Image Feature Detection and Matching Based on SUSAN Method

A new approach to image edge detection and feature matching is proposed. In which, the edge detection employs SUSAN (small univalue segment assimilating nucleus) method at low level image processing. And the integration feature matching is based on NMI (normalized moment of inertia) in combination with intensity and configuration information, achieving object recognition and tracking. The feature detection with SUSAN method locates precisely and not sensibly for local noise. NMI of image has translation invariability, rotation invariability and scale invariability. The simulated experimental results show that the algorithm is efficient for intensity variety, geometry aberration and some noise images