Feature Increase And Terms Optimization Warp For Image Stitching

Image stitching is the process of merging a series of images with overlapping regions into a high-resolution image with a wider field of view.It has been widely used in panoramic stitching,automatic driving,virtual reality and other fields.It is a very important research direction in computer vision.With the rapid development of image processing technology,more and more scenes need high-resolution images with a broader vision.Image stitching technology emerged at the historic moment,which can meet the growing demand and has a far-reaching impact and positive role in the development of various application fields.Although there have been many achievements in the research of image stitching technology,there are still many problems to be solved,such as large parallax,low texture,etc.Feature extraction and correct matching are the basis of image stitching process.The processing results of existing methods for low-texture images are not satisfactory.They use SIFT+RANSAC pattern to extract and match feature points.However,it is difficult to obtain enough correct matching points in low-texture or repeated-texture areas,resulting in insufficient matching points in overlapping regions,and thus leading to incorrect estimation of image warping model.Whether the matching points are correct and the number of matching points directly affect the final stitching result.In addition,when low texture and large parallax coexist in overlapping regions,alignment and distortion often restrict each other and are difficult to balance.Accurate alignment is usually accompanied by projection distortion and perspective distortion.Aiming at the above problems,this paper proposes a novel and flexible method to solve the above problems through feature correspondences increase and hybrid terms optimization warp modules.Feature correspondences increase module is a feature extraction and matching method based on GMS algorithm to ensure that enough correct matching points can be obtained in overlapping regions with low texture to eliminate misalignment;Hybrid terms optimization warp module is to estimate the optimal warp model by adjusting the parameters of each optimization term based on the global homography transformation and the global similarity transformation.The two modules can be separately added to the image stitching process to achieve corresponding functions and solve corresponding problems.The combination of the two can effectively balance alignment and distortion while ensuring alignment.Use the above two modules to improve and innovate the traditional image stitching algorithm,and provide certain ideas and basis for future research.This study mainly compares the single image warping methods in traditional image stitching algorithms.Evaluated on some public image stitching data and some image data collected by ourselves,and compared with the state-of-the-art,the results demonstrate that the alignment accuracy of our proposed method is better than the existing mainstream methods on images with low texture areas in the overlapping regions,and the stitching results have less perspective distortion and projection distortion.