High Performance Video Coding with Massive Parallelism in GPUs and Approximate Computing (HPVC)
Type: International Project Project
Duration: from 2019 Apr 09 to 2022 May 31
Financed by: FCT-CAPES
Prime Contractor: Faculdade de Ciência da Universidade de Lisboa (University) - Lisboa, Portugal
This project aims to explore the complementary profiles of the Brazilian and Portuguese teams to increase the performance of video encoders based on the High Efficiency Video Coding (HEVC) standard and its extensions, as well as future standards under development. New heuristics will be developed to reduce complexity. The parallelism of the encoding algorithms will be explored through GPUs, and dedicated architectures will be designed using approximate computation. The project will focus on processing rate increase and energy consumption reduction, which are critical factors in the various application scenarios for image and video encoding. These factors should become even more critical with the advent of new standards such as the one under consideration in JVET, the AV1 from AOMedia and the JPEG Pleno from JPEG. The cooperative development of this project is based on the accomplishment of work missions in the two countries to guarantee the achievement of the proposed objectives.
Partnerships
- R - INESC-ID Lisboa (Other) - Lisboa, Portugal
- Universidade Federal de Pelotas (University) - Pelotas, Brazil
Principal Investigators
- Nuno Filipe Valentim Roma (HPCAS)