Probabilistic Facial Feature Extraction Using Joint Distribution of Location and Texture Information

In this work, we propose a method which can extract critical points on a face using both location and texture information. This new approach can automatically learn feature information from training data. It finds the best facial feature locations by maximizing the joint distribution of location and texture parameters. We first introduce an independence assumption. Then, we improve upon this model by assuming dependence of location parameters but independence of texture parameters. We model combined location parameters with a multivariate Gaussian for computational reasons. The texture parameters are modeled with a Gaussian mixture model. It is shown that the new method outperforms active appearance models for the same experimental setup.