Low bit rate video compression algorithm using 3-D decomposition

A low bit rate three dimensional decomposition algorithm for video compression with simple computational complexity is presented. This algorithm performs the temporal decomposition of a video sequence in a more efficient way by using 4-tap short symmetric kernel filter (Haar filters) with decimation factor of 4:1 instead of 2:1 used in the classical 3D-wavelet algorithms. The pyramid coding decomposition concept is then used for the spatial domain. The main goal of this paper is to design a simple encoding algorithm with a very high performance. Local adaptive vector quantization (LAVQ) is used to encode some of the spatial subbands. The codebook of LAVQ is simple and robust to the motion which occurs in the video sequences and which seldom captures from a single training sequence. The other subbands are encoded using the very simple coding algorithm called absolute moment block truncation code (AMBTC). The AMBTC is used for the bands that are highly correlated and with no motion or sparks information. Experimental results demonstrate a significant improvement in performance over earlier published algorithms. It gives an excellent image quality at PSNR on the average of 36.9 dB and at a low bit rate of 0.13 bpp for Miss America sequence. Finally, the coding and decoding of the proposed algorithm are of comparable and relatively low complexity and is well suited to parallel implementation.