Automatic Real-Time Localization of Frowning and Smiling Faces under Head Rotation Variations

A new method for real-time face localization from a streaming color video was developed. The method consisted of three stages. First, the face-like skin-colored image region was segmented from the background and transformed into the grey scale representation. Second, the cropped image was convolved with Sobel operator in order to extract local oriented edges at 16 orientations. The extracted local oriented edges were grouped together to form regions of interest which represent landmark candidates. Further, the candidates were matched against edge orientation model to verify the existence of the landmark in the image. Finally, the located landmarks were next spatially arranged into the face–like constellations. The best face-like constellation of the landmark candidates was defined by a new scoring function. The test results demonstrated that the proposed method located expressive faces with high rates in real time from facial images under controlled head rotation variations. Key-Words: Face localization, Facial landmarks, Sobel edge detection, Frontal-view geometrical face model, Facial expression, Head rotation.