Robust Face Recognition and Impostors Detection with Partial Occlusion and Small Number of Training Samples

摘要

This paper includes two contributions. The first, we present a new approach for face recognition subject to partially occlusion with a small number of training images. The new approach is an extension of our previous Posterior Union Model (PUM). It takes place the probability measure with a similarity measure, thereby allowing the use of a small number of images, or even a single image, to represent a class while keeping the robustness to partial occlusion. Base on the new method, this paper further studies the problem how to detect the invalid test samples or called "impostors" and propose a novel rule for impostors detection in face recognition. Experiment results on XM2VTS and AR database indicate that the new system is able to perform as well as an oracle system for dealing with various simulated and realistic partial occlusion without requiring prior information and also show new impostors detection rule outperforms traditional residuals-based decision measurement.