Person re-identification by graph-based metric fusion

Person re-identification is defined as re-identifying individuals across different camera views. This is a very challenging problem since the appearance of a person can vary significantly due to cross-camera changes in viewpoint, pose and illumination. To model the transition between camera views, distance metric learning has been widely used in person re-identification and shown to be effective. However, using one specific metric often suffers from over-fitting and may not be sufficient enough to cope with the cross-camera variations of all different individuals. In this Letter, a powerful metric fusion method is proposed to combine multiple given distance metrics. Specifically, we represent given metrics as different graphs and then formulate the fusion problem as a graph-based learning framework. In this way, our framework can efficiently integrate the complementary information provided by different input metrics.