Introduction to Image Processing
Jacob Furst DePaul CTI
What is an Image? • A digita digitall repres represent entati ation on of a real-worl real-world d scene. scene. (Graph (Graphics ics is the digital representation of an imaginary scene.) • Co Comp mpose osed d of d disc iscret rete e elem element ents s gene general rally ly calle called d pictu picture re element (or pixels for short) • Pixe Pixels ls are are pa para rame mete teri rize zed d by – position – i nten si ty – time
• In all all comb combina inatio tions, ns, the these se para parame meter ters s defin define e still still imag images, es, video, volume data and moving volumes
Digital Photographs
•
•
Two spa pattial param rameters ers – x, or or hori horizo zont ntal al pos posit itio ion n – y, or vert vertic ical al po posi siti tion on Three ree int inten ens sity ity par param amet eter ers s – Red – Green – Bl ue
Ultrasound
•
Two spat spatiial pa para ram meters ers – x and y
•
One int nte ensi nsity pa parram ame ete terr – ultrasound reflection One One tim time e par param amet eter er (ult (ultra raso soun und d printouts don’t show this, but the exam does)
•
image http://whyfiles.org/co olimages/images/csi/ultrasou s/csi/ultrasound.jpg nd.jpg imagefrom fromhttp://whyfiles.org/coolimages/image http://whyfiles.org/coolimages/imag http://whyfiles.org/c oolimages/images/csi/ultrasou es/csi/ultrasound.jpg nd.jpg
Digital X -ray
•
Two spat spatiial pa para ram meters ers – x and y
•
A singl ingle e inte intens nsiity pa para ram met eter er – xray attenuation
image http://www.smm.org/heart/lessons/jpgs/fu rt/lessons/jpgs/full/xray.jpg ll/xray.jpg imagefrom fromhttp://www.smm.org/hea http://www.smm.org/heart/lessons/jpgs/f http://www.smm.org/he art/lessons/jpgs/full/xray.jpg ull/xray.jpg
Digital Video
•
•
•
Two spa pattial param rameters ers – x, or or hori horizo zont ntal al pos posit itio ion n – y, or vert vertic ical al po posi siti tion on Three ree int inten ens sity ity par param amet eter ers s – Red – Green – Bl ue One time parameter – frame #
Other Kinds of Images • Comp Comput uted ed Tomo Tomogr grap aphy hy – 3 dimens dimension ional al x-ra x-ray y imag images es of th the e human body • Sate Satellllit ite e ima image ges s – 4 int inten ensi siti ties es – red, red, gree green, n, blue blue an and d infrared • Func Functi tion onal al Ma Magn gnet etic ic Re Reso sona nanc nce e Ima Image ges s – 3 dim dimen ensio siona nall images of the human body over time • No Nott all all ima image ges s repre represe sent nt visua visuall phe phenom nomena ena bu butt a visualization can be a very effective way to understand the phenomena
Talking about a Pixel • • • •
Give Give th the e iima mage ge a n nam ame; e; usua usualllly y I or f Spec Specif ify y all all the the pa para rame mete ters rs (sp (spac ace, e, time time)) of your your imag image e The The rres esul ultt is is the the inte intens nsit ity y For exampl example, e, a digita digitall image image I might might have have inten intensit sity y (23, (23, 23, 23, 97) at pixel I(32,215) - inte inten nsit sity 23 in red - inte inten nsity sity 23 in green reen - inte inten nsit sity 97 97 in blu blue - at the 32 nd column - at the 215th row
• A CT imag image e f mig might ht hav have e inte intens nsity ity 25 255 5 at ff(6 (67, 7, 95, 95, 13) 13)
What is Image Processing • Image Image p proc rocess essing ing typica typically lly atte attemp mpts ts tto o acco accomp mplish lish one of three things – rest restor orin ing g image images s – en enha hanc ncin ing g image images s – un unde ders rsta tand ndin ing g images images
• Resto Restorat ration ion takes takes a corr corrup upte ted d imag image e and and atte attemp mpts ts to to recreate a clean original • Enhanc Enhanceme ement nt alte alters rs an an image image to make makes s its its meani meaning ng clea clearer rer to human observers • Und Unders erstan tandin ding g usual usually ly atte attemp mpts ts to mimic mimic the the huma human n visual visual system in extracting meaning from an image
