home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Atari Mega Archive 1
/
Atari Mega Archive - Volume 1.iso
/
graphics
/
aim_xtra.arc
/
AIMTUTOR.DOC
< prev
next >
Wrap
Text File
|
1988-04-07
|
46KB
|
957 lines
06600020202800
1AIM TUTORIALPage # ofFile: AIMTUTOR.DOC
2
F0110030
9[...................................................]011
ü
ëAIMTUTORIAL
9[...................................................]011
Ç
by
TomHoeksmaandAdHerweijer
DelftUniversityofTechnology
FacultyofAppliedPhysics
PatternRecognitionGroup
Lorentzweg1
2628CJDELFT
theNetherlands
ëIntroductionê
Ç
TheimageprocessingsoftwarepackageAIM(AtariImageManager)
hasbeendevelopedatDelftUniversityofTechnologybyFrans
Groen(the'imageprocessor')andwasadaptedfortheAtariST
computerincooperationwithRobertdeVries(theGEMprogram⑨
mer).BothweremembersoftheSignal-/System-techniquesGroupin
theFacultyofAppliedPhysics,Fransasaseniorstaffmember
andRobertasagraduatestudent.Recently,Franschangedjobsto
becomeprofessorofAppliedInformaticsintheUniversityof
Amsterdam.
AIMisdesignedtobeanentrylevel'appetizer'fortheartof
imageprocessing.Itgivestheinterestedlaypersonan
opportunitytoacquainthimselfwiththetechniquesand
terminologyusedinthisdynamicfield.Theexperiments
describedmayberegardedasalowerboundofwhatcanbe
accomplishedwithAIM,theupperboundbeinglimitedbythe
user'simaginationonly.
Thistutorialisnotafullcourseondigitalimageprocessing.
SuchcoursesarefrequentlyorganizedbythePatternRecognition
Group.Forcourseinformation,pleasecontacttheSecretaryof
thePatternRecognitionGroup,FacultyofAppliedPhysics,
Lorentzweg1,2628CJDelft,theNetherlands,Phone015-781416.
Notethattheoperationsusedintheexamplesanddemoscovera
minorpartofAIM'scapabilitiesonly.Acomprehensivedescrip⑨
tionofalltheoperationsavailableisgiveninafilecalled
AIMANUAL.DOC.Likethistutorialitiscontainedinacompressed
file(DOCS_IMS.ARC),whichcanbefoundinthefolderCOMPRESSon
thedistributiondisk(togetherwithadecompressingprogram).
Imageprocessingisnotparticularlydifficult;manypeoplewill
havehadtheirfirstimageprocessingexperienceinthe
photographicdarkroom.Note,however,thatAIM,beingdedicated
todigitalimageprocessing,requiressomebasicknowledgeof
digitalsystems.Sothoseofyouwhountilnowhavesuccessfully
avoidedlearningaboutbits,bytesandbinarynumbers,be
preparedtofacetheunavoidable!The15minutesorsospenton
readingabouttheseconceptswillberichlyrewardedbyabetter
understandingofAIM'soperations,ifnotbythedeepadmiration
ofyourenvironmentforthenewlyacquiredvocabulary.
AIMisaresultofmanyyearsofresearchanddevelopmentinthe
PatternRecognitionGroup.Manyoftheimageprocessingroutines
werecontributedby(a.o.)GertvanAntwerpen,FransGroen,Piet
Verbeek,BenVerwer,LucasvanVlietandIgorWeber.This
introductionwaswritten(ifnotcompiledfromexistingtexts)by
TomHoeksmaandAdHerweijer.
PreviousversionsofAIMhavealreadyextensivelypenetratedthe
AtariSTcommunity.However,asthedocumentationsuppliedwith
thosefirstpublicdomainversionswasratherlimited,the
explorationofthesoftwarehasbeenamere'imageprocessing
adventure'tomostofthe'earlybirds'.Theauthorshopethat
thisintroduction,togetherwiththerelatedexamplesanddemos
includedonthedisk,willenableallprospectiveusers(early
birdsandnewcomersalike)tobetterunderstandwhatwas,resp.
isgoingon.
