Research And Implementation Of Dwt-based Video Codec System At Low Bit Rates

In image or video coding, discrete wavelet transform (DWT) compared to the discrete cosine transform (DCT) has scalability, free from blocky artifacts, and is very friendly to human visual system. In this paper, the newest video coding standard H.264 is analyzed as a template to reveal the basic principles and structure of video coding system, and researching JPEG2000 to find out how to use DWT to code still images. After that, a wavelet-based video coding framework is proposed for low bit-rate applications. Two-dimensional DWT at one level is carried out with each frame, which divides the original frame into two parts. Low-frequency part encoded by H.264, high-frequency part is separated into I frames and P frames. Because wavelet decomposition provides a multi-resolution representation of the video frames, which is most suitable for backward motion estimation, backward intra-frame prediction is applied to each P frame, and then use mathematical morphological thinning algorithm to modify. Other more, we enhance the thinned P-frame by motion compensation of I frame, so as to reduce error accumulation. On the other hand, we just code the super significant coefficients of I frames. An adaptive scanning pattern is used before run length coding, then Huffman coding of these super significant coefficients is applied.In addition, implementation and performance of the key modules in the system are studied by simulation experiments, including the selection of wavelet functions, backward prediction of significant coefficients, mathematical morphology thinning algorithm, entropy coding method, coding controller and so on,Experimental results reveal our coder outperforms the state-of-the-art H.264 standard at low bit rates, especially I frames, and the reconstruction video quality with strong capacity to maintain the texture detail, the outline of the image more clearly, and have greater subjective visual quality.