Stereoscopic to Multi-view Video Conversion System FPGA Implementation

摘要

Today, the stereoscopic 3D movie is already mainstream, and almost all new devices support stereoscopic 3D movie. The content generation and the necessity to wear glasses, however, hinder the broader acceptance of the technology. Audiences who have poor vision prefer to the glasses free stereoscopic movie. Multi-view video enables a glasses free perception of stereoscopic 3D movie. To support this, the paper proposes a novel and simplified hardware system to generate multi-view video from stereo video by image domain warping. Additionally, we elaborate the memory management and the algorithm of each module. Relied on the sparse matched feature points and saliency image extracted from the stereoscopic video, the system has already realized the conversion from stereoscopic video to multi-view video real-time based on FPGA.