Shadows analysis and synthesis in natural video sequences

This paper proposes a new method for shadow analysis and synthesis in natural video sequences. The shadows are detected and followed up by using several specific criteria representing their intrinsic characteristics and a modeling of the differences of luminance created by cast shadows. The search area during the detection step is reduced thanks to an estimate of the position of the light source. Finally this shadowing model is used to artificially create coherent shadows with 2D video objects inserted in the scene. This method was tested and validated on real and synthetic sequences.