Sean Cassidy

Abstract

Video is an increasingly ubiquitous part of our lives. Fast and efficient video codecs are necessary to satisfy the increasing demand for video on the web and mobile devices. However, open standards and patent grants are paramount to the adoption of video codecs across different platforms and browsers. Google On2 released VP8 in May 2010 to compete with H.264, the current standard of video codecs, complete with source code, specification and a perpetual patent grant. As the amount of video being created every day is growing rapidly, the decision of which codec to encode this video with is paramount; if a low quality codec or a restrictively licensed codec is used, the video recorded might be of little to no use. We sought to study VP8 and its quality versus its resource consumption compared to H.264 -- the most popular current video codec -- so that reader may make an informed decision for themselves or for their organizations about whether to use H.264 or VP8, or something else entirely. We examined VP8 in detail, compared its theoretical complexity to H.264 and measured the efficiency of its current implementation. VP8 shares many facets of its design with H.264 and other Discrete Cosine Transform (DCT) based video codecs. However, VP8 is both simpler and less feature rich than H.264, which may allow for rapid hardware and software implementations. As it was designed for the Internet and newer mobile devices, it contains fewer legacy features, such as interlacing, than H.264 supports. To perform quality measurements, the open source VP8 implementation libvpx was used. This is the reference implementation. For H.264, the open source H.264 encoder x264 was used. This encoder has very high performance, and is often rated at the top of its field in efficiency. The JM reference encoder was used to establish a baseline quality for H.264. Our findings indicate that VP8 performs very well at low bitrates, at resolutions at and below CIF. VP8 may be able to successfully displace H.264 Baseline in the mobile streaming video domain. It offers higher quality at a lower bitrate for low resolution images due to its high performing entropy coder and non-contiguous macroblock segmentation. At higher resolutions, VP8 still outperforms H.264 Baseline, but H.264 High profile leads. At HD resolution (720p and above), H.264 is significantly better than VP8 due to its superior motion estimation and adaptive coding. There is little significant difference between the intra-coding performance between H.264 and VP8. VP8's in-loop deblocking filter outperforms H.264's version. H.264's inter-coding, with full support for B frames and weighting outperforms VP8's alternate reference scheme, although this may improve in the future. On average, VP8's feature set is less complex than H.264's equivalents, which, along with its open source implementation, may spur development in the future. These findings indicate that VP8 has strong fundamentals when compared with H.264, but that it lacks optimization and maturity. It will likely improve as engineers optimize VP8's reference implementation, or when a competing implementation is developed. We recommend several areas that the VP8 developers should focus on in the future.

Library of Congress Subject Headings

Digital video--Standards; Video compression--Standards; Coding theory

Publication Date

11-1-2011

Document Type

Thesis

Department, Program, or Center

Computer Engineering (KGCOE)

Advisor

Kwasinski, Andres

Comments

Note: imported from RIT’s Digital Media Library running on DSpace to RIT Scholar Works. Physical copy available through RIT's The Wallace Library at: TK6680.5 .C37 2011

Campus

RIT – Main Campus

Download

Share

COinS