Improved Facial Feature Extraction for Model-Based Multimedia
Dwight Kuo, P. Hillman, J.M. Hannah · 2005
Finding faces, locating important facial features and model fitting are required in a range of applications including model based multimedia. Accurate fitting of lip, eye and eyebrow regions is essential where a facial model is to be fitted. We propose improved feature extraction algorithms for lips, eyes and eyebrows using a combination of methods including balloon colour snakes, adaptive colour thresholding, deformable templates, lighting balancing, shadow removal and k-means clustering. Results from applying our algorithm to a large database of facial images (XM2VTS) and synthetic facial animation are presented. These demonstrate the success of our extraction system and its potential usefulness for visual media production. 1