抽象的

Virtual Mirror Rendering and Recognition of 3DObjects using an RGB-D Camera

Gowthami.D, Dr.S.M.Ramesh

Mirror is probably the most shared optical device in our daily life. The virtual mirror rendering by means of a combined camera-display system improved with virtual scenes and objects, has a wide range of applications from cosmetics to medical interventions.Challenges like wide angle viewing of the environment, accurate viewpoint tracking and rendering, as well as real time demonstration to deliver immediate visual feedback are involved in the realistic simulation of a mirror. In this paper, we suggest virtual image rendering system using a commodity structured light RGB-D cameras.These cameras offer the depth information for viewpoint tracking and rendering of scene from altered prospective. A novel depth denoising and completion algorithm is proposed to precise the problem of missing and erroneous depth measurements by structured light cameras. A scalable client server architecture is hosted to render the dynamic scene captured by the static camera network in online with the 3-D background prototype generated by offline.In addition to this labeling of objects are done for the images which must be in a recognized manner. Experimental results demonstrate effective performance of our method in which it provides reliable depth estimation, image rendering and statistics identification.