Graph matching vs mutual information maximization for object detection

Labeled Graph Matching (LGM) has been shown successful in numerous object v ision tasks. This method is the basis for arguably the best face recognitio n system in the world. We present an algorithm for visual pattern recogniti on that is an extension of LGM ('LGM(+)'). We compare the performance of LG M and LGM(+) algorithms with a state of the art statistical method based on Mutual Information Maximization (MIM). We present an adaptation of the MIM method for multi-dimensional Gabor wavelet features. The three pattern rec ognition methods were evaluated on an object detection task, using a set of stimuli on which none of the methods had been tested previously. The resul ts indicate that while the performance of the MIM method operating upon Gab or wavelets is superior to the same method operating on pixels and to LGM, it is surpassed by LGM(+). LGM(+) offers a significant improvement in perfo rmance over LGM without losing LGM's virtues of simplicity, biological plau sibility, and a computational cost that is 2-3 orders of magnitude lower th an that of the MIM algorithm. (C) 2001 Elsevier Science Ltd. All rights res erved.