A method of facial expression synthesis based on cloud model

Abstract

The invention discloses a method for synthesizing human facial expressions based on a cloud model, comprising the steps of: 1. inputting a human face image to be synthesized, and reading the human face image to be synthesized into a data matrix; 2. taking the data matrix as input , using the reverse cloud generator to process the data matrix of each face image to be synthesized separately, so as to obtain the corresponding expected image, entropy image and super-entropy image of each face image to be synthesized; 3. Using the expected image, entropy image and super-entropy image As input, use the forward cloud generator to process the desired image, entropy image and super-entropy image, and obtain the cloud drops corresponding to each face image; 4. Assign weights to each cloud drop, and combine all cloud drops into one Dayundi; 5. Taking Dayundi as input, using the reverse cloud generator in the cloud generator to realize facial expression synthesis. The method of the invention can realize the synthesis of any expression at any ratio by adjusting the weight, so that more comprehensive and rich facial expression images can be synthesized.

Description

technical field [0001] The invention belongs to the technical field of human facial expression synthesis, in particular to a method for synthesizing human facial expression based on a cloud model. Background technique [0002] The cloud model has the characteristics of macro precision, micro fuzziness, macro control, and micro uncontrollability. Its essential unit is the concept cloud composed of cloud droplets. The idea is to take into account both randomness and fuzziness. It organically integrates the randomness and fuzziness in natural language to form a mapping between qualitative and quantitative, which not only breaks through the limitation of "hard calculation" in probability statistics, but also solves the problem that is the cornerstone of fuzzy set theory. The inherent defect of membership function breaks the limitation of rough set boundary set, and provides a new method and new technology to solve the uncertainty problem in data mining. The cloud model has thre...

AI Technical Summary

Problems solved by technology

[0025] At present, there are two main methods to achieve face synthesis: one is to synthesize two-dimensional images through the movement of feature points and the corresponding deformation of surrounding points, but this method ignores the lighting conditions caused by the change of the surface shape of the face The change of facial expressions makes the synthesized facial expressions not realistic enough; the other is to first store a large number of facial expression images in the image library, and synthesize them through the deformation and combination of basic expressions. This method can synthesize highly realistic facial expressions. Expression images and expression animations, but it can only perform expression synthesis on faces in the facial expression image database, and cannot perform expression synthesis on a new face image that is not in the image database

Some scholars have considered the shortcomings of these methods and proposed Expression Ratio Image (ERI), which maps the expression of one person to the face of another person, but the expression ratio image cannot be well realized between the input images. Synthesize in any proportion

Images