A Novel Approach for Video Quantization Using the Spatiotemporal Frequency Characteristics of the Human Visual System

The temporal limitations of the human visual system (HVS) are usually not fully explored in the quantization stage of today’s video encoders. In this paper we present a novel method to improve the compression efficiency of popular video compression algorithms in which the temporal frequency response characteristics of the HVS are taken into account while building macroblockspecific quantization matrices for non-intra-coded (i.e., ‘P’ or ‘B’) frames. These quantization matrices help removing redundant information from the video sequence and, consequently, increase compression efficiency. Results from our experiments show that the proposed approach can achieve improved bit savings over a standard encoder without sacrificing the perceived visual quality of the compressed sequence.