The Study Of Image Stitching Techniques Based On Point-Like Feature And Mutual Information

Image stitching is the process of obtaining a wider field-of-view of a scene from a sequence of partial views.Image stitching combined the complementary information from each image and discarded the redundant information between the images,formed an image which contained the overall information of the targeted scene.This paper studies and analyzes the theory and the state-of-the-art of image stitching.The detailed study contents are as below:1)The image stitching,image registration and image fusion theory were researched.The basic problem of image stitching was analyzed and solutions to the problem were introduced.Six major image transformation models and four image acquisition pattern related to image stitching was introduced.2)Two blob-based image stitching algorithms,based on SIFT(Scale Invariant Feature Transform)and SURF(Speeded-Up Robust Features)were designed and implemented.Both of the two image stitching algorithms can stitch images correctly and obtain a good-looking stitched image.3)An image stitching algorithm based on ORB(Oriented FAST and Rotated BRIEF)was designed and implemented.Experiment results show that the ORB-based algorithm can stitch images properly and the ORB-based algorithm is 7-8 times faster than the SIFT-based algorithm while 5 times faster than the SURF-based algorithm.4)An image sequences stitching algorithm based on normalized mutual information was designed and implemented.The algorithm can stitch images that only translated in horizontal direction.5)An image sequence stitching application was designed and implemented.The application can stitch the image sequences from the project properly,and obtain a 360°of view panorama.