資源描述：
《基于雙正交基字典學(xué)習(xí)的圖像去噪方法》由會(huì)員上傳分享，免費(fèi)在線閱讀，更多相關(guān)內(nèi)容在行業(yè)資料-天天文庫(kù)。
1、基于雙正交基字典學(xué)習(xí)的圖像去噪方法摘要:為了提高圖像去除白高斯噪聲的性能，利用超完備字典作為圖像的稀疏表示。超完備字典的冗余性可以有效地表示圖像的各種幾何奇異特征。在貝葉斯框架下，以圖像塊的稀疏表示定義了全局圖像先驗(yàn)概率模型，給出了最大后驗(yàn)概率模型下的優(yōu)化圖像去噪算法。超完備字典使用兩個(gè)不同的正交基構(gòu)成，給出了基于奇異值分解（svd）的優(yōu)化字典計(jì)算方法。該方法充分利用正交基的特點(diǎn)，采用svd方法進(jìn)行高效的字典學(xué)習(xí)。基于雙正交基字典的去噪算法提高了圖像去噪性能，實(shí)驗(yàn)結(jié)果證實(shí)了所提方法的有效性。關(guān)鍵詞:圖像去噪；字典學(xué)習(xí)；稀疏表示；奇異值分解；貝葉斯估計(jì)imageden
2、oisingmethodbasedondictionarylearningwithunionoftwoorthonormalbasesxiekai*,zhangfenschoolofinformationandmechanicalelectronicengineering,beijinginstituteofgraphiccommunication,beijing102600,chinaabstract:overcompletedictionarywasusedtorepresentanimagesparselyfortheimprovementofimagede
3、noisingperformance.thesparserepresentationmayrepresentefficientlythesingulargeometryoftheimageswiththeredundancyofover-completedictionary.globalimagepriormodelbasedonthesparserepresentationofimagepatcheswaspresentedinbayesianframework.thenmaximumaposterioriprobabilityestimatorfordenoisin
4、gimagewasconstructed.thedictionaryconsistedofthetwoorthonormalbases.amethodbasedonsingularvaluedecompositionwasusedtodictionarylearning.theorthonormalpropertywasmadeuseoftoupdatetheonechosenbasiseffectively.themethodcanimprovetheperformanceofimagedenoising.experimentsresultsshowthevalidi
5、tyofthemethod.overcompletedictionarywasusedtorepresentanimagesparselyinordertoimproveimagedenoisingperformance.thesparserepresentationmayrepresentefficientlythesingulargeometryoftheimageswiththeredundancyofover.completedictionary.globalimagepriormodelbasedonthesparserepresentationofimage
6、patcheswaspresentedinbayesianframework.thenmaximumaposterioriprobabilityestimatorfordenoisingimagewasconstructed.thedictionarywascomposedoftwoorthonormalbases.amethodbasedonsingularvaluedecompositionwasusedfordictionarylearning.theorthonormalpropertywasusedtoupdatetheonechosenbasiseffect
7、ively.themethodcanimprovetheperformanceofimagedenoising.theexperimentalresultsverifythevalidityofthemethod.keywords:imagedenoising;dictionarylearning;sparserepresentation；singularvaluedecomposition(svd)；bayesianestimation圖像去噪的主要目標(biāo)就是把被污染噪聲的圖像恢復(fù)為原始圖像。小波變換用于圖像去噪是近十年來(lái)的主要研究方
