简介：Differentmodalitiesinbiomedicalimages,likeCT,MRIandPETscanners,providedetailedcross-sectionalviewsofhumananatomy.Thispaperintroducesthree-dimensionalbrainreconstructionbasedonCTslices.Itcontainsfiltering,fuzzysegmentation,matchingmethodofcontours,cellarraystructureandimageanimation.Experimentalresultshaveshownitsvalidity.TheinnovationismatchingmethodofcontoursandfuzzysegmentationalgorithmofCTslices.

简介：Thispaperproposesanovelexemplar-basedmethodforreducingnoiseincomputedtomography(CT)images.Intheproposedmethod,denoisingisperformedoneachblockwiththehelpofagivendatabaseofstandardimageblocks.Foreachnoisyblock,itsdenoisedversionisthebestsparsepositivelinearcombinationoftheblocksinthedatabase.Weformulatetheproblemasaconstrainedoptimizationproblemsuchthatthesolutionisthedenoisedblock.Experimentalresultsdemonstratethegoodperformanceoftheproposedmethodovercurrentstate-of-the-artdenoisingmethods,intermsofbothobjectiveandsubjectiveevaluations.

简介：在肺的断层图像中，由于肿瘤和组织之间复杂的关系，肿瘤的边缘往往是模糊的，尤其是在病理区，肿瘤附着于胸和肺的病例。这使得肿瘤分割更加困难。为了准确地对肺CT图像进行分割，提出了一种基于支持向量机的方法和改进的水平集模型。首先，将图像分为若干块单元，然后提取每个块的纹理、灰度和形状特征，然后对其进行训练以检测可疑病灶区域；最后，将可疑边缘提取为优化病变区域后的初始轮廓，并用形态学梯度修正的水平集模型。实验结果表明，该方法能有效快速地分割出具有较高精度的复杂肺部CT图像。