Study On Fast Image Stitching Method Based On SIFT Feature Detection And Matching

With the rapid development of computer technology,the requirements of improving the quality of image are increasing constantly.The technology of image stitching,which can obtain wider field-of-view and high-resolution image,has always been an attractive research in the field of Computer Vision because of its wide ranges of applications,including virtual reality,remote sensing,video monitoring and unmanned aerial vehicle images.Image stitching refers to the alignment of two or multiple overlapping image sequences into a large combination image,which has wide field-of-view and high-resolution image containing all of image sequences information.The approach of image stitching can solve the contradiction between image vision and image resolution.In spite of many experts and scholars are constantly doing the improvement work and different ways of image stitching have been developed in the past,the problems of massive calculation burden and poor speed still exist in the process of image stitching,which limit the applications of image stitching.This thesis focuses mainly on how to speed up the image stitching and gives the in-depth study on it.Firstly,this thesis introduces the background of image stitching,the significance of research and the current research work.The related theories in the process of image stitching and the SIFT(Scale Invariant Feature Transform)feature points detection algorithm are provided.The characteristic and specificity of image stitching are analyzed after reading a large number of relevant literatures in English and Chinese.Then three approaches which improve the speed of image stitching are proposed in this thesis which are the fast image stitching algorithm based on SIFT feature vector image,the fast image stitching method based on SIFT with adaptive local image feature and the fast image stitching algorithm based on SIFT using projection correction.In the traditional image stitching method,the number of SIFT feature points detected is so large that it makes the procedure of feature matching fairly complicated.In view of the above,a fast image stitching algorithm based on SIFT feature vector image is proposed.Firstly,the method combined the phase correlation algorithm to determine the overlapping regions of the images to be stitched,and the detection range of the SIFT feature points is defined.Then we consider the spatial location information of feature points,and construct the feature vector images in order to limit the searching ranges in the feature matching,and get the matching points quickly.Experimental results show that the new method reduces a lot of unnecessary search,and improves the speed of stitching;while it consistently guarantees the quality of stitched image.Aiming at the high dimensionality of SIFT feature points and the massive calculation burden in the process of feature points detection,a fast image stitching method based on fast SIFT algorithm with adaptive local image feature is proposed.Firstly,the images are divided into blocks.And the feature types of these local image blocks are determined.The feature points of the local image blocks are extracted using different simplified method adaptively.Secondly,we use feature matching to get the transform matrix and the RANSAC algorithm is applied to remove the wrong matching point pairs.Finally,the stitched image can be obtained by image blending.Experimental results show that compared with the standard stitching algorithm,this method improves the stitching efficiency and efficiently overcomes the shortcomings of heavy computation in the process of image stitching.A fast image stitching algorithm based on SIFT using projection error correction by analyzing the characteristic and specificity of image stitching is proposed.Firstly,the feature detection area of this approach is only limited to part image blocks in the overlapping area of the to-be-stitched images,where we can obtain the SIFT feature points information.Then,the projection error correction method is applied after feature matching in order to make full use of the finite matching pairs to calculate the high precision projection transformation matrix to avoid unnecessary feature detection and matching,which can greatly accelerate the speed of image stitching.Finally,the evaluation procedure is done according to the quality assessment method of stitched images,which can accurately reflect the real quality of stitched image and the performance of the proposed algorithm.Experimental results show that a great decrease of stitching time comparing to the current fast image stitching method has obtained,and the visual effects of the stitched images are good,which demonstrate the feasibility and effectiveness of our approach.