Corner characterization by statistical analysis of gradient-direction

A new approach to grey-level image corner characterization is proposed in this paper which is based on the statistical analysis of the gradient-direction in an intensity image, and takes the signal to noise ratio (SNR) into account. The proposed approach can detect corner structure, the number of lines and their orientations which construct the corner; as well as the corner location (intersection point of the lines) in subpixel accuracy. Experiments on both synthetic and real images reveal that the proposed approach can cope well with image noise.