Abstract: In image processing, corner point means the dominant point of maximum curvature along the bounding edge of planar object. A method of detecting and localizing corners of planar curvature based on Gaussian scale space is presented. An adaptive curvature estimate method measures the significance of each point on the boundary preprocessed by Gaussian smoothing; then a criterion is established to choose the desired corner points. The method requires no input parameter, and experiments were performed to show that the scale space corner detector based on Gaussian smoothing is reliable for objects with multiple-size features and noisy boundaries.