Describing corners using angle, mean intensity and entropy of informative arcs

Corners are important features in images as they help identify the boundaries of objects. For real-time applications, it is essential that corners can be detected and matched reliably and rapidly. Presented is a descriptor, AMIE, which is compatible with standard corner detectors. It shows that the accuracy of AMIE is similar to that of the popular BRIEF descriptor, yet somewhat quicker to calculate and an order of magnitude faster to match.