Commun ication, 1996, 8 ( 4): 327- 351.
[ 2] Chaudha r,i Kap ilA, Reeves, Stanley J. M osa ic im ag e com pression [ A]. Pro ceedings of SPIE- The International Soc iety fo r Optical Eng ineering [ C] . Bellingham: Internationa l Soc iety for Optical Eng inee ring, 2005. 189- 197.
[ 3] Shum H Y, Ng K T, Ch S C. A v irtual rea lity sy stem using the concentr icm osa ic: Construction, rende ring, and da ta compression [ J]. IEEE Transac tions onM u-l tim ed ia, 2005, 7 ( 1): 85- 95.
[ 4] M ann S, P icard R W. V ideo orb its o f the projectiv e g roup: A simp le appro ach to featureless estim ation o f param eters [ J]. IEEE transaction on im age processing, 1996, 6 ( 9): 1 280- 1 295.
[ 5] Sze liskiR. V ideom osa ics fo r v irtual environm en ts [ J]. IEEE Com puter Graph ics and App lications, 1996, 16 ( 2): 22- 30.
[ 6] H su C T, Cheng T H, Beuker, et a.l Fea ture based v ideo m osa ic [ A ]. Proceedings o f Interna tiona l Conference on Im age Pro cessing [ C]. V ancouver: Institute of E lectrical and E lec tron ics Eng ineers Compu ter Soc ie ty, 2000. 887- 890.
[ 7] Zogham i I, Fauge ras O, Der iche R. Using geom etric corners to build a 2D m osa ic from a set o f images [ A] . Proceed ings o f the IEEE Conference on Com pu-t er V is ion and Pa ttern Recogn ition [ C]. Los A lam itos: IEEE, 1997. 420- 425.
[ 8] H arris C, StephensM. A com bined corner and edg e detecto r [ A ]. Proceed ings Fourth A lvey V ision Conference [ C ]. M anchester: IEEE, 1988. 147- 151.
[ 9] Di Stefano L, M attocc ia S, M o laM. An effic ient a lgorithm for exhaustive tem platem atch ing based on norm a-l ized cross co rre lation [ A]. Proceedings of 12th In ternationa l Conference on Im age Ana ly sis and Processing [ C]. Los A lam itos: Institute of E lectrica l and E lectron ics Eng ineers Computer Society, 2003. 322- 327.
[ 10] F isch le rM A, Bo lles R C. Random samp le consensus: A paradigm fo r mode l fitting w ith application to im ag e ana lysis and autom ated cartography [ J] . Comm un ica tions of the ACM, 1981, 24 ( 6) : 381- 395.
[ 11] 马颂德, 张正友. 计算机视觉[M ]. 北京: 科学出版 社, 1998.


Last Update: 2007-06-30