Are you an EPFL student looking for a semester project?
Work with us on data science and visualisation projects, and deploy your project as an app on top of Graph Search.
MPEG High Efficient Video Coding (HEVC) is likely to emerge as the video coding standard for HD and Ultra-HD TV resolutions. The two elements that push HEVC beyond the previous standards are a higher compression efficiency of about a factor of two, and the introduction of new coding tools, tiles and wavefront that are intended to ease the largely increased encoding complexity particularly for Ultra HD resolutions such as 4K and 8K. However, for HEVC decoder implementations, the achievement of the desired performance on massive parallel platforms cannot rely on the use of such optional (not enforced by MPEG profiles) tools. This paper reports results about the intrinsic parallelism of compliant HEVC decoding algorithms obtained by analyzing a dataflow implementation written using the standard language specified in ISO/IEC 23001-4 and structured attempting to maximize the algorithmic potential parallelism. The experimental results show what is the parallelism achieved by different dataflow architectures and how it can be further combined with the parallelism achieved by relying on tiles and wavefront, whenever they would be available, for porting a compliant HEVC decoder on massive parallel many-core platforms. © 2014 IEEE.
Paolo Ienne, Kubilay Atasu, Radu Ioan Stoica, Jovan Blanusa