Results 1  10
of
17
Unsupervised learning of finite mixture models
 IEEE Transactions on pattern analysis and machine intelligence
, 2002
"... AbstractÐThis paper proposes an unsupervised algorithm for learning a finite mixture model from multivariate data. The adjective ªunsupervisedº is justified by two properties of the algorithm: 1) it is capable of selecting the number of components and 2) unlike the standard expectationmaximization ..."
Abstract

Cited by 267 (20 self)
 Add to MetaCart
AbstractÐThis paper proposes an unsupervised algorithm for learning a finite mixture model from multivariate data. The adjective ªunsupervisedº is justified by two properties of the algorithm: 1) it is capable of selecting the number of components and 2) unlike the standard expectationmaximization (EM) algorithm, it does not require careful initialization. The proposed method also avoids another drawback of EM for mixture fitting: the possibility of convergence toward a singular estimate at the boundary of the parameter space. The novelty of our approach is that we do not use a model selection criterion to choose one among a set of preestimated candidate models; instead, we seamlessly integrate estimation and model selection in a single algorithm. Our technique can be applied to any type of parametric mixture model for which it is possible to write an EM algorithm; in this paper, we illustrate it with experiments involving Gaussian mixtures. These experiments testify for the good performance of our approach. Index TermsÐFinite mixtures, unsupervised learning, model selection, minimum message length criterion, Bayesian methods, expectationmaximization algorithm, clustering. æ 1
Image classification for contentbased indexing
 IEEE Transactions on Image Processing
, 2001
"... Abstract—Grouping images into (semantically) meaningful categories using lowlevel visual features is a challenging and important problem in contentbased image retrieval. Using binary Bayesian classifiers, we attempt to capture highlevel concepts from lowlevel image features under the constraint ..."
Abstract

Cited by 155 (2 self)
 Add to MetaCart
Abstract—Grouping images into (semantically) meaningful categories using lowlevel visual features is a challenging and important problem in contentbased image retrieval. Using binary Bayesian classifiers, we attempt to capture highlevel concepts from lowlevel image features under the constraint that the test image does belong to one of the classes. Specifically, we consider the hierarchical classification of vacation images; at the highest level, images are classified as indoor or outdoor; outdoor images are further classified as city or landscape; finally, a subset of landscape images is classified into sunset, forest, and mountain classes. We demonstrate that a small vector quantizer (whose optimal size is selected using a modified MDL criterion) can be used to model the classconditional densities of the features, required by the Bayesian methodology. The classifiers have been designed and evaluated on a database of 6931 vacation photographs. Our system achieved a classification accuracy of 90.5 % for indoor/outdoor, 95.3 % for city/landscape, 96.6 % for sunset/forest & mountain, and 96 % for forest/mountain classification problems. We further develop a learning method to incrementally train the classifiers as additional data become available. We also show preliminary results for feature reduction using clustering techniques. Our goal is to combine multiple twoclass classifiers into a single hierarchical classifier. Index Terms—Bayesian methods, contentbased retrieval, digital libraries, image content analysis, minimum description length, semantic
Automatic image orientation detection
 in Proc. IEEE ICIP’99
"... Abstract—We present an algorithm for automatic image orientation estimation using a Bayesian learning framework. We demonstrate that a small codebook (the optimal size of codebook is selected using a modified MDL criterion) extracted from a learning vector quantizer (LVQ) can be used to estimate the ..."
Abstract

Cited by 25 (3 self)
 Add to MetaCart
Abstract—We present an algorithm for automatic image orientation estimation using a Bayesian learning framework. We demonstrate that a small codebook (the optimal size of codebook is selected using a modified MDL criterion) extracted from a learning vector quantizer (LVQ) can be used to estimate the classconditional densities of the observed features needed for the Bayesian methodology. We further show how principal component analysis (PCA) and linear discriminant analysis (LDA) can be used as a feature extraction mechanism to remove redundancies in the highdimensional feature vectors used for classification. The proposed method is compared with four different commonly used classifiers, namelynearest neighbor, support vector machine (SVM), a mixture of Gaussians, and hierarchical discriminating regression (HDR) tree. Experiments on a database of 16 344 images have shown that our proposed algorithm achieves an accuracy of approximately 98 % on the training set and over 97% on an independent test set. A slight improvement in classification accuracy is achieved by employing classifier combination techniques. Index Terms—Bayesian learning, classifier combination, expectation maximization, feature extraction, hierarchical discriminant regression, image database, image orientation, learning vector quantization, support vector machine. I.
Hierarchical stochastic image grammars for classification and segmentation
 IEEE Trans. Image Processing
, 2006
"... Abstract—We develop a new class of hierarchical stochastic image models called spatial random trees (SRTs) which admit polynomialcomplexity exact inference algorithms. Our framework of multitree dictionaries is the starting point for this construction. SRTs are stochastic hidden tree models whose l ..."
Abstract

