An Adaptive Active Shape Model for Eye Shape Detection

Abstract

This paper proposes an improved Active Shape Model (ASM) to detect the eye shape. The initial estimation of the eye shape and the poor description of its pixel value variation are considered as common problems in ASM. We employ Principal Component Analysis (PCA) to suggest an initial estimation of the eye shape based on understanding the eye structure in an eye image. In addition, the properties of PCA in pixel values-matching and data reduction are used to describe the pixel value variation around the eye shape and to search on the proper locations of the eye shape in the eye image. The experimental results verified the effectiveness of the proposed method compared to the standard ASM.