Study On Video Compression Technique Based On Vector Quantization

The growth of the multimedia applications in Internet has necessitated the research of efficient video-compression techniques. Video-compression has assumed an increasingly important role in many applications such as video conference, video surveillance, and distant medical treatment and so on. Further, it is one of the key techniques for both satellite remote sensing and sea explorations. In the former case, it is a challenging task to transmit huge volume of image information with the limited bandwidth and satellite power. In the latter case, the elimination of tether of underwater robots also requires the capability to compress a massive amount of live video data to meet the bandwidth limitations of acoustic telemetry. Although the video compression technique has been investigated extensively and developed rapidly, it still can not meet the requirement of live transmission at very low bit rate. It is quite apparent that the video compression technique is not perfect so far. For instance, the existing techniques have not eliminated all the redundancies existed inherently in videos completely. So it is necessary to further study the characteristics of video and compression mechanism deeply and to achieve live transmission of videos at very low bit rate. Aiming at the above goal, this paper investigates the video coding approaches in frequency domain based on VQ. The major achievements are as follows:Firstly, we investigate the video coding method in DCT domain based on VQ. In order to reduce the time redundancy, the space redundancy, and the visual redundancy efficiently, this paper presents a global vector quantization algorithm based on transformed vector. The proposed vector construction scheme is able to reduce the dimension of vector and is capable of adapting to variation in environmental illumination. The adaptive clustering method employed in this paper can increase the searching speed during codebook construction. The global VQ can provide content adaptability for efficient video coding. Experiment results show that the proposed method algorithm is suitable for video-coding applications such as video conference and video surveillance at very low bit rate.Secondly, we also study the video coding method in discrete wavelet transform (DWT) domain based on VQ. To remove the blocking effect in block-based coding methods we apply the VQ method in DCT domain to DWT domain. Lifting scheme is adopted because of its high computing speed and fidelity. The code vectors are formed by wavelet coefficients organized in quad-tree manner. Finally, we conduct some preliminary experiments to evaluate the performance of the method.