A Theoretical Model and Study of Weighted MCTF Residual Energy

This paper presents a theoretical model and study of the impact of the weighted motion-compensated residual energy on the coding efficiency of motion-compensated temporal filtering (MCTF) when MCTF update steps are performed at both encoder and decoder sides and when update steps are executed at the encoder side only. We observe that in the presence of the uncompensated quantization errors due to the MCTF open-loop structure, the decoder must attempt to recover the update weight used by the encoder and this non-linear transform may contribute to the possible magnification of the quantization errors. We apply our analysis to the joint scalable video model (JSVM) 2.0. It is demonstrated that performing the MCTF update step at the decoder side does not contribute significantly to the coding efficiency, except at high spatial-temporal resolutions with high bit rates. This further justifies theoretically the elimination of the MCTF update step as a normative tool for temporal scalability in the current H.264/AVC scalable extension development