基于自适应特性二维经验模式分解的Retinex彩色图像增强.docx

基于自适应特性二维经验模式分解的Retinex彩色图像增强
摘要
Retinex彩色图像增强已成为计算机视觉和图像处理领域的一个重要问题。尽管目前已经有许多优秀的Retinex增强算法被提出，但它们在保留图像细节和增强背景特征等方面仍然存在一些问题。为了克服这些问题，本文提出了一种新的基于自适应特性二维经验模式分解的Retinex彩色图像增强方法。该方法利用Retinex算法和双边滤波器来分离彩色图像的光照分量和反射分量。接下来，通过自适应特性二维经验模式分解，将反射分量进一步分解为基函数和模态系数。最后，通过调整模态系数以达到增强彩色图像的目的，该方法达到了一定的增强效果。实验结果表明，所提出的方法在保留细节和背景特征等方面具有较好的效果。
关键词：Retinex；彩色图像增强；二维经验模式分解；自适应特性
Introduction
Retinexisaclassicaltheoryincolorimageenhancement,whichwasproposedbyLandandMcCannin1971.TheRetinextheoryaimstoseparatetheluminanceandreflectancecomponentsofanimage.Bydoingso,theRetinextheorycanenhancethevisualqualityofanimagewithoutlossofdetailorcolordistortion.Retinexenhancementalgorithmshavebeenextensivelystudiedandhavebecomeanimportanttopicincomputervisionandimageprocessing.
However,existingRetinexenhancementalgorithmsstillhavesomeproblems,suchasdifficultyinpreservingimagedetailsandenhancingbackgroundfeatures.Toovercometheseproblems,weproposeanewRetinexcolorimageenhancementmethodbasedonadaptivecharacteristictwo-dimensionalempiricalmodedecomposition.
Methodology
Theproposedmethodconsistsoftwoparts:theseparationofluminanceandreflectancecomponentsusingRetinexandtheenhancementofthereflectancecomponentusingadaptivecharacteristictwo-dimensionalempiricalmodedecomposition.
Firstly,weseparatetheluminanceandreflectancecomponentsofthecolorimageusingRetinexandbilateralfiltering.Thebilateralfiltercansmooththeimagewhilepreservingitsedges.TheRetinextheorycanseparatetheluminanceandreflectancecomponentsbasedonthelogarithmictransformationoftheimage.Thelogarithmictransformationisdefinedasfollows:
log(R+1)=log(I+1)-log(G+1)-log(B+1)
whereR,G,Barethered,greenandbluecolorvaluesoftheimage,respectively.
Secondly,weuseadaptivecharacteristictwo-dimensionalempiricalmodedecompositiontofurtherdecomposethereflectancecomponentintobasicfunctionsandmodalcoefficients.Thedecompositionprocessisadaptive,whichmeansthatthealgorithmdynamicallyadjuststhedecompositionparametersbasedonthecharacteristicsoftheimage.
Finally,weadjustthemodalco