Forward and inverse 2-D DCT architectures targeting HDTV for H.264/AVC video compression standard

摘要

This paper presents the architecture and the VHDL design of the integer Two-Dimensional Discrete Cosine Transform (2-D DCT) used in the H.264/AVC codecs. The forward and inverse 2-D DCT architectures were designed and their synthesis results mapped to Altera FPGAs are presented. The 2-D DCT calculation is performed by exploring the separability property, in such way, each 2-D DCT architecture is divided in two 1-D DCT calculations that are joined through a transpose buffer. The 1-D DCT transforms implemented and herein described are multiplierless, hence optimized shift-add operations are used. The architectures have a dedicated pipeline, optimized to process one input data per clock cycle. These architectures are able to cope with H.264/AVC encoder or decoder requirements targeting High Definition Digital Television (HDTV), with 1920x1080 pixel/frame at 30 frames per second.