Panorama Generation Technology Research Based On The Feature Points

Panoramic image is a kind of virtual reality technology based on image mapping technology to produce picture truthfully. It is one of the important research directions in the technology of Virtual Reality. Panorama Stitching is the number of the local scene images through cameras or cameras from general access to the true image registration algorithm splicing technology and integration into a wide viewing angle of the image smoothing. As Panorama broaden the perspective of the people in the scientific research and engineering projects are highly practical, widely used in satellite photos, handling medical image processing, virtual scene display, and other fields.This paper studies some Panorama generation algorithms at home and abroad, described key technologies and research status of Panorama Generation Technology. Introducing Panorama generation related technologies from the core technology of Panorama Stitching: image registration techniques and the integration of images commonly used algorithm. Summed up by an ordinary camera access to real images, the splicing of common benchmarks surface selection, interpolation technology, the perspective transformation matrix for linear and nonlinear methods. Based on the analysis of the existing algorithm, in view of characteristics with real image stitching, we achieved an automatic panoramic image stitching algorithm. Image Stitching take into account the robustness of requirements, the speed of registration, the quality of image after image fusion, and various issues such as this. Thinking of this algorithm is: First, pretreatment on multiple images, and extract Harris feature points of each image; filter through the inter-related measure of the metric error with narrow the search scope through the local regional registration; proposed by the operator characteristics of the classification error with RANSAC purification method; recursive 8 parameter transformation matrix using coordinates from selected based on the error function LM nonlinear optimization algorithm; Use of progressive methods in and out of images and the integration of its reunification as a cylindrical projection as the space plane. Experiment proved that the algorithm can obtain satisfactory results of the panorama.