A low complexity picture-in-picture transcoder for video-on-demand

Abstract

The paper presents a low complexity transcoding (LCT) approach with picture-in-picture (PIP) functionality based on H.264/AVC standard for video-on-demand (VOD) application. To address the high complexity of H.264 decoding and encoding tools, the novel and fast LCT approach composes a PIP frame in the transform domain based on an auxiliary bitstream. The auxiliary bitstream is generated along with the primary bitstream to provide the syntax elements including the transform coefficients and motion information for a rapid frame composition. To balance the transcoding speed and the rate-distortion (R-D) performance, the PIP transcoding with the auxiliary bitstream is applied to all inter-coded frames and the partially re-encoding transcoding is used for the infra-coded frames. By removing the high complexity modules including spatial interpolation, motion compensation, in-the-loop deblocking filter and inverse spatial prediction from the transcoding process of all inter-coded frames, the LCT approach can significantly increase the transcoding throughput with very low memory requirement. Our experimental results show that the R-D performance of the LCT scheme is almost identical to that of the cascaded re-encoding approach, which can provide the theoretical best quality with highest complexity. In addition, the LCT approach that allows users to select the location and size of the foreground frame is applicable to the applications with interactive PIP service.