Machine Learning-Driven Efficient Video Compression

Department

Computer Science Department

Main research activities/topics/projects

With video accounting for more than 50% of the Internet traffic, standardized codecs meet high compression efficiency at the cost of high computational complexity. Efficient codec implementations are pivotal for practical codec deployment, and recently machine learning has shown the potential for reducing the encoding complexity. The EIDOS [1] group of University of Torino, computer science dept., has a proven track record in both learning based image processing and image compression. The group has recently established connections with Universidad Federal de Pelotas, factorizing skills with an ongoing resource exchange program. The group aims at strengthening ties by hosting a PhD student in the domain of efficient learning based video compression. The project lies at the boundary between machine learning, video coding and efficient hardware implementations and the sought skills of the PhD student include image feature extraction, data analysis, machine-learning algorithms design (including tree-based algorithms, support vector machines, deep neural networks, etc), programing, hardware implementations and paper writing. The candidate must have good communication skills on either English OR Italian. Reading and writing skills in English are necessary. The research is associated to mathematics and computer sciences fields.

Special entry requirements

Python programing skills
Machine-learning knowledge
Video coding knowledge

Digital image processing

Duration in months (min-max)

PhD sandwich: 6-12