Super-Resolução para Síntese de Vistas em Ambientes de Resolução Mista

Mixed resolution stereoscopic video is a popular format aimed at 3D content consumption. Nevertheless, resolution differences among views can pose significant drawbacks if adapting such a format to other multiview functionalities, such as view synthesis. In this paper we adjust standard view synthesis methodology to mixed resolution conditions. Operating within a mixed resolution, multi-view-plus-depth scenario, we then propose an improved super resolution method capable of enhancing lower-resolution views with high-frequency content available from neighboring full-resolution views. Results on publicly available data sets show that both strategies, adaptation of synthesis to asymmetry and direct enhancement of lowerresolution input views, can improve subjective and objective quality of synthesized views.