Results 1 
6 of
6
On the Efficient Evaluation of Probabilistic Similarity Functions for Image Retrieval
 IEEE Trans. Inf. Theory
, 2004
"... Probabilistic approaches are a promising solution to the image retrieval problem that, when compared to standard retrieval methods, can lead to a significant gain in retrieval accuracy. However, this occurs at the cost of a significant increase in computational complexity. In fact, closedform solut ..."
Abstract

Cited by 28 (1 self)
 Add to MetaCart
Probabilistic approaches are a promising solution to the image retrieval problem that, when compared to standard retrieval methods, can lead to a significant gain in retrieval accuracy. However, this occurs at the cost of a significant increase in computational complexity. In fact, closedform solutions for probabilistic retrieval are currently available only for simple probabilistic models such as the Gaussian or the histogram. We analyze the case of mixture densities and exploit the asymptotic equivalence between likelihood and Kullback–Leibler (KL) divergence to derive solutions for these models. In particular, 1) we show that the divergence can be computed exactly for vector quantizers (VQs) and 2) has an approximate solution for Gauss mixtures (GMs) that, in highdimensional feature spaces, introduces no significant degradation of the resulting similarity judgments. In both cases, the new solutions have closedform and computational complexity equivalent to that of standard retrieval approaches.
Generalized Model Selection For Unsupervised Learning In High Dimensions
 Proceedings of Neural Information Processing Systems
, 1999
"... In this paper we describe an approach to model selection in unsupervised learning. This approach determines both the feature set and the number of clusters. To this end we first derive an objective function that explicitly incorporates this generalization. We then evaluate two schemes for model sele ..."
Abstract

Cited by 16 (2 self)
 Add to MetaCart
In this paper we describe an approach to model selection in unsupervised learning. This approach determines both the feature set and the number of clusters. To this end we first derive an objective function that explicitly incorporates this generalization. We then evaluate two schemes for model selection  one using this objective function (a Bayesian estimation scheme that selects the best model structure using the marginal or integrated likelihood) and the second based on a technique using a crossvalidated likelihood criterion. In the first scheme, for a particular application in document clustering, we derive a closedform solution of the integrated likelihood by assuming an appropriate form of the likelihood function and prior. Extensive experiments are carried out to ascertain the validity of both approaches and all results are verified by comparison against ground truth. In our experiments the Bayesian scheme using our objective function gave better results tha n crossvalidatio...
Contentbased browsing and edition of unstructured video
 in Proc. IEEE ICME
, 2000
"... The focus of this paper is to build a set of tools that analyze, characterize, and prepare footage shot with a camera for unanticipated uses. The idea is to take this ensemble of raw stuff and use it in interesting ways that are not just making a sequence out of them. Some of the possible uses of su ..."
Abstract

Cited by 8 (0 self)
 Add to MetaCart
The focus of this paper is to build a set of tools that analyze, characterize, and prepare footage shot with a camera for unanticipated uses. The idea is to take this ensemble of raw stuff and use it in interesting ways that are not just making a sequence out of them. Some of the possible uses of such unstructured video are: Make a time slice of a person or a series of events, select a group of shots and make a postcard out of them, make a collage of shots, make a SalientStill TM, storyboard etc. We employ novel distribution clustering algorithms to enable browsing unstructured video. The browser permits navigation of content and extraction of stills, collages and summaries from unstructured video. 1.
Humane Interfaces to Video
 In: Proc. 32 nd Asilomar Conf. on Signals, Systems, and Computers. Asilomar, CA, invited Paper
, 1998
"... We present approaches for content characterization along the natural divisions of structured and nonstructured video. We propose probabilistic clustering techniques as an alternative to querybyexample in the unstructured domain, and a Bayesian formulation that exploits knowledge about structure w ..."
Abstract
 Add to MetaCart
We present approaches for content characterization along the natural divisions of structured and nonstructured video. We propose probabilistic clustering techniques as an alternative to querybyexample in the unstructured domain, and a Bayesian formulation that exploits knowledge about structure when it is available. 1. Introduction The explosion in availability of image and video content, due to the high interconnectivity of the new digital media and recent developments in multimedia technology,demands the formulation of powerful paradigms for automated contentbased characterization and efficient algorithms for access into large image and video repositories. Four components are needed in any system that aims to accomplish these goals: a representation that permits ready perusal, a set of robust techniques to select appropriate footage, an interface that maps the analysis to human terms, and an application context in which to work. The application context constrains the choice of re...
Optimization CBIR using KMeans Clustering for Image Database
"... implementation in searching images into image database requires usually for a sufficiently prolonged time because such image searching process is performed with comparison between searched images and individually records in an image database. In this work, it is proposed a KMeans clustering algorit ..."
Abstract
 Add to MetaCart
implementation in searching images into image database requires usually for a sufficiently prolonged time because such image searching process is performed with comparison between searched images and individually records in an image database. In this work, it is proposed a KMeans clustering algorithm aiming to develop clusters from each image database records, it can later be used for optimizing image searching access period. The stored images in image database records are only limited for the JPEGtype images. In this algorithm, cluster formation is based on maximum and minimum PSRN’s (Peak Signal to Noise Ratio) calculation values from individual records on a basic images and it will be treated as key images in every search for records with such cluster utilization. Results of the clustering process in form of cluster table would be made as indexing in early image searching for cluster position determination from searched images to image records.