The Key Technology Of Panorama Mosaic

Image mosaic is a technology to solve the problem of the inadequacy of singlecamera’s view perspective. It has brought a lot of help to people’s daily lives andscientific research. With the development of digital image processing andmicroelectronics technologies, panoramic images have been widely used in virtualreality, computer vision, aerospace application, military application, medical imageanalysis, remote sensing image processing and other fields. The method of stitchingpanoramic image has become a hot research field of virtual reality and imageprocessing.The registration of complex scenes, image distortion correction, the mosaic oflarge exposure differences images and dynamic scene image mosaic are the difficultiesof image mosaic. There is no holistic image mosaic method for the most commonlyused images. This thesis focuses on distortion correction, image registration, imagefusion and ghosting eliminating in dynamic scenes. The main research andachievements present as follows:(1) This thesis studies image geometric distortion correction methods, includingthe common image geometric transformation model, panoramic image projection modeland camera distortion correction. An image distortion correction method based ongeometric model is used to achieve a quick and effective correction of camerageometric distortion.(2) This thesis analysis the image registration method based on frequency domain,the image registration method based on gray scale and feature-based image registrationmethod. To get an effective feature detection method the Harris corners, scale invariantfeature transform and speeded up robust feature are compared. The feature pointspurification and estimation of model transformation parameters are depth researched.The registration of Hand-held camera image can get a good result by using the mehtodin this thesis.(3) The method to solve the problem of fusion ghosting caused by moving objectsin dynamic scene is studied. An improved optimal seam is proposed. The improved optimal seam reduces the impact of exposure differences and takes full advantage of thesimilarity adjacent pixels. It also reduces the amount of computation. Multi-resolutionfusion and Poisson fusion algorithm are deeply studied. The Poisson fusion solved byFourier transform improves its speed. A seamless and natural panorama is achieved byusing the image mosaic method in this thesis.