Efficient Non-Parametric Corner Detection: An Approach Based on Small Eigenvalue

In this paper, we present a fast non-parametric adaptive scheme for detecting corner points on a boundary curve. The accuracy of the proposed method steps from the use of an asymmetric region of support that automatically adopts to the scale of the corner point. The speed of the method is achieved based on a dynamic update of the covariance matrix used to determine the region of support. Several experiments have been conducted to revel the qualities of the proposed method and also to establish its superiority over several other existing methods.