3D Facial Synthesis Based On Grayscale

The development of Realistic 3-D Face Modeling has taken more than thirty years since Park successfully generated a human face model in 1974. Several algorithms tend to maturity, however, there is still no such a relatively mature one with low cost and universal property due to restrictions of conditions and other factors.The information of a human full face is mainly represented by pixel. The former facial modeling were also performed in this way of modeling the tridimensional information collected by means of processing various pixel location information in different pictures.Apart from bidimensional information, the information of a human face also contains such information as hue, saturation and lightness always ignored by former algorithms in that this information in itself excludes pixel location information. But by observing different pixel, one will find pixel location bearing some connection with grey scale. On the basis of the above discovery, this paper introduces a quite different algorithm, that is, Realistic 3-D Face Modeling Generation based upon a single picture, which is mainly built on the grey scale information of a single picture. According to this picture's different grey scale information, the section and normal are calculated, hence to achieve the tridimensional information of the given point. Then various numbers of points can be achieved in accordance with requirements so as to generate realistic 3-D face models with fidelity in various degrees.This new algorithm is usually carried out as follows:1) Pretreating the picture;2) Acquiring the grayscale value of pixel in the processed picture and the inclined information of the realistic face at the given point;3) Calculating the tridimensional coordinates of the given point according to the inclined information;4) Evaluating the tridimensional coordinates of the point at a particular location (the characteristic point);5) Generating 3-D model according to the tridimensional information of the characteristic point;6) Performing texture mapping to the ready-made tridimensional model.This algorithm is featured with a universal property in that on one hand it calls for a lower requirement of data input and on the other hand only one picture of a human full face is needed. Compared with other algorithms, this one can be conducted with an average certificate picture which can be shot easily and of which the shooting request is consistent. In consequence, this algorithm can be applied to full face modeling at most cases with little distortion.This algorithm is applicable to many occasions such as 3-D animation movie, game industry and virtual human synthesis which are all endowed with a promising future at present. For example, 3-D animation movie and 3-D cyber game are the most flourishing industries with a broad market prospect and a high profit margin. And virtual human synthesis, as a branch of computer technology, can satisfy needs for teleconference and virtue host.In addition, this algorithm can be applied to face recognition. According to the former algorithms, human face is recognized by means of bidimensional information, whereas as for this new algorithm, this job can be accomplished with three dimensional information gathered by way of changes in brightness and grey scale of a picture or a realistic person's face. Furthermore, two-eye recognition can be completed if two cameras coordinate at different locations.