Feature Based Sequence Image Stitching Method

Abstract

Image mosaic is useful for the traditional image process. We proposed a novel feature based sequence image stitching method for the image mosaic system. In this method, the whole image registration process has five steps, which includes feature point detection, feature descriptor extraction, feature points matching, estimation of the motion model parameters and stitching process. In these steps, difference of Gaussian image is used to get the extreme points as the feature points, then the SIFT descriptor operator is used to describe the feature, finally random sample consensus method is used for estimating the motion parameters. We test our method on two kinds of sequence images, the artificial images and the real images. The results show the stitching method is robust.