| 期刊名称： |
Geoscience and Remote Sensing Letters |
| 全部作者： |
Linlin Shen,Zexuan Zhu,Sen Jia,Jiasong Zhu*,Yiwen Sun |
| 出版年份： |
2013 |
| 卷 号： |
10 |
| 期 号： |
1 |
| 页 码： |
29-33 |
| 查看全本： |
|
Three-dimensionalGaborwaveletshaverecentlybeensuccessfullyappliedforhyperspectralimageclassificationduetotheirabilitytoextractjointspatialandspectruminformation.However,thedimensionoftheextractedGaborfeatureisincrediblyhuge.Inthisletter,weproposeasymmetrical-uncertainty-basedandMarkov-blanket-basedapproachtoselectinformativeandnonredundantGaborfeaturesforhyperspectralimageclassification.TheextractedGaborfeatureswithlargedimensionarefirstrankedbytheirinformationcontainedforclassificationandthenaddedonebyoneafterinvestigatingtheredundancywithalreadyselectedfeatures.TheproposedapproachwasfullytestedonthewidelyusedIndianPinesitedata.Theresultsshowthattheselectedfeaturesaremuchmoreefficientandcanachievesimilarperformancewithpreviousapproachusingonlyhundredsoffeatures.
