Low-altitude Aerial Photographs Video Image Stitching

Image mosaic is brought for the limitation of camera visual angle, which is impossible to shoot a panorama image once. Computer technology is used to register and stitch these images which have overlap areas. The subject of this dissertation is a representative application for image mosaic.The main contents are as followed:1. Capture images of video streams by DirectShow interface, then the single frames are acquired. The step is base for the whole job.2. Image registration which is the key question for image mosaic is discussed. There are two kind of image registration, for phase correlation and for image feature, which are discussed particular in the later sections.3. In the registration based phase correlation, the steps and principle are given. The improved phase correlation algorithm is discussed.4. For the registration based image feature, there are also two kinds of algorithm, for the area feature and for the point feature. BMMR algorithm is introduced for the first kind. Moravec corner detector, Harris corner detector, SUSAN detection and an improved SUSAN detector which is given by this dissertation are described. The robust estimate is introduced after the corner feature is acquired.5. Image stitching is discussed when the registration information is acquired. Anti-gap algorithm is used to improve the panorama visual quality.