Visual Media Coding and Transmission

4.4 Rate-distortion Analysis of Motion-compensated Interpolation at the Decoder in Distributed Video Coding

(Portions reprinted, with permission, from “Rate-distortion analysis of motion-compensated interpolation at the decoder in distributed video coding”, M. Tagliasacchi, L. Frigerio, S. Tubaro, IEEE Signal Processing Letters, Vol. 14, No. 9, September 2007. © 2007 IEEE.)

The goal of this activity is to introduce a model that allows study of the coding efficiency of DVC-based coding schemes. The analysis is restricted to schemes that compute the side information at the decoder by performing motion-compensated interpolation, starting from two intra-coded key frames. Specifically, focus is only on the generation of the side information, neglecting other factors related to the channel coding tools that are typically used to replace conventional entropy coding. The proposed model is designed in two steps. First, for each Wyner–Ziv-coded frame, the displacement error variance introduced by motion-compensated interpolation is estimated. In fact, the true motion field is not directly available at the decoder, and it must be estimated introducing displacement estimation errors. Then the power spectral density of the motion-compensated prediction error is estimated to obtain the RD curves by inverse water-filling [22]. Armed with the proposed model, the tradeoff between motion-compensated interpolation accuracy and GOP size is investigated, in order to find the optimal GOP size for a ...