Image-space Free-viewpoint Video
Stephan Würmlin, Edouard Lamboray, Michael Waschbüsch, Peter Kaufmann, Aljoša Smolić, Markus H. Groß · 2005
Image-space free-viewpoint video (FVV) is a framework for representation and coding of sparse multi-view video and subsequent re-rendering from arbitrary viewpoints. It is based on the fundamental concept of storing all information describing a scene’s visual appearance in multi-channel video images. Each pixel’s channels define different attributes of discrete point samples of observed surfaces. In this paper, we show how image-space FVV can be implemented with standard video coding tools and ready-available video coding methods can thus be reused. Besides, this representation has been adopted as extension of the MPEG-4 AFX standard. In addition to the representation, we also introduce a novel video coding technique that allows random access and progressive transmission. We evaluate this new codec in the context of FVV and compare it to state-of-the-art image and video codecs like MPEG-4 and JPEG-2000. We present results based on real-world data proving the suitability of the chosen approach.