Abstract
Image based modeling and rendering from single view image is challenging as it requires a considerable amount of user interaction while modeling the objects. We present a novel and very efficient method of automatically modeling the planar surfaces in the single view perspective images based on the edges and intensity representation. By considering the length of the vertical edges, it is possible to estimate the depth approximately. By integrating the depth cueing method of representing the 3D objects with the length of the edges, we automate the process of rnodeling from single view images. The approximate 3D coordinates are extracted from the proposed approach and 3D wireframe is constructed. Then a perspective correction is applied on each surface to get the actual 3D coordinates. Presence of symmetric objects (like circle, square, etc) contributes for obtaining the fronto-parallel view of each surface. Convincing photo realistic models are produced after rendering using texture mapping and results are shown as 3D screen shots.