Automatic Detection of Facial Midline And Its Contributions To Facial Feature Extraction

Abstract

We propose a novel approach for detection of the facial midline froma frontal face image. Using midline as a guide reduces computationalcost required for facial feature extraction (FFE) because the midlineis capable of restricting multi-dimensional searching process intoone-dimensional search. The proposed method detects the facial midlinefrom an edge image as the symmetry axis using the generalized Houghtransformation. Experimental results on the FERET database indicatethat the proposed algorithm can accurately detect facial midlinesover many different scales and rotation. The total computationaltime for facial feature extraction has been reduced by a factor of280 using the midline detected by this method.