Dissertation > Excellent graduate degree dissertation topics show

3D Face Model Reconstruction for Face Recognition

Author: TuYi
Tutor: ZhaoQunFei
School: Shanghai Jiaotong University
Course: Pattern Recognition and Intelligent Systems
Keywords: Three - dimensional face model Eye Location Face model reconstruction Facial Animation
CLC: TP391.41
Type: Master's thesis
Year: 2008
Downloads: 164
Quote: 1
Read: Download Dissertation

Abstract

This paper presents a single face photo for feature extraction, automatic 3D face reconstruction method of deformation of the standard model. The method of single-face photographs (there are some side angle [-10 °, 10 °]) for automatic 3D face reconstruction using a priori statistical knowledge of the deformation model to obtain depth information, and then the general face model deformation. Experimental results show that this method can be generally face photos fast auto-realistic three-dimensional reconstruction, good practical value, can be used for face recognition, animation, and other fields. Using an improved the ASM method of automatically accurate extraction of facial feature points, obtained by making the the sparse deformation model matching flat-screen feature points to the photo face the depth, then the general face model deformation to a specific face. , Making a certain angle photograph of the side face feature extraction based on skin color model optimized ASM reconstruction can have a good effect. The same time, the use of a well-settled based on the texture of the skin model fusion technology side missing information. Experimental results show that the method is quick and easy, with a single photo only fully automated complete the reconstruction, without user interaction, generated three-dimensional model has a better sense of reality. Specific studies have several parts: 1) extract biometric face detection based on local and global features extracted recent noteworthy research focus, which is an important preliminary work model reconstruction. In this paper, we consider local features, eye detection. Wherein the center of the pupil of the extract used for the initial positioning of the ASM algorithm, to prepare the ground for better convergence. For global features, we use the improved ASM algorithm. In order to improve the texture reconstruction effect, the introduction of the the ASM feature point correction based on skin color model, the feature points does not fall outside of the skin, avoiding the side of the phenomenon of missing texture texture reconstruction. 2) new texture Renewal ideas using only a frontal face image reconstruction of the corresponding three-dimensional face model. For the reconstruction process, this article will be divided into shape reconstruction and texture mapping, two parts. Shape reconstruction based on feature points extracted on the basis of two-dimensional and three-dimensional good combination of optimal approximation to obtain three-dimensional feature point, and its deformed shape model flexible algorithm. Insinuate texture, this article used to map facial feature extraction based on skin color model optimized ASM ASM feature points before the correction, the asymmetric texture mapped to the symmetrical reconstruction model to effectively prevent the non-strict frontal face photos lead side missing texture problem, obtain better reconstruction results. 3) more humane application demo framework to experimental results can be applied to the project investor and applications as soon as possible, I use the object-oriented C language core algorithm and its function package object for laboratory and practical. At the same time, in the VS 2005 MFC framework to build a practical and friendly application platform. 4) corresponding to the face model facial animation, in rebuilding good model, not just the three-dimensional face recognition system to provide a good way, a broad range of applications in many areas at the same time. We use the FDP (the Facial DefinitionParameter, face-defined parameters) and FAP (Facial Animation Parameter, facial animation parameters) to achieve facial animation, where the ultimate goal of the research project will generate good face model made talking head, open book robots to help blind reading robot to services for the blind, the paper will examine the advance to the application of the facial expression animation, will be applied to the voice port type animation. For in-depth study of the above aspects, mainly in the following areas achieved a breakthrough and progress, some research.

Related Dissertations

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device
© 2012 www.DissertationTopic.Net Mobile