3D Stereoscopic Image Pairs by Depth-map Generation

This paper presents a new unsupervised technique aimed to generate stereoscopic views estimating depth information from a single input image. Using a single input image, vanishing lines/points are extracted using a few heuristics to generate an approximated depth map. The depth map is then used to generate stereo pairs. The overall method is well suited for real time application and works also on CFA (Colour Filtering Array) data acquired by consumer imaging devices. Experimental results on a large dataset are reported.