Abstract—The paper presents an efficient algorithm for approximating a real world 3D scene captured by a camera by a computer generated 3D graphics environment. First, to extract the 3D edge from an available 2D stereo image pair, straight edges are detected from both the images and a correspondence is established between them through the fundamental matrix based matching scheme. In the next step, we triangulate the end points of the common region of this matched edge pair to find the end points of a 3D straight edge. Using these, we locate the existence of planes between coplanar 3D lines, devising a coplanarity score for pairs of 3D lines and finally eliminating non-physical 3D planes by SIFT based analysis on the basis of discrepancies in depth of SIFT features in those 3D planes.

Index Terms—image to graphics conversion, 3D edges, pseudo-plane, epipolar

Cite: Vishwmitra S. Bhadouria, Disha Prakash, and Venkatesh K. Subramanian, "Stereo Image to Graphics Conversion for Indoor Environments," Journal of Image and Graphics, Vol. 2, No. 1, pp. 83-88, June 2014. doi: 10.12720/joig.2.1.83-88