Research On Matching And Stitching Method Based On2D Image Sequences

Image registration and stitching is throughout the application scope of the military,medicine, remote sensing, computer vision, and other aspects. It is the key issueof image Mosaic and changes the premise of the analysis, pattern recognition and so on.For image registration, there are a large amount of data, calculation complexinformation, demands higher registration precision. So the design and implement ofhigh precision fast image registration algorithm is imperative.This paper is committed to the research of image registration and mosaic algorithm.On the one hand, a angular point matching algorithm is proposed based on a circularneighborhood enhanced on the basis of Harris algorithm. On the other hand, accordingto the requirements of the image registration precision and scale of large images toimprove the image matching algorithm, this paper proposes an image stitching andfusing algorithm based on scale invariant features. The main work and contributionssummed up as follows:(1) This paper introduces the background of image matching technology of thesystem, the definition and understanding of image registration. Image registration isintroduced widely applied and the research of the necessity, as well as the currentresearch status at home and abroad. We expound the basic theory of image registrationand Mosaic.(2) An enhanced angular point matching algorithm is presented based on circularneighborhood, the algorithm was introduced in details of the principle and process, andthe simulation results and experimental data is presented.(3) A kind of image stitching algorithm based on scale invariant features is put forward. First ofall, the advantages of the scale invariant features are analyzed and the improvement based on scaleinvariant feature extraction and image matching algorithm is given. The process of feature extractionand matching is introduced.