Aiming at practical face
recognition systems, we are focusing on any point that may improve the
performance of the system. In sum, the following topics are what we
contribute to.
In the early stage of our
research, we mainly studied the Eigen-features, skin model,
rule-based method. Later on, we have focused on statistical
learning-based method. Especially, we have investigated the GAM-based
training set data re-sampling problem, the spatial histogram-based
object detection method, and ECOC-based multi-class classification
method. For more
information on this topic, please click here.
Under the framework of
sequential Monte Carlo filtering algorithm, we try to get insights
on some key issues in visual tracking with application scenarios on
face and human tracking. Some important technologies and
solutions are studied which are necessary for robust and practical
tracking systems, especially concentrate on how to model the object
appearance variation, how to model and estimate the object complex
dynamics, how to combine low level and high level motion estimation
methods to enhance tracker robust and efficiency.
For more information on this topic,
please click here.
Notice the great importance of
this problem, we have concentrate on this problem from several
aspects including the rule-based methods, Deformable templates,
AdaBoost-based method, Example-based Shape Learning, and Several
improvements on Active Shape Models.
For more information on this topic,
please click here.
Face feature extraction and
classifier designing are two essential parts for face recognition.
For years, we have been working on methods including discriminant
subspace analysis, Face Specific Subspace (FSS), Adaboosted Gabor
Fisher Classifier (AGFC), Support Vectors based Kernel discriminant
analysis (SV-KFDA), Local Gabor Binary Patterns (LGBP), Local Gabor
Phase Pattern (LGPP). For
more information on this topic, please click here.
Basically, we have studied the
image processing-based illumination normalization methods including
histogram modification (equalization or specification), log
transform, Gamma Intensity Correction (GIC), phase image, and
relative gradients edges etc. However, to better model the
problem, we have also investigated model-based methods, such as the
spherical harmonics methods.
For more information on this topic, please
click here.
Besides the efforts to seeking
pose-insensitive face descriptors, we are mainly studying
frontalization technologies that can transform a non-frontal
face images to a virtual frontal view. Based on this basic idea, we
have proposed both statistical learning and appearance based methods
such as Statistical Affine Transform, Local Linear Regression, Face
Warping based on Cylinder face model, and 3D Sparse Deformable
Model. For more
information on this topic, please click
here.
Imprecise alignment (e.g. eye
centers) will greatly degrade the face recognition performance,
which is explicitly defined by us as curse of mis-alignment. we have
investigated this problem and proposed an enhance FDA solution.
For
more information on this topic, please click here.
Please refer to our special
website for the information on the large-scale CAS-PEAL Chinese
face database.
Click here for more information.
|