Scene Image Mosaic Based on Three Freedom Degree
Chengcheng Liu, Yong Shi · Procedia Computer Science · 2013
This study aims to propose a feature matching method, which can stitch together scattered photographs of the same scene (or target), so as to restore a scene (or target) of the picture. For this question, our paper proposes a method based on Scale-invariant feature transform (SIFT) algorithm for image registration. SIFT algorithm is obtained by judging the feature points of local extreme, combined with neighborhood information to describe the feature points to form a feature vector, in order to build the matching relationship between the feature points. First, we extract feature vector via SIFT algorithm, obtain feature point pairs. Then, we filter feature points according to the information provided by the nearest neighbor and the second nearest neighbor, in order to obtain efficient feature point pairs. After that, we get scaling, rotation angle and translation vector based on the relative position between feature points. We get the scaling of feature points by the ratio of the average distance between the accesses, and get the rotation angle by the demand that vectors connecting feature point pairs should be parallel. Finally, we remove some wrong matching feature point pairs by judge parallel of the vector, to improve the accuracy, and use the average angle and its average value as the translation model parameters. The experimental shows that the method is reliable and effective.