Research On Image Registration And Stitching Based On SIFT Feature

In recent years,image stitching becomes a research hotspot in computer vision,which has been widely applied in the fields of medical image processing,video detection,remote sensing,sea floor exploration,video compression and searching,three dimensional virtual scene structuring,national defense security,robot vision and so on.With the deepening of the application,the requirement of registration accuracy and time-consuming has been improved.Feature based Image registration and stitching has high registration accuracy and little influence from illumination and rotation,which is the most popular spot in image stitching research field.This paper will provide some improvement approaches to the existing feature based image registration and stitching solutions,aiming at improve the quality of image stitching.The main research works of this paper are as below:(1)The existing methods of lens distortion correction have been analyzed and compared,and the calibration using checkerboard has been realized.(2)Aiming at improving the image registration accuracy of image stitching,an image enhancement method based on image pyramid has been proposed.(3)The algorithms of image registration based on SIFT feature and PCA-SIFT have been studied and optimized via simplifying the descriptor of SIFT feature.(4)The methods of image fusion have been studied and analyzed.An improved method has been proposed and the result of image fusion has been optimized.Experimental results show that the method proposed by this paper can be used in all kinds of image stitching and video stitching scenario.