3D coordinate extraction from single 2D indoor image

Indoor buildings as part of man made contains many information of linier perspective. Human has great tendency to build rooms with plane between wall to wall, and wall to floor to be perpendicular each other. This condition gives many advantages so we can use those information to reconstruct 2D imaged room back into 3D space. This paper is aimed to automatically extract and reconstruct room into 3D space from single 2D image only. Our method is begin by localizing ground segment from image, then followed by determining which border of ground should be assumed as wall. Once ground and wall border is identified, we calculate corresponding 3D coordinate of ground and wall using homography and cross ratio. The result of proposed algorithm shows its ability to extract and reconstruct 3D rooms based only on single 2D Image.