Pyramid Vector Quantization Of Video Subband With Dithering

Abstract

This paper describes the development of a low complexity and fixed-rate video compression scheme based on three-dimensional subband coding of video signals. The video codec first performs three-dimensional subband decomposition on a group of video frames, and then encode high frequency subbands with pyramid vector quantization and lowest tempo-spatial band with a DPCM coding in time and space. To improve the visual quality of reconstructed video, different types of subtractive and non-subtractive dithering of pyramid vector quantizers were experimented and its effectiveness was proved by a standard pair comparison subjective test. Coder complexity was reduced by using longer filters in the first level of spatial decomposition for better selectivity and coding gain and shorter filter in the second level of decomposition for lower complexity. Results at different low bit-rate (64, 128 and 384 Kbps) for several standard video sequences are reported and compared with ITU standard H.263.