A novel virtual frame synthesis approach for FVV

Free-view Video (FVV) is a new developing video system, which is intended to give the viewers more than one perspective of viewing point to enjoy the video content. Among the whole video processing chain of FVV framework, virtual frame synthesis plays a key role for FVV realization. In our previous work, we proposed a FVV encoding format based on Multi-view Video Coding (MVC). In this paper, we present a novel virtual frame synthesis approach based on Scale Invariant Feature Transform (SIFT) feature detecting and matching technique for FVV design and implementation. The novelty of the proposed approach is that no strict camera calibration is needed when the virtual frames are synthesized. Compared with other virtual frame synthesis approaches, our approach takes use of scale invariant features to better keep the detailed information in the frame. Another advantage of the approach is its robustness in processing different kinds of input images, which is illustrated in the experiment part in this paper.