Technology Research, Computer Vision-based Image Stitching

The purpose of image mosaics is to synthesize small-sized overlapping region images of the same scene into a large-size image,with large field angle,even a panoramic image of some large scene.The technology of image mosaics has been widely adapted in the fields of virtual reality,medical image processing,remote sensing technology,and military,which is also becoming a research focus of computer vision.This paper,on the deep study of traditional approaches of image mosaics,emphatically examined image registration,the key step of image mosaics,and summed up advantages and disadvantages of image registration's methods.On such a base,a new approach of image mosaics is constructed,that is to say,a correspinding relation of coordinate points between image space and object space with camera calibration.According to the consistency priciple of object space coordinate points,homonymy points of adjacent overlapping region images is related to realize the feature matching among adjacent images and image mosaics of adjacent images.This approach greatly avoids the common disadvantages of large calculation, consuming time,and mistake matching;on the other hand,it depends on the results of camera calibration,whose precise degree directly affects image mosaics.Therefore,this paper has selected mature and reliable methods to do deep and detailed research on the technology of camera calibration and finally get the camera parameters.The result of the research is proved to be effective and reliable.