This chapter presents a novel inheritable color space (InCS) and a generalized InCS (GInCS) framework for kinship verification. Unlike conventional color spaces, the proposed InCS is automatically derived by balancing the criterion of minimizing the distance between kinship images and the criterion of maximizing the distance between non-kinship images based on a new color similarity measure (CSM). Two important properties of the InCS, namely, the decorrelation property and the robustness to illumination variations property, are further presented through both theoretical and practical analyses. To utilize other inheritable features, a generalized InCS framework is then presented to extend the InCS from the pixel level to the feature level for improving the verification performance as well as the robustness to illumination variations. Experimental results on four representative datasets, the KinFaceW-I dataset, the KinFaceW-II dataset, the UB KinFace dataset, and the Cornell KinFace dataset, show the effectiveness of the proposed method.
CITATION STYLE
Liu, Q., & Liu, C. (2017). Inheritable color space (InCS) and generalized InCS framework with applications to kinship verification. In Intelligent Systems Reference Library (Vol. 121, pp. 65–89). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-319-52081-0_4
Mendeley helps you to discover research relevant for your work.