改进的小波阈值去噪算法-研究

《改进的小波阈值去噪算法-研究》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。

1、哈尔滨理工大学理学硕士学位论文TheStudyonImprovedWaveletThresholdDenoisingMethodAbstractWaveletthresholddenoisingmethodisasimpleandeffectivedenoisingmethod．buttosomeextent．therearesomedefectsonthetraditionalwaveletthresholdmethodinpracticalapplication．Therefore，itisquitenecessarytodothefurtherresearchonthewavelet

2、thresholdmethod．Thetraditionaluniversalthresholdisfixedoneachscaleanddoesnothavetheadaptivecapacity,whichhasagreatimpactonthedenoisingeffect．Thehardthresholdfunctionisdiscontinuousatthethresholdpoints，forthesoftthresholdfunction，althoughitiscontinuousatthethresholdpoints，thereexiststheconstantdevi

3、ationbetweenthesoftthresholdfunctionandtherealsignal．Thewavelettransformiseffectiveforthedetectionofpointsingularity，butitcannotdealwiththelinesingularityoftheimageeffectively．However,thisdrawbackofwavelettransformCanbecoveredbytheridgelettransform．Thelinesingularitycanbedetectedwellbytheridgelett

4、ransform．Toachievethepurposeofremovingthenoiseeffectively，thewaveletthresholdmethodisimprovedinthispaper．Thespecificworkisasfollows：Firstly,theuniversalthresholdisimprovedinthispaperSOthatitCanbechangedwiththedecompositionlevel．Comparedwithuniversalthreshold，theimprovedthresholdisflexibleandaccord

5、stothedistributionofnoiseineachlayer．Thenthethresholdfunctionisimprovedinthispaper,aparameterisaddedtotheimprovedthresholdfunction．Byadjustingtheparameter，theimprovedthresholdfunctioniSbetweenthetwothresholdfunctions．itcanovercomediscontinuityofthehardthresholdfunctionandreducethedeviationbetweent

6、hesoftthresholdfunctionandtherealsignal．Theimprovedthresholdfunctioniscontinuousanddifferentiable．Inthisway，itisratherconvenienttodosomesubsequentprocessing．Secondly．thewaveletthresholdmethodcannotdealwiththeline哈尔滨理工大学理学硕士学位论文singularityeffectively，buttheridgelettransformcouldmakeupthisshortcomin

7、g．Theimageshavegreat”wrap—around”effect，afterridgeletmethodhasbeenusedtoremovenoise，SOthetraditionalridgeletdenoisingmethodisimprovedinthispaper．Firstofall，noisyimageisdividedintoimageblockswhicharethesamesizeand