On video SNR scalability

We compare two SNR scalable video codecs. The first codec (CODEC1) is a three-layer single-pass quantization algorithm based on H.263 and extends the work presented by Robers (see Master's Thesis Northwestern University, 1997) and Robers, Kondi and Katsaggelos (see Proc. SPIE Conf. on Visual Comm. and Image Proc., p.201-12, 1998). The second codec (CODEC2) implements three layer SNR scalability as described in Annex O of the H.263 standard by requantizing the DCT coefficients of the encoding error using finer quantizers than those used in the previous layer. By testing the two algorithms at various base and enhancement layer rates on both high and low motion sequences, the advantages and disadvantages of each codec are discussed and suitable applications are suggested for each one.