Low Complexity Motion Model Analysis for Distributed Video Coding

Mobile 2.0 is being considered as a natural extension of the Web 2.0 revolution to the mobile networks. One important concept in Web 2.0 is the real-time creation and uplink transmission of rich multimedia contents from individual clients to the Internet. However, due to power constraints of mobile handsets, it is difficult to turn this concept into practice in mobile ecosystems. To realize this vision, distributed video coding (DVC) is proposed in 2002 as a possible solution to the design of low complexity video encoders without sacrificing video quality. DVC relies on highly complex decoders to decode-and-restore high quality video output. However, so far, the coding quality of existing DVC codecs is no where near that of the traditional closed-loop codecs. In this paper, an optical flow-based low complexity motion model analysis process is proposed to aid the generation of side information for DVC decoders. Fast total least-squares analysis is performed on the video data at the encoder so that coding decision can be chosen to provide more useful hints for the generation of high quality side information at the decoder. As shown from the experiments, the proposed technique is very promising for improving the quality of DVC codecs and for mobile 2.0 applications.