Autostereoscopic 3D video generation from stereoscopic videos using FPGA and GPU

We introduce a semi-real time autostereoscopic 3D video generation system using FPGA hardware and GPU processor. Proposed system consists of two major blocks. The first block is a disparity estimation block that is implemented on FPGA. In this block, we extract a disparity map of each frame from inputted stereoscopic video through the video capture board by DP (dynamic programming)-based stereo matching. And post-processing is optionally done in order to improve the quality of the estimated disparity map. Estimated disparity map is transferred to the GPU through USB 3.0 and PCI-express interfaces. Original stereoscopic video is also transferred to the GPU. With these data, disparity-based view interpolation is performed in order to generate virtual multi-view videos. Each pixel of generated multi-view video is re-arranged on the autostereoscopic 3D display panel in order to provide 3D depth perception. The prototype of the proposed system generates 9-view qHD(960×540) video with about 20fps.