Face recognition by elastic bunch graph matching

The paper introduces a system for recognizing human faces from single images drawn from a large database that contains one image per person. Faces are represented as labeled graphs based on a Gabor wavelet transform; image graphs for new faces are extracted through an elastic graph matching process and can then be compared using a simple similarity function.

The system advances earlier work by Lades et al. (1993) in three respects: it uses phase information for accurate node positioning, employs object-adapted graphs to handle large rotations in depth, and bases image graph extraction on a novel data structure, the bunch graph, constructed from a small set of sample image graphs. These contributions made the elastic graph matching approach more robust and became influential in early face recognition research.