Support vector regression for rate prediction in distributed video coding

Abstract

A recent trend in video coding is toward the low-complexity distributed techniques which provide an adaptive way to distribute the computational complexity among the encoder(s) and the decoder. One of the well-known architectures for Distributed Video Coding (DVC) is the Stanford architecture. This structure imposes the presence of a feedback channel from the decoder to the encoder, causing the codec to be impractical in some applications. However, the feedback channel can be eliminated if the encoder estimates the desirable bitrate during the encoding process. In this paper, we introduce a new method for bitrate estimation using ν-SVM regression with the aid of a novel set of features. We also present a Hybrid coding mode which reduces the computational complexity in a conventional Stanford codec. The presented methods are evaluated using three different video sequences. The simulation results for the feedback-free method show that the average decrease in PSNR of the decoded frames is 0.7 dB for low-motion and up to 3 dB for high motion videos. While preserving the same PSNR quality as the conventional Stanford codec, the Hybrid method reduces the computational complexity by a factor 3, thereby speeding up the decoding process by imposing an overhead bitrate of 3 kb/s.