Novel Techniques for Robust Lip Segmentations, Automatic Features Initialization and Tracking

Anima Majumder, Laxmidhar Behera, and Venkatesh K. Subramanian


Pseudo-hue, Adaptive thresholding, Region Of Interest, Segmentation


Lip segmentation and tracking is an essential phase in many multimedia systems, such as automatic clone synthesis, facial expression recognition, talking heads for video conferencing and visual speech recognition. Accurate and robust lip segmentation is very important for lip image analysis. But lip segmentation is still a challenging and difficult problem due to the weak color contrast between lip region and non-lip region. This paper proposes a robust and accurate lip segmentation algorithm with high precision based on pseudo hue plane of lip image and improved adaptive thresholding algorithm. Experiments have been successfully preformed on two hundred frontal face images including images from FEI face database and some difficult images, like badly illuminated images and images with mustaches. The paper also introduces an automatic lip tracking technique in a video sequence of more than four hundred frames which automatically localizes the initial features to be tracked. We also propose an automatic tracking scheme by incorporating Shi-Tomasi's good corner detection technique. A comparative result is shown for lip tracking using Lucas-Kanade's pyramidal optical flow method and our approach. Experimental results of lip tracking using proposed approach brings a significant improvement over that of the optical flow method of Lucas-Kanade.

Important Links:

Go Back