Research on image registration and mosaic based on vector similarity matching principle

Scale invariant feature transform (SIFT) is a better corner extraction algorithm, but there are still mismatching problems in the feature <span class="hit">matching</span> step. A new <span class="hit">matching</span> <span class="hit">principle</span> <span class="hit">based</span> <span class="hit">on</span> <span class="hit">vector</span> <span class="hit">similarity</span> is proposed and then it is compared with traditional <span class="hit">matching</span> <span class="hit">principle</span>. Firstly, the <span class="hit">matching</span> feature points are detected by the new <span class="hit">principle</span>. Mismatching points are further removed by using the mutual mapping theory. Secondly, transformation matrix is calculated by random sample consensus (RANSAC). Furthermore, the matrix is optimized by Levenberg-Marquardt algorithm (L-M). Lastly, <span class="hit">image</span> <span class="hit">mosaic</span> is realized by <span class="hit">image</span> fusion. Experimental results indicate that compared with traditional <span class="hit">matching</span> <span class="hit">principle</span>, new <span class="hit">matching</span> <span class="hit">principle</span> has improved <span class="hit">matching</span> accuracy. It is able to apply new <span class="hit">principle</span> to <span class="hit">image</span> <span class="hit">registration</span> and <span class="hit">image</span> <span class="hit">mosaic