COMPARISON OF IMAGE FEATURE DETECTORS AND EVALUATION OF THEIR STATISTICAL CHARACTERISTICS
A. K. Shakenov
Institute of Automation and Electrometry, Siberian Branch, Russian Academy of Sciences, Novosibirsk, Russia
Keywords: feature detectors and descriptors, image search and matching, image correspondences, BRISK, AKAZE, ORB
Abstract
The BRISK, AKAZE, and ORB feature detectors are compared in a series of experiments on a set of 10,000 images. Several characteristics of the efficiency of these detectors are estimated: the proportion of correctly found positions of features, the proportion of correctly established correspondences, and the number of correspondences in comparisons of two randomly chosen images. The dependence between the average number of false correspondences and the value of proximity of descriptors is estimated. Distributions of the number of correct and false correspondences are plotted.
|
