Section: Overall Objectives

Restoration, super-resolution, editing

Depending on the application and the type of content, various issues are being addressed, such as restoring the data to cope with limitations of sensors or to recover the signal from compression artefacts. The design of efficient methods for enhancing the data resolution in spatial, temporal or angular (in the case of light fields) dimensions is also part of the project-team objectives to cope with limitations of the visual sensors or for anti-aliased image-based rendering, e.g. from light fields. View synthesis remains a difficult ill-posed problem related to angular super-resolution of multi-view data and light fields which keeps evolving with in particular the use of deep learning techniques. Estimating the scene geometry and the scene flow for dynamic scenes is a critical step of the above processing problems. Editing the content, or restoring disoccluded areas in multi-view processing calls for efficient içnpainting techniques.