Image Restoration • Ima Image ge rest restora oratio tion n is imp import ortant ant for two ma main in appl applica icatio tions ns – remo removi ving ng sens sensor or no nois ise e – restoring restoring old, archived archived film and images images
• Many Many senso sensors rs are are subje subject ct to to noise noise,, thus thus prod produci ucing ng corr corrupt upted ed images that don’t reflect the real world scene accurately • Old ph photo otogra graph ph and and film film archiv archives es ofte often n show show cons conside iderab rable le damage
Restoration Example
Images Imagesfrom fromDigital DigitalImage ImageProcessing, Processing,Gonzalez Gonzalezand andWoods Woods
Restoration Example
Images http://www.screenge nes.com/drest_3.html Imagesfrom fromhttp://www.screengenes.com/drest_3.html http://www.screengenes.com/drest_3.html http://www.screeng enes.com/drest_3.html
Image Enhancement
Images Image sscourtesy cour oof Images Image courtesy courtesy tesyof offTobey TobeyThorn Thorn
•
Often Often used used to increase increase the contrast contrast in image images s that that are are overly overly da dark rk or or light light
• •
Enhan Enhancem cement ent algo algorit rithms hms ofte often n play play to huma humans’ ns’ sens sensiti itivit vity y to contrast contrast More sophisti sophisticate cated d algori algorithms thms enhance enhance images images in in a small small neigh neighborh borhood ood,, allowing overall better enhancement.
Image Understanding • Ima Image ge und unders erstan tandin ding g iincl nclude udes s many many differ different ent tasks tasks – segm segmen enta tati tion on – clas classi sifi fica cati tion on – inte interp rpre reta tati tion on
• Segmen Segmentat tation ion involv involves es iden identif tifyin ying g obje objects cts in a an n image image • Classi Classific ficati ation on assign assigns s labe labels ls to indivi individua duall p pixe ixels ls • Int Interp erpret retati ation on extr extract acts s some some mean meaning ing from from the the imag image e as a whole • Lea Leads ds to to such such fiel fields ds as as image image ana analys lysis, is, comput computer er visi vision on and and visual computing
Creating Images • The The crea creati tion on of of imag images es inv invol olve ves s two two main main tas tasks ks – spatial spatial sampling, sampling, which determi determines nes the the resolutio resolution n of an image image – quantizatio quantization, n, which which determines determines how many many intensity intensity levels levels are are allowed
• Spati Spatial al samp samplin ling g deter determi mines nes wha whatt level level of det detail ail can can be be seen seen – finer finer sampl sampling ing allo allows ws for for smaller smaller deta detailil – finer sampling sampling require requires s more pixels pixels and “larger” “larger” images images
• Quanti Quantizat zation ion dete determi rmines nes how how “smo “smooth oth”” the the contr contrast ast change changes s in the image are – finer quanti quantizatio zation n will prevent prevent “false “false contourin contouring” g” (artificia (artificiall edges) – courser courser quantiz quantizatio ation n allows allows for for compres compressing sing images images
Effect of Spatial Sampling
Images PProcessing, Imagesfrom from Digital Digital Image Image Processing, Processing, rocessing, Gonzalez Gonzalezand andWoods Woods
Effect of Spatial Sampling
Images PProcessing, Imagesfrom from Digital Digital Image Image Processing, Processing, rocessing, Gonzalez Gonzalezand andWoods Woods
Effect of Quantization
Images Imagesfrom fromDigital DigitalImage ImageProcessing, Processing,Gonzalez Gonzalezand andWoods Woods
Effect of Quantization
Images Imagesfrom fromDigital DigitalImage ImageProcessing, Processing,Gonzalez Gonzalezand andWoods Woods
