Font Size: a A A

Research Of Automatic Image Mosaic Algorithm Based On Feature Points

Posted on:2012-07-06Degree:MasterType:Thesis
Country:ChinaCandidate:Q J TianFull Text:PDF
GTID:2218330368488299Subject:Communication and Information System
Abstract/Summary:PDF Full Text Request
Image mosaic is an important research topic of image-based rendering and it is a main research direction of computer graphics, computer vision, image processing, virtual reality and so on. It can be used to construct a high resolution panoramic image of super wide view with two or more overlapping sequence of photos. Image mosaic which provides great contribution to solve the contradiction between resolution and size of scene has been used widely in domains of remote sensing technology, medical image processing, military and video compression and transmission.Image mosaic method which was based on feature point has a significant advantage in stability and adaptability. It can be used to process not only the images with low degree of coincidence but also the images with moving scenes and occlusion. This paper has focused on the feature matching and paid attention to the SIFT algorithm which has robust performance on change of rotation, scale and noise. Then we have proposed a robust stitching algorithm of feature point which was based on the registration model of affine transformation.The main contribution of this paper includes improving the accuracy of matching executed by SIFT and purifying the parameters of affine transformation used least square method. At the respect of feature searching, we used BBF against with the shortage of K-D tree in high-dimensional feature. We abandoned outliers determined by RANSAC which was constrained by epipolar geometry. Also, we combined bidirectional matching and constraint geometric relation to analysis the phenomenon of false matches. Others, we have given an experimental result of probabilistic relax matching which was used in the SIFT algorithm. Then we increased the accuracy of image registration by decomposing the matrix of affine transformation using singular.At last, we used method of weighted average to handle with the gap of stitching image. The experimental result shows that the algorithm of this paper can eliminate the gap well and achieved a satisfied consequent of image stitching...
Keywords/Search Tags:Image mosaic, Image registration, Feature point matching, Scale Invariant Feature Transform (SIFT), False matching, Probabilistic relax matching
PDF Full Text Request
Related items