Feature Point Detection
Arthur Ardeshir Goshtasby · 2017
This chapter discusses various methods for detecting feature points in an image. A feature point represents the center of a dark or bright spot/blob, the point of intersection of two or more straight edges, the point of locally maximum curvature along an edge contour, or the center of a unique neighborhood in an image. The method used to detect feature points in images should be insensitive to global radiometric and geometric differences between the images. The chapter discusses various image properties/features suitable for detecting highly repeatable feature points in images with possible radiometric and geometric differences. It also discusses different feature point detectors. The detector based on central moments depends on the orientation and scale of an image, but because it uses intensity gradients rather than intensities, it is insensitive to global intensity changes. Performance measures important in a feature point detector are repeatability, positional accuracy, and speed.