| The colors of the eyebrows, eyes, nostrils, and lips are different from the color of the face skin. The face color and the background color are different, too. Based on these different kinds of contrasts, threshold, Sobel operator, edge searching and thinning, noise removing, and feature extracting algorithms are applied to extract the shapes and positions of the eyes, nose, mouth, and the face outline.; In this dissertation, geometric and feature-based matching are the principles used to recognize different people's faces.; The eyes, eyebrows, mouth, and face outline, for different people, have different kinds of shapes, sizes, and orientations. The distances between the eyes, eye and nose, eye and mouth, and the left chin and the right chin are different for different people, too. Based on these differences, objects' distance, image shifting, image rotating, image interpolating, and image subtraction algorithms are applied to classify different people's faces.; Major axis algorithm and centroid algorithm are used to find the orientations and locations of the feature objects. Under normal circumstances, even though the object's position is shifted, rotated, or the scale is changed, the algorithm will still correctly classify different people's faces.; The centroid and major axis algorithms can very precisely locate the centroids and orientations of the objects. The subtracted results of the gray scale face image and the triangular image formed by the eyes, nose, and mouth have the best results to classify different people's faces. The method of using the length of the major axis has the poorest classifying result. The obtained classifying data can accurately classify different people's faces. |
