Comparative analysis of color- and grayscale-based feature descriptions for image recognition
- 165 Downloads
The method for evaluating the applicability of color- and grayscale-based feature spaces to the image recognition problem is considered. The Histogram of Oriented Gradients (HOG) is used as a descriptor of the image area. Color descriptions involve the gradient calculated from one of the channels of the HSV space or the CIECAM02 model . Parametric optimization is performed for each descriptor to determine the gradient threshold and size of the image area. The Mahalanobis distance between descriptions of images of different classes is used as the optimality criterion. Feature spaces are analyzed in terms of classification of open and closed eyes. The description separability of eye images of different classes has proved to be higher when using color-based descriptors with adaptation to saturation.
Keywordsfeature space descriptor description separability evaluation color model
Unable to display preview. Download preview PDF.
- 3.A. Koschan and M. Abidi, “Detection and classification of edges in color images,” Signal Processing Mag., Special Issue on Color Image Processing 22(1), 64–73 (2005).Google Scholar
- 6.S. A. J. Winder and M. Brown, “Learning local image descriptors,” in Proc. of IEEE Conf. on Computer Vision and Pattern Recognition (Minneapolis, June 2007), pp. 1–8.Google Scholar
- 9.B. Basturk and D. Karaboga, “An Artificial Bee Colony (ABC) algorithm for numeric function optimization,” in Proc. IEEE Swarm Intelligence Symp. (Indianapolis, May 12–14, 2006).Google Scholar