Ultra-High Efficiency Video Coding with Hierarchical Partitioning and Context Modelling

Project Title:

Ultra-High Efficiency Video Coding with Hierarchical Partitioning and Context Modelling

Supervisors:
Professor Manzur Murshed (Principal)
Professor Guojun Lu
Assoc Prof Shyh Wei Teng
Dr Adrian Shatte (ECR)

Contact person and email address:

Professor Manzur Murshed (manzur.murshed@federation.edu.au)

A brief description of the Project:
Video information content and complexity have been growing rapidly, due to 1) increasing display resolutions, where 8K capture is gaining importance; 2) frame rates up to 120 frames per second (fps) for regular full motion capture; and 3) multiple synchronised viewpoints and extra modalities, such as depth. Traditional block-based coding approach Full 360 degree capture and streaming are also now part of the modern media coding world. For example, an array of 128 views, each with 8K cameras operating at 120 frames/s produces 500 Gpels/s; even at a massive compression ratio of 500:1, the resulting bit-rate would far exceed the capacity of a 10Gbit/s optical channel. This project will focus on object-based coding where video frames are hierarchically partitioned into non-overlapping cuboids, roughly representing objects with hyper-rectangular bounding boxes, and space-time-view redundancies are exploited by adapting coding contexts in different cuboids.