期刊名称：International Journal of Advanced Research In Computer Science and Software Engineering
印刷版ISSN：2277-6451
电子版ISSN：2277-128X
出版年度：2012
卷号：2
期号：5
出版社：S.S. Mishra
摘要：Finding landmark positions on facial images is an important step in face registration and normalization, for both 2D and 3D face recognition. In this paper, we inspect shortcomings of existing approaches in the literature and propose a method for automatic land marking of near-frontal faces. The first level of the two-tier method analyzes faces on a coarse scale and employs Gabor features over eight orientations. The second level improves the accuracy of facial feature locations, as inherited from the first tier, by using Discrete Cosine Transform (DCT) coefficients. In addition, a structural analysis subsystem relates the feature positions in an elastic graph, determines outlier positions and can correct them with back-projections. Tests carried on two separate face databases have indicated that our proposed s scheme is robust and outperforms its competitors
