Image Stitching Algorithm Based on SURF and Wavelet Transform

Abstract

With the development of image stitching and its wide application, image stitching has become an important and heated topic in image processing. An effective image stitching algorithm based on SURF feature matching and wavelet transform image fusion is proposed in this paper. Firstly, SURF feature points in the two adjacent images are extracted and matched. Then rapid and accurate image registration can be achieved by adopting the improved RNASCA algorithm to removing the mismatched feature point pairs. The multi-resolution decomposition of the overlapping region is processed by Wavelet Transform. Then the multi-scale image fusion is processed by fade-in and fade-out in order to eliminate the stitching seam better. Experiments show that the fusion results of the overlapping region are natural and there also is a certain robustness for translation, rotation, scale and luminance variant.