A Video compression algorithm using zerotree wavelet and herarchical finite state vector quantization

In this thesis a video compression technique with comparable low bit rates is presented. The proposed encoder is similar to other motion compensated, block based dicrete cosine transform (DCT) video coding standards such as MPEG- 1/2, H.261/3 and MPEG-4/Verification Model (VM). The novel feature of this coding scheme is the combination of Zerotree Wavelet (ZTW) structure with the Hierarchical Finite State Vector Quantization (HFSVQ). In this study, performances of the popular block motion estimation algorithms are compared. The simulation results show that the New Diamond Search (NDS) performs better than the compared algorithms. Because of that as a block motion estimation algorithm, the NDS is used to track the local motion. The motion vectors of the macroblocks are Hufman encoded. The overlapping block motion compensation (OBMC) technique which is an advance scheme in the H.263 standard is used after NDS in order to reduce the block artifacts. A three level discrete wavelet transform (DWT) is used instead of DCT to remove the spatio-temporal correlation. It is seen that the DWT coefficients are Gaussian shape distributed in each level except the lowest frequency subband. Therefore the proposed encoder uses a linear quantizer for the lowest frequency subband and the nonlinear Lloyd-Max quantizer for the rest of the subbands. Since the ZTW structure consists of different size of wavelet blocks that contain similar information, then it is very suitable for these blocks to be coded by HFSVQ. The computer simulation of whole system is done by using the Object Oriented Borland C++ Builder software. The simulation results of the proposed encoder was obtained using three types of standard video sequences; QCIF AMyo which contains a few movements at the rate of 10 kb/s and 5 fr/s, QCIF Carphone which contains relatively more movements than Akiyo at the rate of 30 kb/s and 10 fr/s and QCIFCoastguard which contains more movements than the others at the rate of 48 kb/s and 7.5 ît/s. Average PSNR over the entire coded sequences show that the proposed ZTW-HFSVQ video compression technique achieves comparable performance over MPEG-4/VM, H.263, ZTE and EZW for both I and P frames. Keywords: Video compression, Zerotree Wavelet, Vector Quantization