Abstract:Video panorama was generated by moving camera.Because of the temporal redundancy in video and misregistration in the large angle rotation,many existing algorithms perform poorly.To solve these challenging situation,this paper proposed an parallel synthesis approach to generating panorama which used texture feature to select key-frames.This method first extracted repeatable features from the selected key-frames using parallel threads,and then estimated the perspective motion model to align the adjacent key-frames by features.Employed bilinear blending to reduce the parallax effect in the overlapped region.Meanwhile,parallel threads programming improved the synthesis of speed.The experimental results show that the proposed method can effectively synthesize panorama with misregistration and parallax.

Parallel Generation for Video Panorama in Mobile Environment