Multistage lattice vector quantization video coding with adaptive subband thresholding algorithm

Abstract

In this paper a new non-predictive wavelet based video coding scheme is presented. The scheme omits motion estimation/compensation process (prediction step) in the temporal axis. The video sequence is grouped into a number of frames. An adaptive thresholding algorithm is used to optimize the subband threshold settings, and use to identify the significance of the subband coefficients. The significant vectors of the same subband within the group frames are joined together in order to exploit their spatial-temporal redundancies. The significant vectors are quantized using a novel multistage lattice vector quantization (MLVQ). This process reduces quantization errors and enhances the reconstructed frame quality. Incremental results of the new non-predictive video coding scheme with adaptive thresholding algorithm are obtained with more than 1 dB. Also it is significantly better and than MPEG-2 with I frames only for several test sequences.