Research Of Image Registration Based On Feature Extraction And Matching Method

Image registration technique is a very important step in the image processing,in terms of image registration techniques,officials have given attention. Both in the field of image processing and other fields such as medical image registration, image registration techniques are widely researched and applied.In the first chapter of this article introduces the research purpose and meaning of this subject, then image registration techniques are summarized in detail for nearly 10 years of development situation at home and abroad, mainly for image registration process is applied to the algorithm and application are reviewed. Finally illustrates the application of this technology in the change tracking and morphological aspects. Then in the second chapter introduces the related theory of rigid body registration technology, verify the non-rigid image registration in principle.Research on feature point extraction and matching method, include Scale-invariant feature transform feature extraction and Harris corner detection method. This chapter expounds on the two methods of feature extraction and feature matching theory method, and respectively simulation results and analysis results is given at the end using SIFT algorithm and Harris corner detection for image matching. Finally summarizes the advantages and disadvantages of two algorithms. In the end of the paper, basing on the method of Scale-invariant feature transform algorithm is put forward, called Speeded Up Robust Feature algorithm. In this paper, at last this paper expounds the SURF for feature extraction of whole steps, then after eliminating false matching points combined with RANSAC algorithm using least square method to complete parameter transformation, finally using bilinear interpolation method to complete the resampling and the image registration. This article uses MATLAB computer simulation platform is 2010. It is found that the SURF not only faster, but also perfect complete the whole registration process than the SIFT from the simulation result, it is for researches on the feature extraction of a big stride.