SELECTION OF FEATURE POINTS RESISTANT TO BLUR AND GEOMETRIC TRANSFORMATION OF IMAGES
A. K. Shakenov
Institute of Automation and Electrometry, Siberian Branch, Russian Academy of Sciences, Novosibirsk, Russia
Keywords: feature points selection, stable feature points, feature point detectors and descriptors, image search and matching, image matching, BRISK, AKAZE
Abstract
The article proposes a method for selecting stable feature points of images obtained using the BRISK and AKAZE algorithms. The method is based on the selection of points that remain in the image after its significant transformation. The results of numerical experiments, based on 1000 images, confirming the effectiveness of the proposed method are presented. It is found that the sets of points obtained using the proposed method contain three to five times more stable points than random sets of the same size. It is shown that, for the selection of stable points, it is more efficient to use significant transformations of images.
|
