Corner detection is an important task in various computer vision and image-understanding system. Traditional methods are based on chain-code and curvature computation of curves

which suffer the dependence either on the correctness of region segmentation or on the susceptivity of noise. In this paper

a novel corner detection method based on mathematical morphology is proposed

which is very different from traditional chain-code based corner detection methods. This method is based on morphological skeleton principle

and uses a modified opening operator to detect the convex and concave corner of the image. The result of the corner detection is achieved by compose the result of the two-corner sets of the source image and its complement set. The multi-scale morphological filter is used to eliminate noise. The uniform model of corner detection has also been established. Experiments show that this method leads to accurate detection

low computational cost

and rotational invariance.