Lip Segmentation And Curve Fitting In Mouth-Shape Recognition

In the paper, segmentation and contour extraction of lip are focused during theoverview of mouth-shape recognition. Firstly, Fisher transform is presented and everypixel in the selected region is Fisher transformed to compute the best projection vector.Secondly, method of threshold segmentation combine the speciality of the lip-face colorspace is used to segment the lip from the whole face. But there are several noise in theprocessed lip picture which will influence the latter processing. Here the template isutilized to smooth the picture to decrease the disturbance of the noise. After getting theclear contour of the lip, the lip edge is gained through contour extraction or contourtracking. Finally, a smooth mouth-shape curve is drew using curve fitting after gettingseveral points on the edge.