Context adaptive variable length decoder of H.264

Abstract

This paper presents a software implementation of context adaptive variable length decoder (CAVLD) and context adaptive variable length coder (CAVLC). Efficient algorithms for coding and decoding of Coeff_Token, levels, total zeros, and run_before have been developed, which are the five vital syntax elements of CAVLC and CAVLD. The necessary blocks namely Integer Transform, quantization and their inverses are also coded. This implementation can process video sequences of any size. Experiments have been conducted to process video sequences up to High Definition 720p (1280×720 pixels) resolution and up to 30fps frame rate and also for high resolution pictures. The reconstructed pictures are of acceptable video quality with PSNR values greater than 34dB. The design can be conFig.d to different Qp values. The proposed work is coded using matlab and serves as a basis for FPGA implementation of H.264 Base line video decoder.