A novel weighted Hausdorff distance for face localization

In this paper, a new weighted Hausdorff distance measure is proposed for face localization. The weighting function in the new Hausdorff distance measure has large values at important locations, in which the values are proportional to the frequency of edge points appearing. Thus, the weighting function can reflect the common properties of face edge maps effectively. When comparing the edge maps of the face model and that of a candidate face region in face localization, the weighted Hausdorff distance incorporates information about the location of important edge points so that the face can be localized more accurately. Experimental results show the new weighted Hausdorff distance measure achieves accurate face localization.