Abstract

The latest video compression standard is a joint effort between ITU and MPEG known as H.264/AVC. As with any video compression standard the H.264/AVC uses computationally intensive algorithms to maximize performance. During decompression these algorithms must be applied in real-time, processing 30 frames a second. This can be done in software, specialized hardware, or a combination of the two. Software solutions allow for maximum portability and ease of design, but General Purpose Processors (GPP) can not take full advantage of the parallelizable algorithms that the H.264 decoder is based upon. Specialized hardware solutions, on the other hand, allow concurrent data and instruction paths, but do not offer a high level of abstraction for cross platform development. Recent work by Xilinx has resulted in the advent of the MicroBlaze soft-processor that is a stand alone microcontroller built from an FPGA. The MicroBlaze provides a specialized hardware medium to run software on-chip with VHDL entities. The goal of this thesis was to model and simulate a software hardware hybrid H.264/AVC Baseline Profile decoder using VHDL and a soft-processor. It was proposed to divide all highly sequential calculations (run-length and CALVC decoding) and control data flow into software and perform the remaining calculations (prediction, inverse transform, inverse quantization, etc.) in hardware modules. The software runs on Xilinx' s MicroBlaze soft-processor and the hardware was designed using VHDL. A major advantage of soft-processors over GPP's, is that it hardware instantiations reside on-chip with the processor. The software and MicroBlaze soft-processor were simulated in a test bench and the results proved that the MicroBlaze could not handle the encoded bit-stream in real-time. For this reason the hardware interface and hardware decoder were never fully implemented. The scope of the thesis covers the H.264 Baseline Profile standard, MicroBlaze processor, the implemented software solution, and the proposed hardware counterpart.

Library of Congress Subject Headings

Decoders (Electronics); Video compression--Standards; VHDL (Computer hardware description language)

Publication Date

2005

Document Type

Thesis

Student Type

Graduate

Degree Name

Computer Engineering (MS)

Department, Program, or Center

Computer Engineering (KGCOE)

Advisor

Marcin Lukowiak

Advisor/Committee Member

Roy Czernikowski

Advisor/Committee Member

Kenneth Hsu

Campus

RIT – Main Campus

Share

COinS