Deformable 3D face tracking in real world scenarios.
Finally, a performance driven face animation system is introduced. The proposed system consists of two key components: a robust non-rigid 3D tracking module and a MPEG4 compliant facial animation module. Firstly, the facial motion is tracked from source videos which contain both the rigid 3D head mo...
Other Authors: | |
---|---|
Format: | Others |
Language: | English Chinese |
Published: |
2010
|
Subjects: | |
Online Access: | http://library.cuhk.edu.hk/record=b6074893 http://repository.lib.cuhk.edu.hk/en/item/cuhk-344526 |
Summary: | Finally, a performance driven face animation system is introduced. The proposed system consists of two key components: a robust non-rigid 3D tracking module and a MPEG4 compliant facial animation module. Firstly, the facial motion is tracked from source videos which contain both the rigid 3D head motion (6 DOF) and the non-rigid expression variation. Afterward, the tracked facial motion is parameterized via estimating a set of MPEG4 facial animation parameters (FAP) and applied to drive the animation of the target avatar. === In the first part of the thesis, the problem of tracking a non-rigid 3D face is studied. A novel framework for non-rigid 3D face tracking is proposed for applications in live scenarios. In order to extract more information of feature correspondences, the proposed framework integrates three types of features which discriminate face deformation across different views. The integration of these complementary features is important for robust estimation of the 3D parameters. In order to estimate the high dimensional 3D deformation parameters, we develop a hierarchical parameter estimation algorithm to robustly estimate both rigid and non-rigid 3D parameters. We show the importance of both features fusion and hierarchical parameter estimation for reliable tracking 3D face deformation. Experiments demonstrate the robustness and accuracy of the proposed algorithm especially in the cases of agile head motion, drastic illumination change, and large pose change up to profile view. === The video based face recognition is studied in the second part of the thesis. Compared to the still image based recognition methods, the video based methods share the merits of spatial temporal coherence among image sequences and overcomplete training samples. We propose a framework for the task of face recognition in real-world noisy videos based on 3D deformable face tracking, which can directly estimate face pose for a view-based face recognition scheme. Meanwhile, the precise non-rigid tracking provides well-aligned face samples for the subsequent recognizer. At the recognition stage, three types of feature descriptors, including Regularized LDA, LE and sparse representation, are exploited. Extensive experiments conducted on the real world videos demonstrate that the proposed recognition framework can achieve the state-of-the art recognition results, even with the usage of a simple classifier. === Three dimensional face tracking is a crucial task for many applications in computer vision. Problem like face recognition, facial expression analysis and animation, are more likely to be solved by if the geometry and appearance properties are available through a 3D face tracker. === Zhang, Wei. === Adviser: Xiaoon Tang. === Source: Dissertation Abstracts International, Volume: 72-04, Section: B, page: . === Thesis (Ph.D.)--Chinese University of Hong Kong, 2010. === Includes bibliographical references (leaves 102-113). === Electronic reproduction. Hong Kong : Chinese University of Hong Kong, [2012] System requirements: Adobe Acrobat Reader. Available via World Wide Web. === Electronic reproduction. Ann Arbor, MI : ProQuest Information and Learning Company, [200-] System requirements: Adobe Acrobat Reader. Available via World Wide Web. === Electronic reproduction. Ann Arbor, MI : ProQuest Information and Learning Company, [200-] System requirements: Adobe Acrobat Reader. Available via World Wide Web. === Abstract also in Chinese. |
---|