Facial caricature has a potential for realizing a new media of man-machine interface and for modeling facial impression. We have been proposing a method of 3D caricature generation which is based the shape features of facial parts for the 3D facial image. Our method generates a caricature by using the individual feature vector that is defined by the difference between the input face and the mean face which is prepared from many input faces. This method is sometimes likely to suffer from fatal inadequate deformations both of the local and global shapes of the face. This fatal deformation is caused by the fact that the variances of the feature points of the mean face are not uniform among facial parts. Taking this fact in consideration, in this paper, we propose a new caricaturing method by introducing sample variances inherent in the mean face into the deformation process. Experimental results will show the novelty of the proposed method which is friendly adapted to the respective user.
