Combining classifiers

Kittler, Josef; Hater, Mary A.; Duin, Rpw

doi:10.1109/icpr.1996.547205

Cited by 672 publications

(875 citation statements)

References 13 publications

Supporting

Mentioning

846

Contrasting

Unclassified

Order By: Relevance

“…Given the posterior estimates from the individual classifiers, the goal of the combining stage is to produce a single estimate that maximizes the probability for localized object detection while reducing clutter and false alarms. Various integration methods have been proposed in the past [5]. We formulate a supra-Bayesian integration in which the posterior estimates from each classifier are assumed to have a probability distribution and, based on the means and variances of the outputs, we can formulate an optimal decision scheme.…”

Section: Bayesian Classificationmentioning

confidence: 99%

Segmenting Biological Particles in Multispectral Microscopy Images

Shah

2007

2007 IEEE Workshop on Applications of Computer Vision (WAC v '07)

View full text Add to dashboard Cite

This paper presents a methodology and results for segmentation of biological particles in multispectral images by learning disparate models from each spectra for pixel classification coupled with contour evolution based on the use of level set theory. Traditional contour models have some limitations on the segmentation of complicated images whose sub-regions consist of multiple components. The segmentation of multispectral images is even a more difficult problem. Our proposed model overcomes these limitations and uses multiple classifiers, each of which solves the problem independently based on its input observations. Each classifier module is trained to detect distinct regions and a higher order decision integrator collects evidence from each of the modules to delineate a final region.

show abstract

Section: Bayesian Classificationmentioning

confidence: 99%

Segmenting Biological Particles in Multispectral Microscopy Images

Shah

2007

2007 IEEE Workshop on Applications of Computer Vision (WAC v '07)

View full text Add to dashboard Cite

show abstract

“…. .,Xm) Pi(P(CilXi),P(CiX2),...,P(CiIXm)) (9) Considering that Pi has to obey the laws of probability if an integrated measure of the posterior is to be estimated, the integration may be written as:…”

Section: Multifeature Integrationmentioning

confidence: 99%

“…Various integration methods have been proposed in the past. 9 We formulate a supra-Bayesian integration in which the posterior estimates from each classifier are assumed to have a probability distribution and, based on the means and variances of the outputs, we can formulate an optimal decision scheme. Strictly speaking, Bayesian theory holds true only for individual decision makers, but if the group decision is viewed as a collaborative effort, the effect is externally Bayesian.…”

Section: Multifeature Integrationmentioning

confidence: 99%

Hierarchical multifeature integration for automatic target recognition

Shah

Aggarwal

1999

SPIE Proceedings

View full text Add to dashboard Cite

This paper presents a methodology for object recognition in complex scenes by learning multiple feature object representations in second generation Forward Looking InfraRed (FUR) images. A hierarchical recognition framework is developed which solves the recognition task by performing classification using decisions at the lower levels and the input features. The system uses new algorithms for detection and segmentation of objects and a Bayesian formulation for combining multiple object features for improved discrimination. Experimental results on a large database of FUR images is presented to validate the robustness of the system, and its applicability to FUR imagery obtained from real scenes.

show abstract

“…As the term ½ does not depend on , we may neglect it for classification purposes. Note that this motivation for the sum rule differs from that proposed by KITTLER in [8]. Using multiple classifiers to classify a single test pattern, he assumed that the a-posteriori probabilities computed by the respective classifiers do not differ much from the a-priori probabilities to justify the sum rule.…”

Section: The Virtual Test Sample Methodsmentioning

confidence: 99%

“…Many authors such as SCHWENK use TD within artificial neural nets [7]. Finally, the virtual test sample method proposed in Section 3 was motivated by KITTLER's research on classifier combination [8].…”

Section: Related Workmentioning

confidence: 99%