Video compression coding based on the improved 3D-SPIHT

In this paper, an improved 3D-SPIHT algorithm has been proposed based on a traditional hierarchical encoding of the spatial-temporal trees of wavelet coefficients. We first encode the coefficient of LLLn using Huffman coding, meanwhile divide the other 3D-wavlet transform coefficients into three groups of horizontal, vertical and diagonal directions, which can be encoded independently by the 3D-SPIHT algorithm. This procedure brings the added benefit of parallelization of the compression and decompression algorithms. Second we use the symbol array of fixed allocation of memory instead of the lists of LSP, LIP, LIS. Then a threshold is set respectively to each of seven high-pass subbands of the highest level decomposition and also the information of all descendants is saved orderly in an array. When taking the significant test to every coefficient every time, we can quickly get the significant coefficients only need to binary search to the array. As shown by simulations, the improved algorithm shows significant improvements of the coding efficiency and PSNR, reducing memory usage, while helping for the hardware design. It is true that the bit stream is the output of a fully embedded wavelet coder which renders it capable of delivering progressive buildup of fidelity and scalability in frame size and rate.