A Real-Time Image Registration Algorithm Based on Feature-Index

A Feature-Index based algorithm that uses similarity information and spatial coherence is presented in the paper. It especially fits to the case that the base image is very large and real-time property is needed. The algorithm is divided into four steps: firstly extract features from base image, and establish feature-index table beforehand; then extract corresponding features from sensed image and use the feature-index table to match them with those from base image; thirdly, fuse the matching results of Main line pairs (MLPs) and contours; finally, solve the transformation parameters using the intersection points of MLPs and centers of contours and complete registration. The experiment results show that the proposed algorithm is accurate and effective.