David Tien

Relevance Units Latent Variable Model and Nonlinear Dimensionality Reduction

A new dimensionality reduction method, called relevance units latent variable model (RULVM), is proposed in this paper. RULVM has a close link with the framework of Gaussian process latent variable model (GPLVM) and it originates from a recently developed sparse kernel model called relevance units machine (RUM). RUM follows the idea of relevance vector machine (RVM) under the Bayesian framework but releases the constraint that relevance vectors (RVs) have to be selected from the input vectors. RUM treats relevance units (RUs) as part of the parameters to be learned from the data. As a result, a RUM maintains all the advantages of RVM and offers superior sparsity. RULVM inherits the advantages of sparseness offered by the RUM and the experimental result shows that RULVM algorithm possesses considerable computational advantages over GPLVM algorithm.

Tensor LRR and Sparse Coding-Based Subspace Clustering

Subspace clustering groups a set of samples from a union of several linear subspaces into clusters, so that the samples in the same cluster are drawn from the same linear subspace. In the majority of the existing work on subspace clustering, clusters are built based on feature information, while sample correlations in their original spatial structure are simply ignored. Besides, original high-dimensional feature vector contains noisy/redundant information, and the time complexity grows exponentially with the number of dimensions. To address these issues, we propose a tensor low-rank representation (TLRR) and sparse coding-based (TLRRSC) subspace clustering method by simultaneously considering feature information and spatial structures. TLRR seeks the lowest rank representation over original spatial structures along all spatial directions. Sparse coding learns a dictionary along feature spaces, so that each sample can be represented by a few atoms of the learned dictionary. The affinity matrix used for spectral clustering is built from the joint similarities in both spatial and feature spaces. TLRRSC can well capture the global structure and inherent feature information of data and provide a robust subspace segmentation from corrupted data. Experimental results on both synthetic and real-world data sets show that TLRRSC outperforms several established stateof- the-art methods.

Low rank representation on Riemannian manifold of symmetric positive definite matrices

Sparse coding aims to find a more compact representation based on a set of dictionary atoms. A well-known technique looking at 2D sparsity is the low rank representation (LRR). However, in many computer vision applications, data often originate from a manifold, which is equipped with some Riemannian geometry. In this case, the existing LRR becomes inappropriate for modeling and incorporating the intrinsic geometry of the manifold that is potentially important and critical to applications. In this paper, we generalize the LRR over the Euclidean space to the LRR model over a specific Rimannian manifold—the manifold of symmetric positive matrices (SPD). Experiments on several computer vision datasets showcase its noise robustness and superior performance on classification and segmentation compared with state-of-the-art approaches.

Blind image deblurring via coupled sparse representation

Abstract The problem of blind image deblurring is more challenging than that of non-blind image deblurring, due to the lack of knowledge about the point spread function in the imaging process. In this paper, a learning-based method of estimating blur kernel under the l 0 regularization sparsity constraint is proposed for blind image deblurring. Specifically, we model the patch-based matching between the blurred image and its sharp counterpart via a coupled sparse representation. Once the blur kernel is obtained, a non-blind deblurring algorithm can be applied to the final recovery of the sharp image. Our experimental results show that the visual quality of restored sharp images is competitive with the state-of-the-art algorithms for both synthetic and real images.

Spoken communication with computer game characters

This paper explores the characteristics of a new language for communication with game characters. This new language is called game pidgin language (GPL) (Rudra et al., 2003). A GPL is a sub-class of computer pidgin language (CPL) (Hinde and Belrose) which is a new spoken language taught to the game player and is efficient for dialogues with the computer. The GPL is extended by adding three conditions to the grammar. We add some attributes in an already established eXtensible Markup Language (XML) to encapsulate a cognitive response from the non-player-character (NPC). A complete game pidgin model is proposed in this paper.

Tensor LRR based subspace clustering

Subspace clustering groups a set of samples (vectors) into clusters by approximating this set with a mixture of several linear subspaces, so that the samples in the same cluster are drawn from the same linear subspace. In majority of existing works on subspace clustering, samples are simply regarded as being independent and identically distributed, that is, arbitrarily ordering samples when necessary. However, this setting ignores sample correlations in their original spatial structure. To address this issue, we propose a tensor low-rank representation (TLRR) for subspace clustering by keeping available spatial information of data. TLRR seeks a lowest-rank representation over all the candidates while maintaining the inherent spatial structures among samples, and the affinity matrix used for spectral clustering is built from the combination of similarities along all data spatial directions. TLRR better captures the global structures of data and provides a robust subspace segmentation from corrupted data. Experimental results on both synthetic and real-world datasets show that TLRR outperforms several established state-of-the-art methods.

An automated image analysis approach for classification and mapping of woody vegetation from digital aerial photograph

This paper presents a recent study on woody vegetation delineation and mapping using digital aerial photograph and geographic information system (GIS) in Hunter Region, Australia. The aim of the study was to develop automated and repeatable digital image processing methods for woody vegetation classification and mapping using aerial photograph or high-resolution satellite images and GIS. Parallelepiped classification or density slice method was used to classify woody and non-woody vegetation, and ancillary GIS data were used as quality controls in the classification processing. Specific scripts were developed for automated image processing in a GIS environment. The classification accuracy was assessed against traditional aerial photograph interpretation using adequate random points. The automated process reached an overall classification accuracy of 94% and 97% after post-classification correction. The automated approach can be applied to any other type of high-resolution imagery such as SPOT 5, ALOS, IKONOS and QuickBird images.

Swimming Pool Identification from Digital Sensor Imagery Using SVM

bushfires. In January 2003 it witnessed a significant natural disaster when Canberra was struck by firestorms. The Ash Wednesday fires in Victoria resulted in 75 deaths twenty years earlier. Events like these have made environmental and emergency management research a focus in Australia. In this paper, we introduce a Support Vector Machine (SVM) technique to classify small area water bodies, swimming pools in particular. These features provide a valuable source of water for Emergency Services and play a crucial role in fighting bushfires in urban areas of Australia.

Enhancement of semantics in CBIR

Although much research has been done in the area of content based image retrieval (CBIR), little progress has been made to fully implement an engine solely based on the search of image content. This paper examines one of the basic problems in pattern recognition which highlights the difficulty in the area of content understanding in CBIR, i.e. the inability of current systems to fully incorporate low level features of image, such as intensity, colour, texture, shape and spatial constraints characteristics, with the high level features such as semantic content. To further the development of content based image processing, semantic algorithms should be combined with low level features and be used to process the image objects.

Development of a portable electrical capacitance tomography system

Industrial process tomography is a non-invasive visualisation technique to obtain cross-sectional images of dynamic industrial processes. Electrical capacitance tomography (ECT) is the first developed and the most mature tomographic imaging technique used for industrial applications. Comparing with other tomographic techniques, it is relatively cheap, fast and safe. This paper describes the process of developing a portable ECT system in terms of sensor design, capacitance measuring circuit design, capacitance calibration and image reconstruction. Different image reconstruction algorithms were verified and compared.