Automatic lip model extraction for constrained contour-based tracking

Contour model-based tracking is more robust if an accurate reference shape model of the underlying object is available. Since lip shapes vary, the ability to automatically extract user-dependent lip models from input images is desirable. We present an unsupervised segmentation method to hierarchically locate the user´s face and then the lips. Techniques employed include modeling in the hue/saturation color space using Gaussian mixture models and the use of geometric constraints. With the region of interest automatically located, the model extraction problem is then formulated as a regularized model-fitting problem. The use of a generic shape as prior information improves the accuracy of the extracted lip model which is based an a cubic B-spline representation. We also describe a method to compute automatically an optimal linear color space transform needed to obtain raw estimates of the lip boundary locations, as required by the fitting procedure.