Research On An Robust Image Stitching Algorithm

Image Stitching is the process which builds the panorama with a series of images having parallax from one or multi-sensor. Image Stitching has important applications in image processing field, e.g. digital video, computer vision, computer graphic, virtual reality, remote image processing, medical image processing and so on.Generally, to get a panorama with high quality, we need some special, expensive and carefully calibrated equipment. Image stitching is a cheap and flexible approach to get panorama with a simple hand-held camera.The photos taken with a hand-held camera usually have large perspective distortion, small overlap, brightness difference and camera rotations. These characteristics make both image alignment and panorama building more difficult than using photos taken by cameras calibrated by special equipment. In this paper, we propose a robust algorithm which composes of feature points extracting, matching, estimating the transform relationship and blending finally. Firstly, we analyze variousapproach to extract feature points—corner. Then, we purify the inliers which areextracted by Harris Detector using a robust filter strategy, and insure the selection after purifying be able to estimate the transform relationship precisely. We decrease the constraint in the aspect of extracting feature points and matching nicely, so the images still can be matched precisely from those obtained in the flexible condition.Fmally, I implement an image stitching system and demonstrate the steps with experimental pictures. The steps demonstration of the image stitching system are realized by VC++ program. Some experimental pictures are showed in this paper.