Real-time corner detection algorithm for motion estimation

Abstract

This paper presents a fast algorithm for corner detection based on the observation that the total curvature of the grey-level image is proportional to the second order directional derivative in the direction tangential to edge normal, and inversely proportional to the edge strength (norm of the edge normal). This algorithm simply takes the difference of the second tangential derivative with the edge strength, where the first term is the cornerness measurement and the second is called a false corner suppression. A subpixel addressing mechanism (called linear interpolation) is utilized for intermediate pixel addressing in the differentiation step, which results in improved accuracy of corner localization and reduced computational complexity. The analysis of corner dislocation leads to a subpixel implementation. The corner finder is implemented on a hybrid parallel processor PARADOX with a performance of 14 frames/s for the vision algorithm Droid.