Panorama Automatically Generated Technical Studies

Panorama image has important application in Virtual-Reality and ComputerVision, and automatic panorama generation has become the hot point of the researchon image this years. So it is sense to find a new algorithm of automatic panoramageneration which has more precision and faster speed.This paper mentions past results of others' research, and describes almost all thetechniques used in the process of automatic panorama generation. Firstly, interestpoints are detected and matched to judge weather there is similar relationship betweentwo images. Considering that it has several disadvantages to use the parameters byexPerience, a new automatic method to detect the threshold of interest point is broughtinto use. Secondly we aim to find the possible position of all the images belong topanorama. Then two operating methods named serial method and parallel method arebrought forward, and they can both get the position of the messy images in the rough.Thirdly, this paper solves the problem of image registration in frequency domain,three parameters are calculated here. At last, some methods of image fusion witch cangenerate a panorama with no gap are introduces. In a word, this paper mainlyintroduces the algorithm of interest points detection, the algorithm of position surmise,the algorithm of image registration based on FFT and the algorithm of image fusionbased on tower frame.At the end of this paper, a prototype system for automatic panorama generationby Visual C++6.0 is introduced and implemented. Experiments have demonstratedthe efficiency of this algorithm.