3D Scene Completion
By gelautz - Posted on November 24th, 2011
To create 2D or 3D novel views from a sparse set of input images has been a long-standing goal in computer vision. The range of applications is vast (computer games, 3D Photosynth browsing, 3D image manipulation, etc.), especially in the light of latest developments in the hardware domain (3D television, stereo point-and-shoot cameras). This proposal revisits this fundamental problem from a new perspective. One main research challenge for new view synthesis is to reconstruct both the depth and the colour of those pixels which have not been observed in any of the input images. We term this problem 3D scene completion. Our hypothesis is that by extracting and modelling physically aspects of the scene, such as geometry, light and camera, it is possible to improve on competing methods which operate purely in the image space.