A New Video Coder using Multiwavelets

Abstract

Video compression takes the advent of spatial, temporal and psycho-visual redundancies. These redundancies can be compensated efficiently by transform, motion estimation/compensation and quantization respectively. The existing video coding standards (MPEG) use block based discrete cosine transform (DCT). In the DCT the input image needs to be blocked. So correlation across the block boundaries is not eliminated resulting in noticeable and annoying blocking artifacts. This is overcome by using wavelets. For better performance in compression, filters used in wavelet transforms should have the property of orthogonality, symmetry, short support and higher approximation order. Due to implementation constraints scalar wavelets do not satisfy all these properties simultaneously. Also compression scheme based on wavelets is slow. A new class of wavelets called 'multiwavelets' which possess more than one scaling filters overcomes this problem. The objective of this paper is to develop and propose an efficient video compression scheme which provides better quality, high compression and also faster. This is achieved by using multiwavelets for transform, kite cross diamond search (KCDS) algorithm for block matching and 'novel scheme' for quantization. Novel scheme is a combination of set partitioning in hierarchical trees algorithm (SPIHT) and set partitioning in embedded block coding (SPECK). Hence the name hybrid coding