Study On Color Video Compression Technique Based On Vector Quantization

Video compression has been the key technology for transmitting massive amount of real-time video data via limited bandwidth channels. Video-compression has assumed an increasingly important role in many applications such as videoconference, 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. For example, unmanned vehicles are employed more and more frequently for a variety of scientific and commercial undersea applications. However, the elimination of the tether requires the capability to compress a massive amount of live video data to meet the bandwidth limitations of acoustic telemetry. Although video-coding technology has been developed rapidly, the existing coding methods still cannot meet the requirements of video transmission at very low bit rate. Therefore, it is necessary to investigate and develop more effective video coding techniques to satisfy the requirement of real-time video transmission in some practical applications such as ocean and space explorations. Focusing on the above objective, this paper studies the color video coding methods based on vector quantization (VQ) in discrete cosine transform (DCT) domain and discrete wavelet transform (DWT) domain. The following achievements have been accomplished.(1) Color Video Compression in DCT domain Based on VQWe present an efficient VQ-based color video coding method in DCT domain. In order to remove the inherent correlation among the three components of color image, the RGB space is first transformed to the YUV space. For the U and V chrominance components, sub-sampling operations and average computation are performed to remove the visual redundancy. On the other hand, for the luminance Y image, we first partition it into 8×8 blocks which are not overlapped, and then apply DCT to these blocks, thus forming the transformed vectors. Based on HVS, the transformed vectors are then automatically classified and truncated. Finally, for each type of the vectors, the global codebook is designed accordingly, which can adaptively update and replace the code words of the cookbook with the change of scene according to the inter-frame correlation and the using frequency of code words.(2) Color Video Compression in DWT domain Based on VQThis paper also presents an efficient VQ based color video coding method in DWT domain. The input frames are first transformed from the RGB space to the YUV space, and then the three components are decomposed using the integer-to-integer versions of lifting form of Daub5/3. The transformed Y component is coded by employing classified tree-structure vectors and global codebook scheme. The global codebook designed can adaptively update and replace the code words of the cookbook with the change of scene according to the inter-frame correlation and the frequency of code words. The coding of chrominance components of U and V images is achieved by throw off all the high frequency sub-bands based on HVS.The experimental results show that while maintaining the acceptable quality of the reconstructed images, the both proposed algorithms can provide high coding efficiency, especially applying to such scenarios as videoconference and video observation of underwater scenes.