Real-time 3DTV system for autostereoscopic displays

This Paper focuses on the real-time implementation of 3DTV system, of which the core algorithms are stereo matching and multi-view rendering. The former algorithm is specially designed for the generation of high-quality disparity map on the graphics processing unit (GPU) through CUDA (Compute Unified Device Architecture) API, it is structured in a way that exposes as much data parallelism as possible and the power of shared memory and data parallel programming in GPU is exploited. The latter algorithm is specially designed in a new hardware architecture consisting of dual parallel DIBR, merging, hole filling, and synthesis, for the generation of multi-view images on the FPGA. We make the communication between PC and FPGA across network. Experiment results show that, for each 450 × 375 video with 60 disparity levels, stereo matching achieve about 16 frames per second on an NVIDIA Tesla C2050 graphics card and speed can be up to 30 frames per second after doing downsapling. Multi-view rendering can achieve 30 frames per second for processing full HD (1920 × 1080) video on Xilinx Virtex IV XC4VSX25.