A New Method for Uncertainty Evaluation of Corner Detection

As a method of image feature extraction, corner detection algorithm has been applied in many fields. Uncertainty evaluation of corner detection is an important approach to evaluating the reliability of corner detection. This paper presents a new method for uncertainty evaluation of corner detection. A mathematical model which relates the uncertainty of pixel intensity with the pixel intensity and image gradient is presented. To evaluate the uncertainty of corner detection, the uncertainty associated with the intensity of each pixel, which belongs to the target to be detected, is firstly evaluated by using the mathematical model presented in the paper. Then the uncertainties associated with the output of a corner detector are evaluated by using Monte Carlo Simulation. The method proposed in this paper has been validated by using classical SUSAN corner detector as an example. The experimental results show that the uncertainty of corner detection can be evaluated accurately using this method.