Performed 3D reconstruction of an environment by using a pair of omnidirectional images. The process involved generating a virtual camera with a cylindrical image plane.
Applications for recovering 3D information using single viewpoint imaging systems require the exact location of the viewing point and a distortion free image plane.
By employing a calibrated omnidirectional camera, I implemented a new algorithm that would reconstruct the 3D structure based on the idea that color histograms of projections coming from two optical centers would match on the surface of the object.
Based on this method, a rough depth map of the surrounding environment could be generated. In the experimentation, I have applied this method to indoor environments obtaining promising results.
Comments