Research On Image Stitching Algorithms Based On Novel Energy Function And Adaptive Smooth-transition

Image stitching is an important research topic in the field of computer vision.Image stitching is widely used in many fields,such as remote sensing imaging,medical imaging,underwater mapping,virtual reality,etc.In recent years,scholars have proposed many image stitching methods with superior performance.Given two or more single images with overlapping areas of narrow field of view,image stitching combines them into a whole image of wide field of view.The goal of image stitching is to generate visually reasonable stitching images.The problems are artifacts in overlapping region and distortion from overlapping region to non-overlapping region.This paper is dedicated to solving the above problems and proposes a robust image stitching algorithm.The main research work of this paper is as follows:(1)Aiming at the problem that the traditional seam-cutting method cannot effectively avoid the inconsistent regions of color difference,an artifactfree image stitching method based on the novel normed seam-cutting energy function is proposed.This method can effectively distinguish the inconsistent regions of color difference,so as to find a more suitable seam for the original problem.Experiments show that the proposed method can achieve better stitching results than traditional seam-cutting methods.Compared with other existing seam-cutting methods,the proposed method can obtain the same or even more satisfactory stitching results without requiring iteration or complicated saliency calculations.(2)Aiming at the problem of transition distortion in existing robust elastic deformation method,an adaptive smooth-transition stitching method is proposed.This method can eliminate the distortion of the transition region and realize a smooth-transition from the overlapping region to the non-overlapping region.Qualitative and quantitative experiments on public and our data sets show that the proposed method is better than the original method and improves the quality of stitching.(3)The two methods proposed in this paper are highly compatible.In many cases,only depending on one of these methods may not get the ideal result.More often,they are combined with each other to get better stitching results.