基于SIFT特征匹配的图像拼接算法_李云霞.pdfVIP

19 　 1 Vol.19　No.1 2009 1 　　　　　　　　　COM PUTER TEC NOLOGY AND DEVELOPMENT 　　　　　　　　 Jan.　2009 SIFT 李云霞, 曾　毅, 钟瑞艳, 郭　涛 (山东大学控制科学与工程学院, 山东济南 250061) 　:SIFT , 、。SIFT , , 、、, 。 , , , 。, , , 。 :SIFT ;;;RANSAC ;L-M :TP301.6　　　　　　　:A　　　　　　　:1673-629X(2009)01-0043-03 Algorithm of Image Stitching Based on SIFT Feature Matching LI Yun-xia, ZENG Yi, Z ONG Rui-yan,GUO Tao (School of Control Science and Engineering, Shandong University, Jinan 250061, China) Abstract:A fast and accurate image stitching algorithm based on SIFT feature matching is proposed.SIFT featurematching algorithm is hot and difficult in the field of matching feature points research at home and abroad.Its match ability is strong, and can be dealt with translation, rotation, affine transformation circumstances matching problem between the two images.To someextent, images also have a more stable feature matching capacity evenin the perspective of arbitrary capture.In this paper, a new imagestitching algorithm is pro- posed on the basis of this match thinkingand multi-resolution spline.It uses multi-resolution splineto determine the width and splicing regional weighting function, so that images are blended at different resolutions, thus the blended-image is clear and smooth seamless. Algorithm in most cases can be automatically completed, the experimental results show that it improves the stitching speed and reduces the differenceof image brightness.It makes the blended-image natural and realistic. Key words:SIFT feature matching;image stitching;multi-resolution blending;RANSAC algorithm;L-M algorithm 0　　 , , 。 。, 1　SIFT , 、