Automatic eyebrow features detection and realization of avatar for real time eyebrow movement

This paper focuses on extraction of eyebrow key points in video sequences and realization of an avatar that uses the located key point's coordinates to show automatic movements of eyebrow in accordance with the video clip. The work uses a color based method for automatic segmentation of eyebrows from video clips and extraction of the key eyebrow features like, two corners of the eyebrows and upper and lower mid points, from eyebrow contour. The color based method of eyebrow segmentation extracts pseudo-hue plane of eyebrow region to separates eyebrow from skin region. The second part of the work focuses on the realization of avatar for eyebrow movements using the obtained eyebrow features. The eyebrow segmentation and features extraction method is much faster than other existing model based methods, as it is simple and uses no prior information about the eyebrow, also it needs no learning method. Around 200 frontal face images from FEI face database [1] and around 100 facial expression videos containing nearly frontal face images from MMI database [2] are tested. Accurate eyebrow segmented region key point detection result is obtained. Finally an avatar is realized which shows the automatic eyebrow movements in accordance to that of the video clip.