Joint Estimation of Object Motion and Point Light Illumination for 3D–Object–Based Analysis–Synthesis Coding

In video telephone scenes to be transmitted at low data rates non–diffuse scene illumination has to be expected. In this contribution, non–diffuse scene illumination and 3D object motion is estimated concurrently from a monocular image sequence. Image analysis uses the source model ”3D rigid moving objects with non–diffuse illumination”. The illumination is modeled by a distant point light source and ambient light. Experimental results of estimating and compensating non–diffuse scene illumination and 3D object motion using real and artificially animated image sequences are given. For a real image sequence, a prediction gain of about 0.6 dB can be stated compared to the source model ”3D rigid moving objects with diffuse illumination”.