Reconstruction of Polyhedral Objects by Correction of Images and Using Shape from Shading Approach

This paper presents a robust and simple approach for reconstruction of 3D polyhedral objects from a single shaded image of a polyhedral scene. Edges and vertices in image are found by detecting the edges from the image. The shape of polyhedral object is considered as the composition of planar polygonal faces. The consistency of picture is checked by the rank condition of shape matrix, computed from image. In case of incorrect line drawing, vertices of object image are corrected by using the singular value decomposition method, keeping the shape of object preserved. Normals of each face are computed by exploiting the shading information contained in the image. The shading in the image is considered as an approximation of Lambertian reflectance map. The local minimization technique is used for the computation of normals of each face of object from image intensity, which varies on change of plane. From these normals depth value is calculated for each vertices of polyhedral objects. The algorithm is tested on various polyhedral images and the results are shown.