Research Of Fearure-Based Image Stitching

Image stitching or photo stitching is the process of combining multiple photographic images with overlapping fields of view to produce a segmented panorama or high-resolution image. The basic requirements of Image stitching are that we can get a high-definition image without apparent junctures after image stitching, and the image reserves the information of the original images as far as possible.The commonly used approaches in image registration processing are methods based on the outline of feature extraction, methods based on the angular points of feature extraction and SIFT algorithm based on the scale space. SIFT has very good scale invariance and it also has good stability in changing light and various transforms such as image translation, rotation, scaling and so on, therefore this paper adopts SIFT as feature extraction algorithm and improve it in data purification with RANSAC algorithm, and eventually makes many images get together.In order to eliminate exposure differences existing in the overlaps fields after image stitching processing, this paper adopts the multiresolution fusion method. And the experimental results show that the stitching algorithm in this paper has achieved seamless mosaic effect successfully.