DVC coding and CUDA decoding acceleration scheme for network multi-view stereoscopic TV

Network multi-view stereoscopic TV remains an unfilled blank, because multi-view stereoscopic TV is characterized by high bit rate and low tolerance for bit error which are the restriction of the development of network multi-view stereoscopic TV. This paper aims to solve the problem by distributed video coding (DVC) and GPU acceleration. Distributed video coding is an effective approach of Wyner-Ziv video coding for multi-view stereoscopic video encoded by LDPC. The distributed video coding based on channel coding offers flexible partitioning of the complexity between the encoder and the decoder to achieve a low complexity encoder for some applications, which are limited on calculation capability, memory capability and power consumption. Therefore, the DVC is suitable for wireless video capturing device, camera and so forth which the computation ability, memory size, power consumption are limited. The DVC decrease the complexity of encoding process, but increase the complexity of decoding process. To guarantee the fluent of the multi-view stereoscopic TV play, GPU is employed to accelerate the decoding process. The experiment result shows that joint decoding of eight views to generate the side information is more effective than conventional decoding of single view. And the CUDA could accelerate the decoding process efficiently.