Description And Analysis Based On Similarity Of Face Shape Characteristics

With the high-speed development of computer network and multimedia technology, the face detection, identification, retrieval techniques based on images have gradually become quite active research categories. One of the important research topics is the human facial similarity measure, which is the key foundation and important content of face detection, identification, retrieval techniques; therefore, the research on the facial similarity has practical value and research significance.This thesis studies the principles of Angle Context algorithm, Shape Context algorithm, orthogonal Procrustes algorithms and facial geometric features algorithm. We have applied these four methods to obtain the corresponding facial similarity and conducted an empirical study on the first three algorithms. The similarity experiments mainly include face detection, feature extraction, similarity calculation.This thesis also presents two human face similarity algorithms, namely:the principle of Angle Context combines with that of orthogonal Procrustes algorithms and the principle of shape context algorithm combines with the orthogonal Procrustes. We performed experiments to obtain the corresponding facial similarity. Though mutiple experiments, we find that the two algorithms proposed in this paper can significantly improve the efficiency and accuracy of facial similarity measure. Meanwhile, we have established a marriage database based on the hit shows "If You Are the One", which includes the original database, numerical database, image database. This thesis applied the face similarity algorithm into the face images of male and female guests, and verifies the hypothesis of Couples Faces.