Treść książki
Przejdź do opcji czytnikaPrzejdź do nawigacjiPrzejdź do informacjiPrzejdź do stopki
vii
caseofafirst-orderneighbourhood,appearedtobeidenticaltotheSobeledge
detector.
ContentsThebookisdividedintotwopartsandsevenchapters.
IntheIntroduction(Chapter1),theaccumulationprincipleisformulated.
Thisprinciplesetsthedirectionfollowedintheremainingchapters.Thedispersed
characteroftheimageinformationischaracterisedandillustratedwithimages
someofwhichwillbeanalysedinthefollowingpartsofthebook.
InPartIthefoundationsofthemethodsusedaregiven.
InChapter2thenecessarynotionsareintroduced,startingwiththemea-
surementandthehistogramandconcludingwiththebasicversionoftheHough
transform.Relationsofaccumulationandstatisticsarebriefedandsomeextended
versionsoftheHT,tobeusedlater,areshown.
InChapter3theproblemofestimationwiththeuseofahistogramisdiscussed.
Fuzzificationofahistograminthecaseofaperiodicandperiodicdataispresented,
accordingmainlyto[Chm06].Thenewnotionsofthelimitfuzzificationandthe
degreeoffuzzificationareintroducedinthischapter,andtherecommendations
onthedegreeoffuzzification,furtherusedinspecifyingthescaleofthefuzzifica-
tionfunctionsuchthatthesolutionsobtainedwiththeaccumulationmethodsare
robustagainstnoiseinthedata,areformulated.
PartIIisdevotedtoparticularproblemssolvedwiththeaccumulationmeth-
ods.
InChapter4asimpleaccumulationedgedetectorisintroduced.Thisdetector
istreatedasanexerciseinthedesignofaccumulationmethods.Relationsbetween
thethreevariantsofthedetector–theaccumulation,themedianandtheaveraging
one–areshown.Thelattervariantofthedetectorappearedtobeequivalentto
theSobeldetector.
Inthefollowingchaptersthesolutionsoftwopracticalproblemshavebeen
presented,oneparametricandonenon-parametric.
Chapter5discussestheimageregistrationproblem,intheversionwithonlythe
locationsofthepixelstoregistergiven.Threemethodsknownfromtheliterature
arecompared.Thequickestofthemappearedtobethelessrobust.Comparisons
andanexampleofamedicalapplicationofthemostrobustmethodhavebeen
foundedonthepaper[Chm04a].Theattentionisfocusedontheverificationofthe
recommendationsconcerningthedegreeoffuzzificationformulatedbefore.The
resultispositive:inthetestproblemstherobustnessuptoaround70-80%share
ofnoiseinthedatahasbeenattained.
Chapter6inwhichthedetectoroflines,thatis,theelongatedobjects,has
beendescribed,completesthebook.Thepresentednewaccumulationmethodis
anon-parametricdetectorworkingintheimagespace.Thedetectorisdescribed
andthoroughlytestedintheapplicationtoartificialtestimages.Resultsforsev-
eralselectedmammographicimagescontainingobjectsdifficulttodetectarealso
shown.Thedetectorappearedtoberobustagainstnormalnoisewhichcanbe
describedastwiceasstrongastheimagesignal(SNR=−3dB).Inthecaseof
anormalnoisemixedwithastrongpointnoisetheresultsarenotworse.Such
noiselevels,accordingtoasubjectiveestimation,significantlyexceedthelevels
foundinmammograms,forexample.
