Face contour extraction using Snake

Face contour represents important information commonly applied in biometric applications using human face information. Here, we propose an improvement to the conventional active contour model, called Snake, in the application of face contour tracking. We propose using face detection method developed by P. Viola as a quick initialization method. A new energy term called gradient angular difference (GAD) is introduced here. GAD energy term is used to constrain the Snake contour to the specific face region. The Snake method obtains the optimal contour by iteratively minimizing its energy function. The long iteration process leads to long computation time. Therefore, we proposed a normal line searching strategy to accelerate the face contour extraction process. The simulation results show that the accurate Snake contour can be detected quickly, and with minimal computation. The method can also be used in other applications such as general organ contour extraction from X-ray.