北邮国院视频图像处理Word文件下载.docx

北邮国院视频图像处理Word文件下载.docx

《北邮国院视频图像处理Word文件下载.docx》由会员分享，可在线阅读，更多相关《北邮国院视频图像处理Word文件下载.docx（29页珍藏版）》请在冰点文库上搜索。

PortableGreyMap."

Thename"

PPM"

istheacronymfor"

PortablePixelMap."

Imagesinthisformat（oraprecursorofit）wereoncealsocalled"

portablepixmaps."

Itisahighlyredundantformat,andcontainsalotofinformationthattheHumanVisualSystem（HVS）cannotevendiscern.However,asPGM,PPMisveryeasytowriteandanalyze.Thegoalofthefirstpartoftoday’slabistoletstudentsbecomecomfortablewiththesetwoformats.

YouwillimplementfunctionstoreadandtowritePPMandPGMimages.Thefinaldemonstrationoftheimplementedsoftwarewillbedoneusingwell-knowntestimages:

LENA,BABOON,PEPPERS,etc.YoucanfindPPMandPGMversionsoftheseimagesintheEBU723Uintranetpages.Thewritingfunctionmustaddascommentintheheader:

“imagecreatedbyyour_nameyour_surname”.

Includeinyoursubmissionthefileresultingfromreadingtheimagesprovidedandwritingthembackintheiroriginalformat.

Summarizein5pointstheoperationsnecessarytoreadaPGM/PPMimage:

1.Opentheimageinreadmode.

2.Ignorecommentsinthefile

3.DistinguishtheimageisaPGMorPPM.

4.Readinformationoftheimage.

5.Readmatrix.

Summarizein5pointstheoperationsnecessarytowriteaPGM/PPMimage:

1.Opentheimageinwritemode.

2.Writecommentsinthefile.

3.DecidewrittingaPGMorPPMimage.

4.Writeinformationoftheimage.

5.Writematrix.

WhatisthedifferencebetweentheidentifierP3andP6?

ForidentifierP3,theimageisstoredasASCIIandeachlineislessthan70characters.

ForidentifierP6,theimageisstoredasbinary.

P6formatissmallerthanP3formatandreadP6formatfileisfasterthanP3formatfile.

Exercise1（b）

Formatconversions:

inthispartofthelab,theimageswillbeconvertedfromcolourtogreyscale;

inotherwordsaPPMimagewillbeconvertedtothePGMformat.Youwillimplementafunctioncalled“BUPT_format_converter”whichtransformsimagesfromcolortogrey-scaleusingthefollowingYUVconversion:

Y=0.257R+0.504G+0.098B+16

U=0.439R+0.368G–0.071B+128

V=-0.148R–0.291G+0.439B+128

Whatcomponentrepresentstheluminance,i.e.thegrey-levels,ofanimage?

Yrepresentstheluminance.

Usetheeboxestodisplaytheresultsforthecolourtogrey-scaleconversion.

Lenacolour（RGB）

Lenagrey

Baboongrey

Babooncolour（RGB）

Isthetransformationbetweenthetwocolour-spaceslinear?

Motivateyouranswer

Thetransformationbetweenthetwocolour-spacesislinear.

Transformeachlayer,thencombineeachoutcome.OrTransformlayerstogether.Thesetwooutcomesaresame.

f（ax+by）=af（x）+bf（y）

DisplayintheboxtheLenaimageconvertedtoYUV3channelsformat.

Arethecolorsofthepreviouspicturedistorted?

Ifyeswhy?

Theyaredifferent.BecauseRGBusedforPCandYUVusedforTV.IfweshowYUVimageonPC,itwillbedistorted.

BasedontheformulafortheRGBtoYUVconversion,derivetheformulafortheYUVtoRGBconversion?

R=1.164*（Y-16）+1.596*（V-128）

G=1.164*（Y-16）-0.813*（V-128）-0.391*（U-128）

B=1.164*（Y-16）+2.018*（U-128）

UsetheformulayouderivedattheprevioussteptoconverttheYUVimagebacktotheoriginalRGBformat.Displaytheresultinthebox.

Exercise1（c）

Sub-sampling:

TheHVSisincapableofperceivingcertaindetailsinanimage.ThereforehighcompressionratioscanbeachievedbyexploitingthecharacteristicsoftheHVS,thusdiscardingwhathasalowvisualrelevance.However,thisprocesscanintroducedistortionsinthecompresseddata.AsimplewaytoexploitthecharacteristicsoftheHVScompressionpurposesistosub-sampleanimage.Adrawbackofthisapproachisthatitispossibletoincurinwell-knowproblemsofadiscreterepresentation,suchasaliasing.Thispartofthelabcoverssomesimplesub-samplingoperations.

Implementafunctionthatsub-samplesgreylevelimagesbyafactorn,withnamultipleof2.Thefunctionshouldbeabletosub-sampleindependentlyinthehorizontalandintheverticaldirectionorinbothdirectionsatthesametime.