Cited by 16 (3 self)
 Add to MetaCart
Abstract—We develop a new class of hierarchical stochastic image models called spatial random trees (SRTs) which admit polynomialcomplexity exact inference algorithms. Our framework of multitree dictionaries is the starting point for this construction. SRTs are stochastic hidden tree models whose leaves are associated with image data. The states at the tree nodes are random variables, and, in addition, the structure of the tree is random and is generated by a probabilistic grammar. We describe an efficient recursive algorithm for obtaining the maximum a posteriori estimate of both the tree structure and the tree states given an image. We also develop an efficient procedure for performing one iteration of the expectationmaximization algorithm and use it to estimate the model parameters from a set of training images. We address other inference problems arising in applications such as maximization of posterior marginals and hypothesis testing. Our models and algorithms are illustrated through several image classification and segmentation experiments, ranging from the segmentation of synthetic images to the classification of natural photographs and the segmentation of scanned documents. In each case, we show that our method substantially improves accuracy over a variety of existing methods. Index Terms—Dictionary, estimation, grammar, hierarchical model, image classification, probabilistic contextfree grammar, segmentation, statistical image model, stochastic contextfree grammar, tree model. I.
Unsupervised selection and estimation of finite mixture models
 in Proc. Int. Conf. Pattern Recognition
, 2000
"... We describe a new method for fitting mixture models to multivariate data which performs component selection and does not require external initialization. The novelty of our approach includes: an MMLlike (minimum message length) model selection criterion; inclusion of the criterion into the expectat ..."
Abstract

Cited by 13 (3 self)
 Add to MetaCart
We describe a new method for fitting mixture models to multivariate data which performs component selection and does not require external initialization. The novelty of our approach includes: an MMLlike (minimum message length) model selection criterion; inclusion of the criterion into the expectationmaximization (EM) algorithm (increasing its ability to escape from local maxima); an initialization strategy supported on the interpretation of EM as a selfannealing algorithm. 1.
A probability hypothesis densitybased multitarget tracker using multiple bistatic range and velocity measurements
 Proceedings of the ThirtySixth Southeastern Symposium on , March 1416, 2004
, 2004
"... Ronald Mahler’s Probability Hypothesis Density (PHD) provides a promising framework for the passive coherent location of targets observed via multiple bistatic radar measurements. We apply a particle filter implementation of the Bayesian PHD filter to target tracking using both range and Doppler mea ..."
Abstract

Cited by 13 (1 self)
 Add to MetaCart
Ronald Mahler’s Probability Hypothesis Density (PHD) provides a promising framework for the passive coherent location of targets observed via multiple bistatic radar measurements. We apply a particle filter implementation of the Bayesian PHD filter to target tracking using both range and Doppler measurements from a simple nondirectional receiver that exploits noncoöperative FM radio transmitters as its “illuminators of opportunity”. Signaltonoise ratios, probabilities of detection and false alarm and bistatic range and Doppler variances are incorporated into a realistic twotarget scenario. Bistatic range cells are used in calculating the birth particle proposal density. The tracking results are compared to those obtained when the same tracker is used with rangeonly measurements. This is done for two different probabilities of false alarm. The PHD particle filter handles ghost targets well and has improved tracking performance when incorporating Doppler measurements along with the range measurements. This improved tracking performance, however, comes at the cost of requiring more particles and additional computation. A. The PHD and Passive Radar I.
Online Handwriting Recognition Using Multiple Pattern Class Models
, 2000
"... The field of personal computing has begun to make a transition from the desktop to handheld devices, thereby requiring input paradigms that are more suited for single hand entry than a keyboard and recent developments in online handwriting recognition allow for such input modalities. Data entry usin ..."
Abstract

