A Fully Automatic and Haar like Feature Extraction-Based Method for Lip Contour Detection
Zahedi Morteza, Mohamadian Zahra · 2013
In this paper we propose a fully automatic and efficient method for lip contour detection. At first face region is separated from any background using a variation of AdaBoost classifiers trained with Haarlike features extracted from the face. After that by applying the second classifier trained with mouth Haarlike features, on face region mouth region is extracted, at last sobel edge detection operator applies on mouth region and lip contour is detected. Most of previous methods are based on image color intensity, such methods act very weak on pictures with low contrast and noise, furthermore they are very time consuming. Our method by using Haarlike features is very robust against low contrast and noisy images. It is also very fast, efficient and fully automatic.