High-Order Luminance Modeling for 3-D Model-Based Coding of Multiview Image Sequences

An interframe coding method for multiview image sequences using 3-D models is presented. The 3-D motion is estimated using an optical ow formulation which also takes into consideration the photometric e ects of rotation, using a specular illumination re ection model. More speci cally, the Phong specular re ection model is used, resulting to a luminance model which depends on the surface normal at each 3-D point, six illumination parameters, as well as the viewing angle. A Gauss-Newton optimization technique is used to estimate the illumination parameters. During the motion estimation stage, the same technique is used to estimate the 3-D motion parameters. Experimental results show that the use of this illumination model for motion compensation results in improved image reconstruction, when compared to standard 2-D and 3-D motion estimation techniques.