Research On Image Stitching Algorithm Based On Thin Plate Spline

Image stitching is a technique of splicing two or more images with overlapping regions,and is also a key technology related to image processing in the field of computer vision.Image stitching is critical for generating high-resolution panoramic images as well as video.In recent years,with the introduction of some excellent image stitching techniques,related problems have gradually become a hot spot in the field of computer graphics image processing.The most direct effect of image stitching is to increase the field of view of the image.In order to obtain a large field of view image,there are two feasible solutions.One is to use a wide-angle lens for shooting,and the other is to use a normal angle of view lens for multiple shots.splice.Although the wideangle lens can capture a large field of view image at one time,its expensive equipment price can only be used in specific situations.The limitation of image stitching is the stitching effect and calculation speed.In this paper,the image stitching process is divided into two parts,namely registration and fusion.The common algorithms for the relevant steps are introduced respectively,and the image alignment process is studied in detail.In order to align images with parallax tolerance,some superb image stitching techniques have been proposed in recent years.One of the methods is the Elastic Local Alignment(ELA)algorithm based on the approximating Thin-Plate Splines,which greatly improves the speed of image stitching while ensuring the alignment accuracy.However,in some cases with large parallax,it is easy to produce visible distortion at the edge of the overlapping region,making the result look unnatural.The main work of this paper is as follows:(1)The current international status of image stitching problem is studied,and the representative method of Elastic Local Alignment(ELA)is analyzed.(2)The image alignment algorithm based on the approximating Thin-Plate Splines is studied,and a weight optimization algorithm based on deformation function is proposed.The algorithm effectively alleviates the image distortion phenomenon near the overlapping area of the stitched image,and further improves the visual effect of the stitched image.(3)In view of the actual situation that the current image resolution is getting higher and higher,an algorithm for adaptive sparse matching feature point accelerated image stitching is proposed.The algorithm improves the speed of image stitching and ensures the alignment accuracy of stitched image.In the experiment,the 8 megapixel image was stitched,and the image deformation alignment time was shortened from the original 20 seconds to 3 seconds.