Abstract
The automatic detection of the lip contour is relatively a difficult problem in computer vision due to the variation amongst humans and environmental conditions. In this paper we improve upon the classical methods by introducing fusion. Two separate methods are first applied, one based on edge detection and the other on region segmentation to detect the outer lip contour, the results from them are then combined. Empirical evaluation of the detection process is also presented on an image subset of the Valid database, which contains lighting, pose and speech variation with promising results.