Image Registration Approach Based on Log-Polar and SIFT

To resolve the problem of large angle and large scale image registration, an improved approach combining log-polar and SIFT is proposed. Firstly, the log-polar technique is implemented in order to achieve the preliminary registration result as well as estimate the arbitrary rotations parameters and large scale changes. Secondly, image is segmented into sub-blocks and six candidates of sub-blocks are extracted by information entropy, where SIFT features and moment features are fused to form a new feature descriptor. Finally, registration result is calculated by the matching points obtained by measuring the Euclidean distance. The experimental results show that the proposed algorithm is robust, fast with high-precision.