HEVC / H.265 Video Coding Standard Overview

Seminar
Friday, March 29, 2013
11:00am

Network bandwidth consumption is projected to accelerate as CE and mobile devices provisioned with network connectivity continue to evolve with diminishing silicon geometries and the advent of multi-core architectures. In particular, these technology advancements extend the benefits of lower power consumption and higher processing capabilities that enable consumers with the means to access or send video content via the network anywhere at any time. The need for superior video compression technology is deemed inevitable as video data traffic is anticipated to grow perpetually and account for a significant portion of the network bandwidth boom.High Efficiency Video Coding (HEVC) is the latest video coding specification jointly developed by ITU-T VCEG and ISO/IEC MPEG. It features a comprehensive suite of coding tools enabling an improvement factor of two in coding performance over the current video coding standard deemed most efficient, the incumbent - AVC. After a brief review of video coding fundamentals and terminology, the HEVC coding framework is contrasted to AVC at a high level. We highlight HEVC's key coding tools versus their AVC counterparts: Intra and Non-Intra modes, CABAC, and in-loop filtering methods. HEVC formally introduces Coding Units (CUs), Prediction Units (PUs), and Transform Units (TUs), each potentially in a respective different size. The Coding Tree Unit (CTU) replaces the former macroblock but can be sized in one of three constant square block sizes, two of which are larger. CUs are mapped via quadtree decomposition inside each CTU. PUs are mapped throughout the spatial span of each CU, either as one or four squares, or as one of six preconfigured sets of rectangular PUs. The residual signal in the CU that results after prediction is decorrelated by applying TUs via quadtree decomposition. The TUs need not be aligned to the PUs, thereby resulting in flexible mapping combinations within the CU, which coupled with the ability to adapt the size of the CUs within the larger CTU, lead to superior coding performance. HEVC features other welcomed enhancements, including means for extraction of decodable sub-sequences to provision applications with random access and playback modes such as fast or reverse playback. We compare complexity and discuss HEVC provisions for parallel processing. In closing, we touch on Profiles and Levels, and reflect on objective and subjective performance.

Speaker

Arturo A Rodriguez joined Cisco Systems via the Scientific Atlanta acquisition in 2006. He has worked on video coding, computer vision, graphics, embedded real-time computing, and interactive multimedia applications. He has 94 issued patents and others pending related to these areas of expertise. Prior to joining Scientific Atlanta, he worked at Kaleida Labs and IBM. He obtained the MS and Ph.D. in electrical engineering from Purdue University and the BSEE from the University of Florida. Arturo has served as Chair or Program Committee member of multimedia and video coding conferences, and served on the editorial board of transactional journals. He has published several papers and book chapters. He is an active member and contributor to video standards.