Universal robust algorithm for detection of image features
Konstantin E. Rumyantsev, Dmitry Petrov · 2016
Effective detection of the feature points in images requires usage of detectors that produces similar results in variable observation conditions. Authors present a feature detector which is based on a uniformly most powerful invariant algorithm that uses t-test. The authors obtained equations that allow evaluating the effectiveness of the robust feature detector. Simulation proved the robustness (constant rate of false alarms) of the proposed feature point detection algorithm to changes in mean value and standard deviation of background pixels' intensity. The study proved invariance of the synthesized detector structure to the angle of the image rotation. The uniqueness of detected features is determined by the formation of support and analyzed samples.