Architecture Design of Context-Based Adaptive Variable-Length Coding for H.264/AVC

摘要

Context-based adaptive variable-length coding (CAVLC) is a new and important feature of the latest video coding standard, H.264/AVC. The direct VLSI implementation of CAVLC modified from the conventional run-length coding architecture will lead to low throughput and utilization. In this brief, an efficient CAVLC design is proposed. The main concept is the two-stage block pipelining scheme for parallel processing of two 4$times$4 blocks. When one block is processed by the scanning engine to collect the required symbols, its previous block is handled by the coding engine to translate symbols into bitstream. Our dual-block-pipelined architecture doubles the throughput and utilization of CAVLC at high bit rates. Moreover, a zero skipping technique is adopted to reduce up to 90% of cycles at low bit rates. Last but not least, Exp-Golomb coding for other general symbols and bitstream encapsulation for the network abstraction layer are integrated with CAVLC as a complete H.264/AVC baseline profile entropy coder. Simulation shows that our design is capable of real-time processing for 1920$times$1088 30-fps videos with 23.6 K logic gates at 100 MHz.