Priority and Segmentation algorithm for the 3-D motion estimation of two heads partly overlapped

Though many passive sensing methods have been presented for the analysis of 3D facial motion, generally those methods are useful only when the motion of a single face is analyzed. This paper presents an efficient method to analyze 3D head motion of two faces in the scene even when the areas of two faces are partly overlapped. To solve the problem which comes from overlapping of facial areas, PriSe (Priority and Segmentation) algorithm with the GRIP method is newly suggested in this paper. With our proposed method, the 3D head motion of two faces can be stably analyzed at the same time even though some parts of the faces are overlapped. Preliminary experiments for a series of real and synthesized image sequences show the effectiveness of the proposed method with sufficient accuracy to analyze two facial motions in 3D space.