Abstract

From vision models to cinema applications, and back

Marcelo Bertalmío
Universitat Pompeu Fabra

This talk will present an overview of the research done in our lab, using vision science models for perception and neural responses alongside variational methods to tackle several very challenging image processing problems in cinema production, post-production and exhibition. We'll remark the importance of model fine-tuning by cinema professionals, the limitations of Linear+Nonlinear cascades to model vision, and current opinions in vision science suggesting that a change of paradigm might be needed.