Advanced Technology Research On Image Stitching

As one of the fast developing digital image processing technology, image stitching solves the contradictions that a camera is not able to balance the wide viewing angle with high resolution at the same time. It outputs a panorama image with wide viewing angle and high resolution that contains all the information for the scenario. It has brought a lot of help to daily lives,industrial manufacture and security applications. As this technology involves numerous fields of image processing, traditional stitching method can not work well in some special cases. Through a deep research on each processing state and based on the analysis of present algorithms of panorama image stitching, an improved solution system was proposed based on image feature points.The research contents and contributions are summarized as follows:1. This thesis studies the whole system of image stitching and each processing state are analyzed in details, including the image distortion correction ? image registration? bundle adjustment and image blending. Combined with the practices of our own project, this paper proposed a basic system framework built with the most suitable methods. This basic system obtained a satisfied performance in most cases.2. A brightness adjustment method for images sequence based on image feature points was proposed to solve the problem that the traditional blending ways are unable to deal with image sequence of large exposure difference. Since the matched feature points from different image corresponds to the same point in the real world, the pair of feature points should have the same or consistent brightness. This method makes the most of brightness information of feature point to adjust the brightness of disqualified images. The method effectively solves the inconsistent problem lies in the luminance of panorama image, and obtained good performance in practice.3. A Kruppa equation based method was proposed to determine the focal length of cylinder projection model, overcoming the severe flaw of cylindrical image stitching that focal length must be known. This method adopted the Kruppa equation from the theory of camera self-calibration, and simplified the camera intrinsic parameters. This method can not only reduce the computation burden, but also is practicable in engineering application.