Adaptive Coding of High-resolution Panoramic Video Using Visual Saliency
16 June 2013
In this paper, we describe how visual saliency can be used in a panoramic content interactive delivery framework. We first introduce the interactive panoramic video delivery paradigm, which has recently emerged in applications requiring beyond Ultra-HD content interactive streaming to mobile devices. We then show that these interactive delivery solutions are often based on a segmentation of the video content into independent spatio-temporal tiles that can be separately encoded, decoded and recomposed based on user interactions. This enables parallel encoding and spatial interactive navigation into the content but also comes with new issues related to quality adaption over tiles under bandwidth constraints. We show that tile-level rate control is required for ensuring uniform quality when live encoding each panorama tile on different cores with minimal inter-core communication based on prior information.
