Still to video face recognition using a heterogeneous matching approach

Abstract

In this paper, we address the problem of still-to-video (S2V) face recognition. Still images usually have high qualities, captured from cooperative users under controlled environment, such as the mugshot photos. On the contrary, video clips may be acquired with low resolutions and low qualities, from non-cooperative users under uncontrolled environment. Because of these significant differences, we consider the S2V as a heterogeneous matching problem, and propose to develop a method to bridge the gap between the two heterogeneous modalities. A Grassmann manifold learning method is developed to construct subspaces for the purpose of bridging the gap between the two face modalities smoothly. We conduct extensive experiments on two large scale benchmark databases, COX-S2V and PaSC, with different recognition tasks: face identification and verification. The experimental results show that the proposed approach outperforms the state-of-the-art methods under the same experimental settings.