Tracking and recognizing people in colour using the earth mover's distance

This paper proposes a method that tracks and recognizes people in indoor scenes. Background subtraction and foreground silhouette analysis are used to detect people. The colour of each person's clothes is used as a distinguishing feature for the purpose of tracking and recognizing people. Clothing colour is extracted using histograms on a luminance and perceptually uniform chrominance colour space. A measure of dissimilarity between different histograms is computed using the earth mover's distance. This approach has been used in an application consisting in monitoring people entering or leaving a room.