Showing papers by "James Z. Wang published in 2004"

Image Categorization by Learning and Reasoning with Regions

Abstract: Designing computer programs to automatically categorize images using low-level features is a challenging research topic in computer vision. In this paper, we present a new learning technique, which extends Multiple-Instance Learning (MIL), and its application to the problem of region-based image categorization. Images are viewed as bags, each of which contains a number of instances corresponding to regions obtained from image segmentation. The standard MIL problem assumes that a bag is labeled positive if at least one of its instances is positive; otherwise, the bag is negative. In the proposed MIL framework, DD-SVM, a bag label is determined by some number of instances satisfying various properties. DD-SVM first learns a collection of instance prototypes according to a Diverse Density (DD) function. Each instance prototype represents a class of instances that is more likely to appear in bags with the specific label than in the other bags. A nonlinear mapping is then defined using the instance prototypes and maps every bag to a point in a new feature space, named the bag feature space. Finally, standard support vector machines are trained in the bag feature space. We provide experimental results on an image categorization problem and a drug activity prediction problem.

Studying digital imagery of ancient paintings by mixtures of stochastic models

Abstract: The paper addresses learning-based characterization of fine art painting styles. The research has the potential to provide a powerful tool to art historians for studying connections among artists or periods in the history of art. Depending on specific applications, paintings can be categorized in different ways. We focus on comparing the painting styles of artists. To profile the style of an artist, a mixture of stochastic models is estimated using training images. The two-dimensional (2D) multiresolution hidden Markov model (MHMM) is used in the experiment. These models form an artist's distinct digital signature. For certain types of paintings, only strokes provide reliable information to distinguish artists. Chinese ink paintings are a prime example of the above phenomenon; they do not have colors or even tones. The 2D MHMM analyzes relatively large regions in an image, which in turn makes it more likely to capture properties of the painting strokes. The mixtures of 2D MHMMs established for artists can be further used to classify paintings and compare paintings or artists. We implemented and tested the system using high-resolution digital photographs of some of China's most renowned artists. Experiments have demonstrated good potential of our approach in automatic analysis of paintings. Our work can be applied to other domains.

The story picturing engine: finding elite images to illustrate a story using mutual reinforcement

Abstract: In this paper, we present an approach towards automated story picturing based on mutual reinforcement principle. Story picturing refers to the process of illustrating a story with suitable pictures. In our approach, semantic keywords are extracted from the story text and an annotated image database is searched to form an initial picture pool. Thereafter, a novel image ranking scheme automatically determines the importance of each image. Both lexical annotations and visual content of an image play a role in determining its rank. Annotations are processed using the Wordnet to derive a lexical signature for each image. An integrated region based similarity is also calculated between each pair of images. An overall similarity measure is formed using lexical and visual features. In the end, a mutual reinforcement based rank is calculated for each image using the image similarity matrix. We also present a human behavior model based on a discrete state Markov process which captures the intuition for our technique. Experimental results have demonstrated the effectiveness of our scheme

Machine Learning and Statistical Modeling Approaches to Image Retrieval

Abstract: In the early 1990s, the establishment of the Internet brought forth a revolutionary viewpoint of information storage, distribution, and processing: the World Wide Web is becoming an enormous and expanding distributed digital library. Along with the development of the Web, image indexing and retrieval have grown into research areas sharing a vision of intelligent agents. Far beyond Web searching, image indexing and retrieval can potentially be applied to many other areas, including biomedicine, space science, biometric identification, digital libraries, the military, education, commerce, culture and entertainment. Machine Learning and Statistical Modeling Approaches to Image Retrieval describes several approaches of integrating machine learning and statistical modeling into an image retrieval and indexing system that demonstrates promising results. The topics of this book reflect authors' experiences of machine learning and statistical modeling based image indexing and retrieval. This book contains detailed references for further reading and research in this field as well.

Clustering of Time-Course Gene Expression Data

Abstract: Microarray experiments have been used to measure genes’ expression levels under different cellular conditions or along certain time course. Initial attempts to interpret these data begin with grouping genes according to similarity in their expression profiles. The widely adopted clustering techniques for gene expression data include hierarchical clustering, self-organizing maps, and K-means clustering. Bayesian networks and neural networks have also been applied to gene clustering. Sharan & Shamir [3] provided a survey on this topic. Clustering techniques typically discover the inherent structure of the genes expression profiles based on some similarity measures. The clustering results largely depend on how the similarity measure corresponds to the biological correlation between genes. Before reliable conclusion about biological functions can be drawn from the data, the gene clusters obtained from microarray analysis must be investigated with respect to known biological roles of those clusters. The current analysis of whole-genome expression focuses on relationships based on global correlation over a whole time-course, identifying clusters of genes whose expression levels simultaneously rise and fall. However, genes may be regulated by different regulators in a long time course. Co-regulating in part of the long time course does not guarantee a global similarity in gene profiles. Biclustering of microarray gene expression data has recently been introduced by Chen & Church [1] as a means to discover sets of genes that co-expressed in only part of the experiment conditions under study. Essentially, overlapping in gene clusters is allowed, and many subtle gene clusters are revealed. Since then, several other algorithms have been developed to bicluster gene expression data [4]. However, existing biclustering algorithms do not consider the differences between time-series gene expression data and multi-condition gene expression data. The relations between time points are ignored, and the time points are clustered independently. It is marginally biologically meaningful if two genes show similar expression pattern in non-consecutive time points. It is therefore necessary to preserve the time locality in time-course gene expression data.

Stochastic modeling of volume images with a 3-D hidden Markov model

Abstract: Over the years, researchers in the image analysis community have successfully used various statistical modeling methods to segment, classify and annotate digital images. In this paper, we propose a 3-D hidden Markov model (HMM) for volume image modeling. A computationally efficient algorithm is developed to estimate the model. The 3-D HMM is applied to volume image segmentation and tested using synthetic images with ground truth. Experiments have demonstrated that 3-D HMM outperforms Gaussian mixture model based clustering by an order of magnitude in accuracy.