An algorithm and architecture for video compression.
|Title:||An algorithm and architecture for video compression.|
|Authors:||Idris, Fayez M.|
|Abstract:||In this thesis, we present a new frame adaptive vector quantization technique and an architecture for real-time video compression. Video compression is becoming increasingly important with several applications. There are two kinds of redundancies in a video sequence namely, spatial and temporal. Vector quantization (VQ) is an efficient technique for exploiting the spatial correlation. The temporal redundancies are usually removed by using motion estimation/compensation techniques. The coding performance of VQ may be improved by employing adaptive techniques at the expense of increases in computational complexity. We propose a new technique for video compression using adaptive VQ (VC-FAVQ). This technique exploits the interframe as well as intraframe correlations in order to reduce the bit rate. In addition, a dynamic self organized codebook is used to track the local statistics from frame to frame. Computer simulations using standard CCITT video sequences were performed. Simulation results demonstrate the superior coding performance of VC-FAVQ. We note that both VQ and motion estimation algorithms are essentially template matching operations. However, they are compute intensive necessitating the use of special purpose architectures for real-time implementation. Associative memories are efficient for template matching in parallel. We propose an unified associative memory architecture for real implementation of VC-FAVQ. This architecture is based on a novel storage concept where image data is stored by association rather than by contents. The architecture has the advantages of simplicity, partitionability and modularity and is hence suitable for VLSI implementation.|
|Collection||Thèses, 1910 - 2010 // Theses, 1910 - 2010|