Feature-based image registration and mosaicing
Weilin Song, Han Wang · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2000
The technique of image mosaicing was analyzed in this paper. The progress of image mosaicing comprises two parts, namely image registration and image blending. The key issue of image mosacing is to calculate the transformations (Homography) from the input image pairs correctly and accurately. Feature based methods are adopted for image registration. Firstly, corners are extracted from the images to be glued together. Matches were found from these features manually. During the process of mosaicing, three different algorithms have been investigated including (1) the four-point algorithm, (2) the least square algorithm and (3) the unbiased least square algorithm which appliced Kanatani's statistical optimization theory. Experimental results show the effectiveness of the unbiased least square algorithm.