Earn Khwang Teoh

Structure-adaptive B-snake for segmenting complex objects

In this paper, we presented a structure-adaptive B-snake model for segmenting the complex structures in medical images. A strategy of automatic control-point insertion, adaptive to the structure of the studied object, has been proposed. Furthermore, a method of minimum mean square energy (MMSE) is developed to iteratively estimate the position of those control points in the B-snake model. By applying the proposed structure-adaptive B-snake model to medical images, we show that our method is robust and accurate in object contour extraction.

A novel 2D shape matching algorithm based on B-spline modeling

This paper presents a novel algorithm for 2D planar curve recognition based on B-spline modeling. It combines the advantages of the B-spline which are continuous curve representation and affine invariant, and the robustness of the CSS (curvature scale space) matching with respect to noise and affine transformation. It solves the problem of the nonuniqueness of the B-spline in curve matching. A new algorithm, which contains degree increase and reduction, is proposed for smoothing the B-spline and constructing the CSS image. The proposed algorithm has been tested and good performance has been obtained.

A novel Bayesian shape model for facial feature extraction

This paper presents a novel application of the Bayesian shape model (BSM) for facial feature extraction. First, a full-face model is designed to describe the shape of a face, and the PCA is used to estimate the shape variance of the face model. Then, the BSM is applied to match and extract the face patch from input face images. Finally, using the face model, the extracted face patches are easily warped or normalized to a standard view. Applications of this facial feature extraction algorithm include face recognition, face video coding and retrieval, face animation and multimedia.

Recognition of Planar Shapes Using Algebraic Invariants from Higher Degree Implicit Polynominals

Higher degree implicit polynomials and moments have been used earlier for characterizing shapes. However, computation of invariants of implicit polynomials of degree greater than four has been known to be a complex problem. In this paper, an computationally efficient method for obtaining affine invariants from coefficients of higher degree implicit polynomials has been proposed. The algorithm is based on tensors and unlike the previous works, it is not based on partial derivative forms or symbolic computation. The algebraic invariants from the higher degree implicit polynomials can be directly used for obtaining moment invariants.

Soft-biometric detection based on supervised learning

In the past 5 years, people re-identification has been a popular topic as an application using computer vision techniques. Among the models used for people re-identification, soft-biometric traits based models have great potential due to the semantic meaning and robust performance they have. In this paper, we will exploit the performance of supervised learning based method on the detection of three soft-biometric traits: Glasses, Cap and Clothes Pattern. Simple features like edge and frequency are extracted from sample images and used for learning. Two supervised learning methods - Support Vector Machine (SVM) and Extreme Learning Machine (ELM) are employed and compared. Different normalization methods are compared as well. Experiments are carried out on images from FERET dataset and images collected online, and discussion is provided.

Applying training hidden features to joint curve evolution for brain MRI segmentation

According to the level of information provided in images, segmentation techniques can be categorized into two groups. One is region-labeling, which obeys the intensity-based classification methods. Although modeling the tissue intensity is straightforward by applying local statistical methods and spatial dependencies, the results might suffer from noise and incomplete data. The second group of techniques applies active contour models, in which the objective is to find the optimal partition of the image domain using a closed or open curve by using prior constraints on the shape variation. However, estimating optimal curve is intractable due to the incomplete observation data. This paper extends a previously reported joint active contour model for medical image segmentation in a new Expectation-Maximization (EM) framework, wherein the evolution curve is constrained not only by a shape-based statistical model but also by applying a hidden variable model from the image observation. In this approach, the hidden variable model is defined by the local voxel labeling computed from its likelihood function, depended on the image functions and the prior anatomical knowledge. Comparative results on segmenting putamen and caudate shapes in MR brain images confirmed both robustness and accuracy of the proposed curve evolution algorithm.

3D growing deformable B-surface model for object detection

A new method, called 3D growing deformable B-surface model, is proposed for object detection which works in 3D space directly. First, the coarse boundary of the subject is extracted. The 3D external force field of the subject is generated based on this coarse boundary using modified GVF (gradient vector flow). After the initialization of a surface patch, growing B-surface model starts to deform it to locate the boundary of the object. Next, this surface patch is anchored to the surface of the subject and a new surface patch grows up based it. This process is repeated until a closed surface of the subject is obtained. 3D growing deformable B-surface model overcomes the difficulty that comes from analyzing 3D volume image slice by slice. And the computation load of B-surface is reduced since the internal force is not necessary in every iteration deformation step. Next, the geometric information on every surface point can be calculated easily. And it has the ability Io achieve high compression ratio (the ratio of data to parameters) by presenting the whole surface with only a relatively small number of control points. Growing B-surface model simplifies the initialization step of the surface model.