Displaytheresultsofsub-samplingtheimageLenausingthefollowingfactors:

2horizontal,2vertical,2verticaland8horizontal,4verticaland4horizontal.Includethefilesoftheresultsinthesubmission.

Boxforthe4images

2horizontal

2vertical

2verticaland8horizontal

4verticaland4horizontal

Describeusingyourownwordsthealiasingproblemandhowtoavoidit,asappliedtosignalprocessing

Aliasingoccurswhenthesamplefrequencyisnothighenoughsothatathesampledsignalcannotrepresentatheoriginalasignals,liketheblurofimageedgesandapitchchangesaofmusic.

Toavoidit,weshouldchooseathesamplefrequencymorethantwiceofhighestfrequencyofsignalaccordingtoNyquistsampleatheory.Forpracticalasolutions,wealwaysincreasesamplefrequencyordecreasehighestsignalfrequency.

Givenascenesampledbyaccdwithminimumhorizontalsamplingfrequency10cm-1,whatisthemaximumhorizontalfrequencyintheimagethatcanbecorrectlyrepresented?

Fs=10cm-1

Fs=2*Fmax

SoFmax=5cm-1

Ifyousub-sampleanimage,whydoyouhavemoreproblemsrelatedtoaliasing?

Whenweadothis,thepixelsaareareduced.Asaresult,theresolutionaaredecresed.

Pastebelowaclearexampleofartifactsgeneratedbyaliasing.Forthistaskyoucanuseyourownchoiceofimage.Usetheboxbelowfortheimageandcomments.

4verticaland2horizontal

Exercise2（a）

Quantize:

Quantizationistheprocessofapproximatingthecontinuousvaluesintheimagedatawithafinitesetofdiscretevalues.Theinputofaquantizeristheoriginaldataandtheoutputisoneamongthefinitenumberoflevels.Thisprocessisanapproximation,andagoodquantizerisonewhichrepresentstheoriginalsignalwithminimumloss（quantizationerror）.Inthislab,youwillworkwithascalaruniformquantizerappliedtogrey-scaleimages.

Lena,quantizationfactor2

Baboon,quantizationfactor8

Implementafunctionthatuniformlyquantizesgreylevelimages.Thefunctionwillallowthereductionofthenumberofgreylevelvaluesbyagivenfactorn（apowerof2）.

Note.Tovisualizetheimage,youneedtore-mapitinthe8-bit-per-pixelrepresentation.Showtheresultsintheboxesbelow.

Peppers,quantizationfactor128

Peppers,quantizationfactor32

Usethegivenspacesforyouranswers.Donotextendthespacesorchangethefont-size.Page13

Isquantizationareversibleprocess?

Canyourecoverwhatyoudiscarded?

Brieflyexplain.

No,forthequantizationprocess,somecolourlevelarereducedandwecannotagetthembackbecauseweadonotknowwhichapixeldoweachangeintonearcolour.

WritetheresultsbacktoPGM/PPMfilesusingthefunctionyoucreated.Makesurethatyourwritingfunctionallocatesacorrectnumberofbitsperpixel.Whatisthesizeofthefilescomparedwiththeoriginal?

Giventheresults,whatisatypicalapplicationfieldforquantization?

Includeinyoursubmissiontheoutputfilesandcommentontheresults.

Thesizeoffilesbecomessmaller.Becausewereducesomecolourlevel,thebitsthatareusedforrepresentacolourarebecomesless.Thefileaarecompressed.IfaweuesfactorN,wegetanewfilehave1/Nsize.

Exercise2（b）

Histograms:

Thispartofthelabisdedicatedtoimageprocessingusinghistograms.Ahistogramisastatisticalrepresentationofthedatawithinanimage.Thehistogramcanberepresentedasaplotofthefrequencyofeachgreylevel.Thisrepresentationshowsthedistributionoftheimagedatavalues.Bymanipulatingahistogram,itispossibletoimprovethecontrastinanimageandtheoverallbrightnessortosegmentdifferentareasoftheimagebyapplyingoneormorethresholdstothehistogramitself.

Implementafunctiontooutputthehistogramvaluesofagivengreylevelimage.Displayintheboxestheresultinghistograms.

Lena

Baboon

Peppers

Ifyounormalizethevaluesofthehistogramsothattheysumto1,whatdoesthevalueofabinrepresent?

Representtheprobabilityofeachgreylevel

Exercise2（c）

Equalize:

Equalizationisoneofthepossibleimageprocessingalgorithmsimplementedusinghistograms.Histogramequalizationallowsuseenhancingthecontrastofimages.Histogramequalizationemploysamonotonic,non-linearmappingwhichre-assignstheintensityvaluesofpixelsintheinputimagesuchthattheoutputimagecontainsauniformdistributionofintensities（i.e.aflathistogram）.

Implementafunctionthatequalizesgrey-scaleimagesbasedontheirhistogram.Theinputisagivengreylevelimage;

theoutputisthederivedimage