Detecting and Tracking Eyes Through Dynamic Terrain Feature Matching

Automatic eye detection and tracking is an important component in the advanced human-computer interface design. In this paper, we present a novel approach for detecting and tracking eyes through matching their terrain features. Regarded as a 3D terrain surface, eye region exhibits certain intrinsic traits when using a so-called topographic representation. With the topographic classification of terrain features, we generate a terrain map for each facial image and extract eye candidates from the terrain map. Our algorithm mainly consists of two parts. First, eye locations are estimated from the candidate positions using an appearancebased object recognition technique. Second, a mutual information based fitting function is defined to describe the similarity between two terrain surfaces. By optimizing the fitting function, eye locations are updated for each frame in a video sequence. The distinction of the proposed approach lies in that both eye detection and eye tracking are performed in a terrain map domain rather than an original intensity image domain. The robustness of the approach is demonstrated under various imaging conditions and with different facial appearances using a web camera.