Extending the Scale Invariant Feature Transform Descriptor into the Color Domain
Robert H. Luke, James M. Keller, Jesús Chamorro-Martínez · 2008
In recent years, Lowe’s Scale Invariant Feature Transform (SIFT) algorithm has become a widely used tool for object recognition. One shortcoming is that it only works on grayscale image intensities. In this paper, we propose extending the SIFT descriptor to also incorporate color information in a novel fashion. The keypoint descriptor proposed in this paper is a superset of the original feature vector and works exactly the same on grayscale or single band images as well as in regions of a color image that are essentially monochromatic. Examples are presented to demonstrate the properties and matching capability of the new descriptor in color image processing.