Paper Info
Title
Entropic selection of histogram features for efficient classification
Abstract
This paper addresses the problem of local histogram-based image feature selection for learning binary classifiers. We show a novel technique which efficiently combines histogram feature projection with the conditional mutual information (CMI) based classifier selection scheme. Moreover, we investigate cost-sensitive modifications of the CMI-based selection procedure, which further improves the classification performance. Extensive evaluations show that the proposed methods are suitable for object detection and recognition tasks.
Year
DOI
Venue
2012
10.1007/978-3-642-34166-3_29
SSPR/SPR
Keywords
Field
DocType
novel technique,binary classifier,histogram feature projection,entropic selection,extensive evaluation,local histogram-based image feature,classifier selection scheme,cmi-based selection procedure,cost-sensitive modification,conditional mutual information,efficient classification,classification performance
Data mining,Histogram,Pattern recognition,Feature selection,Feature (computer vision),Histogram matching,Adaptive histogram equalization,Mutual information,Artificial intelligence,Balanced histogram thresholding,Conditional mutual information,Mathematics
Conference
Volume
ISSN
Citations 
7626
0302-9743
0
PageRank 
References 
Authors
0.34
13
1
Name
Order
Citations
PageRank
Ákos Utasi1496.40