Hierarchical filtering scheme for the detection of facial keypoints

Usually, the rst processing step in computer vision systems consists of a spatial convolution with only a few simple lters. Therefore, information is lost or it is not represented explicitly for the following processing steps. This paper proposes a new hierarchical lter scheme that can e ciently synthesize the responses for a large number of speci c lters. The scheme is based on steerable lters. It also allows for an e cient on-line adjustment of the trade o between the speed and the accuracy of the lters. We apply this method to the detection of facial keypoints, especially the eye corners. These anatomically de ned keypoints exhibit a large variability in their corresponding image structures so that a exible low level feature extraction is required.