Presentation of a new color-based descriptor for videos, called Space-Time Histogram. We apply this model on persons extracted from news videos and we link all the appearance of each person throughout the video using the histograms signature. We compare our results with those obtained by histograms and spatiograms.