In terms of the deficiency in the aspects that the higher computational complexity caused by excessive iterations and the easy happened stitching dislocation caused by the difficult-to-determine parameters. In this paper, an improved RANSANC algorithm based similarity degree is proposed and is applied in image mosaic. This improved algorithm includes that sorting rough matched points by similarity degree, calculating transformation matrix, rejecting obviously wrong matched points and executing classical RANSAC algorithm. It is demonstrated by the experiments that this algorithm can effectively remove wrong matched pairs, reduce iteration times and shorten the calculation time, meanwhile ensure the accuracy of requested matrix transformation. By this method can get high quality stitching images.
展开▼