Cited by 10 (1 self)
 Add to MetaCart
The field of personal computing has begun to make a transition from the desktop to handheld devices, thereby requiring input paradigms that are more suited for single hand entry than a keyboard and recent developments in online handwriting recognition allow for such input modalities. Data entry using a pen forms a natural, convenient interface. The large number of writing styles and the variability between them makes the problem of writerindependent unconstrained handwriting recognition a very challenging pattern recognition problem. The stateoftheart in online handwriting recognition is such that it has found practical success in very constrained problems. In this thesis, a method of identifying different writing styles, referred to as lexemes, is described. Approaches for constructing both nonparametric and parametric classifiers are described that take advantage of the identified lexemes to f...
Probabilistic Modeling of SingleTrial fMRI Data
 In Second International Conference on medical Image Computing and CompuerAssisted Intervention
, 2000
"... This paper describes a probabilistic framework for modeling singletrial functional magnetic resonance (fMR) images based on a parametric model for the hemodynamic response and Markov random field (MRF) image models. The model is fitted to image data by maximizing a lower bound on the log likelihood ..."
Abstract

Cited by 8 (0 self)
 Add to MetaCart
This paper describes a probabilistic framework for modeling singletrial functional magnetic resonance (fMR) images based on a parametric model for the hemodynamic response and Markov random field (MRF) image models. The model is fitted to image data by maximizing a lower bound on the log likelihood. The result is an approximate maximum a posteriori estimate of the joint distribution over the model parameters and pixel labels. Examples show how this technique can used to segment twodimensional (2D) fMR images, or parts thereof, into regions with different characteristics of their hemodynamic response. Index TermsHemodynamic response, image segmentation, Markov random field, mean field theory. I.
An ExpectationMaximisation Framework for Segmentation and Grouping
 Image and Vision Computing
, 2002
"... This paper casts the problem of perceptual grouping into an evidence combining setting using the apparatus of the EM algorithm. We are concerned with recovering a perceptual arrangement graph for linesegments using evidence provided by a raw perceptual grouping field. The perceptual grouping proces ..."
Abstract

Cited by 5 (3 self)
 Add to MetaCart
This paper casts the problem of perceptual grouping into an evidence combining setting using the apparatus of the EM algorithm. We are concerned with recovering a perceptual arrangement graph for linesegments using evidence provided by a raw perceptual grouping field. The perceptual grouping process is posed as one of pairwise relational clustering. The task is to assign linesegments (or other image tokens) to clusters in which there is strong relational affinity between token pairs. The parameters of our model are the cluster memberships and the pairwise affinities or linkweights for the nodes of a perceptual relation graph. Commencing from a simple probability distribution for these parameters, we show how they may be estimated using the apparatus of the EM algorithm. The new method is demonstrated on linesegment
Summarising contextual activity and detecting unusual inactivity in a supportive home environment
 PATTERN ANALYSIS APPLICATION
, 2004
"... Interpretation of human activity and the detection of associated events are eased if appropriate models of context are available. A method is presented for automatically learning a contextspecific spatial model in terms of semantic regions, specifically inactivity zones and entry zones. Maximium a ..."
Abstract

Cited by 4 (0 self)
 Add to MetaCart
Interpretation of human activity and the detection of associated events are eased if appropriate models of context are available. A method is presented for automatically learning a contextspecific spatial model in terms of semantic regions, specifically inactivity zones and entry zones. Maximium a posteriori estimation of Gaussian mixtures is used in conjunction with minumum description length for selection of the number of mixture components. Learning is performed using EM algorithms to maximise penalised likelihood functions that incorporate prior knowledge of the size and shape of the semantic regions. This encourages a onetoone correspondence between the Gaussian mixture components and the regions. The resulting contextual model enables humanreadable summaries of activity to be produced and unusual inactivity to be detected. Results are presented using overhead camera sequences tracked using a particle filter. The method is developed and described within the context of supportive home environments which have as their aim the extension of independent, quality living for older people.