EYE DETECTION UNDER UNCONSTRAINED BACKGROUND BY THE TERRAIN FEATURE (FriPmPO1)

Author(s) : Jun Wang (State Univerisity of New York at Binghamton, United States of America) Lijun Yin (State Univerisity of New York at Binghamton, United States of America)

Abstract : Locating eyes in face images is an important step for further automated face analysis and recognition. In this paper, we present a novel approach for eye detection without finding face region based on the terrain features. First, we argue that the eyes always have some certain topographic features if the gray-level face image is treated as a 3D terrain surface. Then some terrain maps, which denote the terrain feature of each pixels, are derived from the original images by applying topographic primal sketch theory. The pixels are classified and labeled based on their terrain features. From the terrain map, the eyes always locate in the region close to pit-labeled pixels because of its intrinsic terrain feature. At last, we construct Gaussian Mixture Model based probabilistic model to describe the distribution of pit-labeled candidates. The eye pair detection problem is transform to maximum the possibility belongs to eye space by select a particular pair of candidates. Experiments show that our method has certain robustness to the unconstrained background.

Menu