Automatic Tagging Of Facial Feature Points And Expression Generation

Facial expression generation can be defined as making a neutral face into a face image with an expression by the expression computing algorithms. The critical work of the expression generation is the tagging of the facial features. To solve the problems that manual tagging costs too much time, this research aims at automatic tagging for the application of facial expression generation. It has a grate value for fast and accurate facial expression generation.The work and innovation of this paper mainly includes:1. Firstly, I improve CANDIDE‐3 face mode according to face geometry. Two auxiliary contour points and two pupil points are added into the model to accurately express the human face geometry and position the facial features.2. Secondly, according to the five sense organs'geometry and movements in expression changing, I propose a method for automatic tagging of feature points which need segment the organs from face at first. All the organs are processed as the following steps: coarse positioning—fine positioning—the automatic tagging of facial feature points. That is to segment the face into blocks, and then get the organs'binary image or contours on the basis of which I tag the features automatically. Specifically, I introduce the watershed segmentation algorithm into the segmentation of eyebrows. Taking the relation of expression changing and muscle movements into consideration, I position the eyebrows'feature points. The segmented binarization method based on gray integral projection is proposed for eyes features'positioning. Seeing the magnitude of the open mouth as the stretch of the face and making a combination with the geometric active contour model, I can tag the mouth feature points automatically.3. Finally, I use the example‐based expression generation method. The facial expression mapping is established upon the automatically tagged facial feature points, then load the sample s expression mapping onto the target person s face to drive him producing the expression. A realistic facial expression can be obtained through the OpenGL texture mapping mechanism.