Automatic conversion of two-dimensional video into stereoscopic video
Manbae Kim, Sanghoon Park, Hak‐Soo Kim, Artem Ignatov · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2005
Stereoscopic conversion of two-dimensional (2-D) video is considered based upon image motions. In general, a stereoscopic camera with two imaging sensors is required for stereoscopic video, while the stereoscopic conversion directly converts 2-D video to 3-D stereoscopic video. Image motions that are needed to generate a stereoscopic image are computed by any motion estimation algorithms. However, it is well know that motion vectors are generally far from a true motion, thereby posing a difficulty in the utilization of such data. To overcome the aforementioned problem, each image frame is classified into either a primary frame (PF) or a secondary frame (SF) depending upon the reliability of motion vectors. Furthermore, the stereoscopic generation method of the PF is presented. For the performance evaluation of our proposed method, we apply it to five test sequences and measure the accuracy of our proposed method. Experimental results show that our proposed method has the accuracy more than about 90 percent in terms of the detection ratio of primary frames. Furthermore, a variety of test sequences encoded in MPEG are directly applied and support that our proposed method is well designed.