Study On Facial Aging Image Synthesis

The simulation of facial appearance variation caused by aging has been found many applications in the areas such as criminal investigation, face recognition, make-up design aid in film&TV making and digital entertainment. So far, the synthesis of facial aging image usually based on a single mode, with no consideration of different aging trajactories for various persons. Although foreign researchers had built aging database with considerable bulk, the facial images are mainly Caucasians, which are not appliable to our yellow race. In the meanwhile, methods based on 3D face models are too complicate and expensive to generalize to real-life applications.In this thesis we propose a new image based individual prototyping method for facial aging image synthesis, which can realize several important aged features such as wrinkles, pouches, darken skin, and somewhat white hair and eyebrows. Experiments show that our method can be easily manipulated and generalized to multiple fields and successfully synthesize the anile facial images at different ages for various kinds of people realistically.This thesis is organized as follows:Chapter 1 introduces the motivation and difficulties of our topic, reviews previous related work in this area, and briefly describes our main research work in this thesis.In Chapter 2, we describe our individual face shape matching algorithm based on the local standard deviation of curvatures of facial contour. Fisrt of all we build a facial image database which contains various faces of yellow race at different ages, then by calculating the local standard deviation of those curvatures, we match the input facial image with these at different age groups and find the corresponding facial image from the face database, preparing for the generation of individual prototypes.In Chapter 3, we present our implementation of individual prototyping and texture enhancement based on those facial images selected from database in the previous chapter. By the shape and color transformation model we generate the aged facial images. Our method can synthesize the aged facial images at different ages for various kinds of people realistically.The summary and conclusion of this thesis are given in Chapter 5, as well as the discussion of the further research work.