Video transcoding techniques for wireless video communications

The transmission of compressed video over channels with different capacities may require a reduction in bit rate if the transmission channel has a lower capacity than the capacity required by the video bit-stream, or when the channel capacity is changing over time. The process of converting a compressed video format into another compressed format is known as transcoding.; This thesis addresses the specific transcoding problem of dynamic bit-rate adaptation for transmission over low bandwidth wireless channels. Transmitting compressed video over lower bandwidth wireless channels require accurate and efficient rate-control schemes. In this thesis, we propose several techniques to improve transcoding performance. Based on our experimental results, we present an approximate linear bit allocation model and macroblock layer rate-control algorithm, which can achieve accurate transcoding bit-rate. By reusing useful statistics information from the incoming compressed video, the bit-rate of the transcoded video can be determined according to the video scene context.; Considering a specific bursty error wireless channel, we propose a solution which combines video transcoding and an ARQ protocol to transmit compressed video over this channel. In order to make sure that the end decoder can decode and play the transcoded video within the required end-to-end delay, we analyze the rate and buffer constraints of the transcoder and derive the conditions that have to be met by the transcoder. In order to test the proposed solution, we use a statistical channel model to simulate the wireless channel and use this model and channel observation to estimate the effective channel bandwidth, which will be fed back to the transcoder for better rate control.; In this thesis, we discuss two applications. For real time video communication over wireless channel, we propose an algorithm that determines the transcoding scaling factor considering end-to-end delay, buffer fullness and effective channel bandwidth. For pre-encoded video distribution over wireless channels, we propose an algorithm which can determine the transcoding bit budget based on end-to-end delay, effective bandwidth, and original video bit profile. The proposed algorithm outperforms H.263 TMN8 in terms of video quality and buffer behavior with the same computational requirements.