Xuetao Feng

Robust facial expression tracking based on composite constraints AAM

Facial expression tracking is a challenging task because the head pose may change in a large range and the expression is highly non-rigid. It can be formulized as an energy minimization problem. The two most important issues are the construction of the cost function and the selection of the initial value. In this paper, we present a fast and robust expression tracking algorithm called Composite Constraints AAM. Firstly, a novel cost function is proposed to enhance the convergence by combining multiple constraints in a unified framework. Secondly, widely used local features are strictly tested with face videos, and an efficient motion estimation method is presented to provide a good initial value to the iterative optimization process. Experimental result demonstrates that our system can track the head pose and facial expression with very high stability in real time speed.

Learning based ensemble segmentation of anatomical structures in liver ultrasound image

Automatic segmentation of anatomical structure is crucial for computer aided diagnosis and image guided online treatment. In this paper, we present a novel approach for fully automatic segmentation of all anatomical structures from a target liver organ in a coherent framework. Firstly, all regional anatomical structures such as vessel, tumor, diaphragm and liver parenchyma are detected simultaneously using random forest classifiers. They share the same feature set and classification procedure. Secondly, an efficient region segmentation algorithm is used to obtain the precise shape of these regional structures. It is based on level set with proposed active set evolution and multiple features handling which achieves 10 times speedup over existing algorithms. Thirdly, the liver boundary curve is extracted via a graph-based model. The segmentation results of regional structures are incorporated into the graph as constraints to improve the robustness and accuracy. Experiment is carried out on an ultrasound image dataset with 942 images captured with liver motion and deformation from a number of different views. Quantitative results demonstrate the efficiency and effectiveness of the proposed algorithm.

Robust pose normalization for face recognition under varying views

Unconstrained face recognition under varying views is one of the most challenging tasks, since the difference in appearances caused by poses may be even larger than that due to identity. In this paper, we exploit and analyze a novel pose normalization scheme for facial images under varying views via robust 3D shape reconstruction from single, unconstrained photos in the wild. Specifically, to address the problem of ambiguous 2D-to-3D landmark correspondence and imperfect landmark detector, for each input 2D face, the 3D shape is suggested to be learned by iteratively refining the 3D landmarks and the weighting coefficients of each landmark. Experimental results on both LFW and a large-scale self-collected face databases demonstrate that the proposed approach performs better than the existing representative technologies.

An extensible framework for facial motion tracking

Facial motion tracking is a challenging task because of highly flexible head pose and facial expression. An extensible tracking framework is proposed in this paper. Within the framework, proper models are selected according to requirements and restrictions of the application, and different trackers can be constructed to handle different tasks. Experimental result shows that our tracker outperforms the existing commercial software.

Combining nonuniform sampling, hybrid super vector, and random forest with discriminative decision trees for action recognition

Trajectory-based features have become popular for action recognition and achieve the state-of-the-art results on a variety of datasets. In this paper, we propose a novel framework to improve the performance of action recognition. Specifically, we first apply the nonuniform sampling method to efficiently select features for given actions. The proposed hybrid super vector, namely fisher vector (FV) combined with vector of locally aggregated descriptors (VLAD), is then employed to encode sampled trajectories. A random forest with discriminative decision trees, where every tree node is a discriminative classifier, is finally applied to predict action labels. We have achieved 88.2% in average accuracy on the UCF101 dataset, which outperforms the best results that have been reported in the literature.