数字图像处置冈萨雷斯英文版Chapter图像分割专题培训课件.ppt

数字图像处置冈萨雷斯英文版Chapter10图像分割

PreprocessImageacquisition,restoration,andenhancementIntermediateprocessImagesegmentationandfeatureextractionHighlevelprocessImageinterpretationandrecognitionISAN-DSPGROUPElementofImageAnalysis

Imagesegmentationisusedtoseparateanimageintoconstituentpartsbasedonsomeimageattributes.ImagesegmentationisanimportantstepinimageanalysisBenefit 1.Imagesegmentationreduceshugeamountofunnecessarydatawhileretainingonlyimportancedataforimageanalysis 2.ImagesegmentationconvertsbitmapdataintobetterstructureddatawhichiseasiertobeinterpretedImportanceofImageSegmentation

1.Similaritypropertiesofpixelsinsidetheobjectareusedtogrouppixelsintothesameset.2.Discontinuityofpixelpropertiesattheboundarybetweenobjectandbackgroundisusedtodistinguishbetweenpixelsbelongingtotheobjectandthoseofbackground.Discontinuity:IntensitychangeatboundarySimilarity:InternalpixelssharethesameintensityImageAttributesforImageSegmentation

SpatialFilteringApplicationtoShapeDetectionOneapplicationofspatialfilteringisshapedetection:findinglocationsofobjectswiththedesiredshape.Unlikefrequencyselectivemasksthataredesignedbasedontheconceptoffrequency,shapedetectionmasksarederivedfromtheshapestobedetectedthemselves.Amaskforshapedetectionusuallycontainstheshapeorapartoftheshapetobedetected.Thelocationthatismostcorrelatedtothemaskisthelocationwherethehighestfilterresponseoccurs.Theshapeismostlikelytoexistthere.

PointDetectionWecanuseLaplacianmasksforpointdetection.Laplacianmaskshavethelargestcoefficientatthecenterofthemaskwhileneighborpixelshaveanoppositesign.Thismaskwillgivethehighresponsetotheobjectthathasthesimilarshapeasthemasksuchasisolatedpoints.Noticethatsumofallcoefficientsofthemaskisequaltozero.Thisisdue