Image Registration of Blown and Swayed Trees
This paper concerns the problem of blown and swayed tree image registration. The swayed tree leaves and little branches contain plenty of details or feature points, but matching with these points may result in error image registration. In this work, we delete some details of leaves and little and thin branches by using of morphological image processing methods, then extract the relatively invariant crotch features of trees from thinning images, lastly we adjusted the matching points with L-M algorithm. Results show our method is insensitive to the ordering, rotation and scale of the input images so it can be used in image stitching and retrieval of images & videos.
D. Y. Han and X. Y. Huang, "Image Registration of Blown and Swayed Trees", Key Engineering Materials, Vols. 474-476, pp. 2183-2188, 2011