ABSTRACT
Harris algorithm is a method for extracting point feature based on signal firstly proposed by C. Harris and Stephens [6] in 1988. Harris algorithm is an effective method to extract point feature [7]. We can obtain many corner points when using Harris corner detection for full facial image. Because the emphases of eye feature extraction is to find the eye corner, we must select eye corner points from the points we obtained previously, here we fix the area for selecting eye corner points in the eye rough location area previously. In this area, we select the furthest left and furthest right points as the eye corner points. Simultaneity, the vertical position height must match the criteria: Rtop < Height < Rbottom, here Rtop and Rbottom are the top position and bottom position of pupil fixed previously. From our experiment results, we can find that the points detected in facial area using Harris corner detection distribute intricately. If we only use Harris corner detection to extract facial feature, we must select the points we need after detection. Unfortunately, the algorithms using for selecting points we need are usually very intricate, and the result is usually not satisfying [8]. So only a few researchers use single corner detection to extract facial feature. In view of the factors above, we use Harris corner detection to detect eye corner position after roughly locating the eye area and accurately locating pupil position. Because we narrow down limit the searching area, and the position of eye corner is also restricted by the position of pupil and the size of eye after normalization, the method we presented can locate eye corner quickly and accurately.
