Volume 9 - Issue 14
An eye location algorithm based on the image marking and curve blending
Abstract
Traditional projection algorithms have a bad adaptation to face image background and illumination, and are affected by eyebrow, bangs and eye wrinkle. Due to above problems, an algorithm of eye location based on image marking and curve blending is proposed in this paper. Firstly, according to the specialty of eye gray distribution and direction selectivity of certain operators, two image marking methods are defined to overcome the influence of background and illumination. On the basis of image marking, in order to make full use of strong change of eye gray and low gray value, adaptive weight projection curve blending method is proposed in this paper. In addition, we have analyzed the projection curve after blending to ensure horizontal and vertical positions, and then realized the eye accurate location. According to the experiments on YaleB and FERET face databases, our algorithm can overcome the influence of complex illumination etc, and the correct rate has achieved 99. 9% when measure standard Derr<0.25.
Paper Details
PaperID: 84880342535
Author's Name: Zhang, H., Yang, F., Kong, Z.
Volume: Volume 9
Issues: Issue 14
Keywords: Curve blending, Eye location, Gray projection, Image marking
Year: 2013
Month: July
Pages: 5827-5835