Non-predictive multi-scaled based MLVQ video coding for robust transmission over mobile channels

A new non-predictive video codec for mobile applications is presented. The scheme omits the prediction step in the temporal axis and increases robustness of its transmission in the mobile channel. The same subbands from each frame in the group of frames (GOP) are joined together to exploit their spatial-temporal redundancies. The significant vectors of the joined subbands within the GOP are quantised using a novel multi-stage lattice vector quantisation. This process reduces quantisation errors and enhances the reconstructed frame quality. Experimental results are shown to be significantly better than H.263 and comparable to the current H.264 standards in an erroneous hilly terrain mobile environment using the TETRA channel simulator for some test video sequences.