Deep neural networks have achieved unprecedented success in several tasks related to computer vision, such as recognition, classification, tracking, pose estimation.

Even when lower-level tasks are considered, NNs have considerably improved the state of the art (optical flow, super-resolution…).

Recently, NN-based methods have been proposed also for image and video compression, since they have the potential for a breaktrough in this field.
In this talk we will review some of the recently proposed image and video compression techniques using NNs, including some recent work carried out at the Multimedia team in Telecom-Paris.

A perspective on the most promising approaches will be given.

Short Bio:

Marco Cagnazzo obtained the Ph.D. degree from the Federico II university (Napoli) and the university of Nice-Sophia Antipolis (France) in 2005, and has achieved the “Habilitation à Diriger des Recherches” form Université Pierre et Marie Curie (Paris) in September 2013.

He is currently full professor at Telecom Paris within the Multimedia team, where he arrived as associate professor in 2008.

He is the responsible of the Multimedia team since December 2015. Marco Cagnazzo has (co-)supervised 12 PhD candidates and currently (co-)supervises 6.

His research mainly focuses on visual data communication, including immersive video processing and compression, hologram compression and transmission, semantic video compression, compression helped by learning-based methods.

Marco Cagnazzo is an Associate Editor for IEEE Open Journal on Circuits ans Systems and for  Elsevier Signal Processing: Image Communication an was an Associate Editor for IEEE Transactions on Circuits and Systems for Video Technology, IEEE Signal Processing Letters and Elsevier Signal Processing. Since 2017, he is a member of the IEEE SPS Technical committee on Image, Video and Multidimensional Signal Processing (IVMSP).