Ç
Feedbackplaysanimportantroleintheeffectiveuseofimage
processingaswellasinsoftwaredevelopment.Therefore,we
inviteyoutosendyoureventualcommentsandsuggestionsto:
Prof.dr.ir.F.C.A.Groen
c/oDelftUniversityofTechnology
FacultyofAppliedPhysics
PatternRecognitionGroup
Lorentzweg1,2628CJDELFT
theNetherlands
ë
ëWHATCANWEEXPECTOFIMAGEPROCESSING?Ç
9[...................................................]011
MaybeyouhavealreadyhadsomeexperiencewithAIMorseen
somebodyelseworkwithit.Youmayhavewondered,then,how
imageprocessingcouldmakefeaturesappearinthedisplaythat
weren'ttherebeforetheoperation.Textthatatfirstwas
unreadable,suddenlycouldbereadeasily;frecklesinagirl's
face,thatdidn'tshowintheoriginaldisplay,could(toher
distress!)becountedaftersomeappropriatekeystrokes.Youwere
perhapsconvincedtohaveacquiredalittlemagician,hiddenin
thattinyfloppydisk.
Bewarned,however!Imageprocessingcanneverreproducefeatures
oftheoriginalimage,thatsomehowhavebeenlost.So,be
preparedforsurprises,butdon'texpectmiracles!Image
processingcanenhanceparticulardetailsinanimage,but
usuallyotherdetailswillhavetobesacrificed.
If'Apictureisworthathousandwords',imageprocessingallows
youtoarrangethosewordstomakethepicturerevealits
(secret)'stories'.Butremember:onepicture'sthousandwords
aredifferentfromanother's.Everystorywillinheritthe
characteristicsofthepictureitwasextractedfrom.Theartof
imageprocessingis:tomakethepicturetelltheonestory
that'susefultoyou.
ë
STAGESINANIMAGEPROCESSINGAPPLICATIONÇ
Inalmosteveryimageprocessingapplicationwecanrecognize
somestages,thateventuallyleadtothefinalresult:ürecordingÇ,
üpreprocessingÇ,üsegmentationÇ,üpostprocessingÇ,üanalysisÇandüinter⑨Ç
üpretationÇ.
Probablythemostimportantstageistherecordingofthe
originalpicture.Asstatedabove,informationthatgetslostin
therecordingprocesscanneverberecovered.Fortherecordinga
video-cameraisrequired,capableofconvertingthelight,
reflectedfromtheobject,intoanelectricalsignal.Asthe
videosignalcannotbereaddirectlybythecomputer,thissignal
mustbesampledanddigitizedandthedigitalinformationmustbe
transferredinanorderlymannertothecomputer'smemory.This
isdonebyaspecialpieceofhardware,aso-called'üframe
grabberÇ'.VariousframegrabbersareavailablefortheAtariST.
InthePatternRecognitionGroupmonochromeandcolorframe
grabbershavebeendevelopedthateventuallywillbemade
commerciallyavailable.
Aftertheimageisstoredinthecomputeritmustbepre⑨
processed.Heretheusertriestoenhanceorfilterout
particularwantedresp.unwantedfeatures,tosatisfy
requirementsoffurtheranalysis.Examplesofpreprocessingare
thecorrectionforchanginggreyvaluesinthebackground
(shading),suppressingnoise,determinationofedgesinthe
image,etc.
ÇInordertobeabletodeterminepropertiesofsingleobjectsin
theimage,theseobjectsmustfirstbemadedetectable.Ifagrey
valueimagecontainsdarkobjectsonalightbackground,
segmentationcan,forinstance,resultinabinaryimagewith
whiteobjectsinablackbackground.
Postprocessingcanimprovetheresultofthesegmentation.
Examplesarethemorfological(binary)operations.Holesin
objectscanbeclosed,edgespeeledoff,etc.
Intheanalysisstage,measurementscanbeperformedonthe
segmentedobjects.Examples:area,perimeter,curvature.
Intheinterpretationstagetheresultsofthemeasurementsare
evaluated.
An'intelligentrobot'inafactor