3D object database: Multi-viewpoint image coding & view synthesis

Abstract

Image-based rendering is a powerful and promising approach for 3D object representation. This approach considers 3D object or scene as a collection of images called key frames taken from the reference viewpoints and generates arbitrary views of the object using these key frames. In this paper, we present an object-based encoding scheme for the multi-view sequences (key frames) of 3D object and a view synthesis algorithm for predicting new views.