Facial Expression Synthesis Based on Cloud Model

Abstract

There are many differences between human faces, but still having common characteristics. The person's facial contour can be approximated as ellipses, and the relative position of eyebrows, eyes, nose, mouth and other organs is stable in the whole face. Such shapes are similar and can provide the basis for the realization of human face synthesis. Whether in technology or in the application, human face synthesis with computer has broad prospects. As mathematical conversion model of uncertain knowledge, cloud model integrates the fuzziness and randomness to constitute the mapping between qualitative and quantitative, while the facial expression is a kind of uncertainty data. This paper proposes face synthesis technology based on cloud model. First of all, expand the cloud model algorithm from data points to data set and then put each piece of face image as a M'N (M rows, N columns are actually the image positioning) grid in order to make each image grid have a grayscale value (0-255). Secondly, extract cloud numerical characteristics (Ex, En, He) of inputted human face image with backward cloud generator. Thirdly, by positive cloud generator, generate a set of cloud droplets which have corresponding figures feature. And finally, achieve human face synthesis with backward cloud generator. Human face synthesis technology based on cloud model, realizes human face synthesis of multi-face expression sources based on different weighting ratio. The experimental results show that it can obtain different expression of modes, and enrich the connotation of the performance of facial expression by adjusting values of the weight vector.