CMSC 635Basic ideaTexture, not ImagesUnsuccessful ApproachesMarkov Random FieldMarkov TextMarkov Texture SynthesisMultiresolution SynthesisSome SamplesSlide 10Slide 11Image AnalogiesMulti-scale Markov MethodEnhancementsBlurringEmbossingSuper-resolutionSlide 18Texture TransferSlide 20Line ArtPaintingSlide 23Slide 24Slide 25Texture-by-numbersSlide 27CMSC 635CMSC 635Texture SynthesisTexture SynthesisBasic ideaBasic ideaQuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.Have a patchof textureQuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.Want moreImage: Wei & Levoy, Fast Texture Synthesis Using Tree-Structured Quantization, SIGGRAPH 2000Texture, not ImagesTexture, not ImagesImage: Wei & Levoy, Fast Texture Synthesis Using Tree-Structured Quantization, SIGGRAPH 2000StationaryLocalStationaryLocalQuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.Unsuccessful ApproachesUnsuccessful ApproachesImage: De Bonet, Multiresolution Sampling Procedure for Analysis and Synthesis of Texture Images, SIGGRAPH 97QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.RepeatQuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.RandomPixelsQuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.SpectralSynthesisMarkov Random FieldMarkov Random FieldNew texture should look like it was generated by the same processDo not know the processCan model as a stochastic processNew texture should look like it was generated by the same processDo not know the processCan model as a stochastic process• Order 2: Order 2: Ther I the heingoind of-pleat, blur it dwere Ther I the heingoind of-pleat, blur it dwere wing waske hat trooss. Yout lar on wassing, an sit." wing waske hat trooss. Yout lar on wassing, an sit." "Yould," "I that vide was nots ther."Yould," "I that vide was nots ther.• Order 3: Order 3: I has them the saw the secorrow. And I has them the saw the secorrow. And wintails on my my ent, thinks, fore voyager lanated wintails on my my ent, thinks, fore voyager lanated the been elsed helder was of him a very free the been elsed helder was of him a very free bottlemarkable,bottlemarkable,• Order 4: Order 4: His heard." "Exactly he very glad trouble, His heard." "Exactly he very glad trouble, and by Hopkins! That it on of the who difficentralia. and by Hopkins! That it on of the who difficentralia. He rushed likely?" "Blood night that.He rushed likely?" "Blood night that.Markov TextMarkov TextModel as probabilistic state machine:State: previous n lettersTransition: select next letterModel as random processProbability based on frequencyModel as probabilistic state machine:State: previous n lettersTransition: select next letterModel as random processProbability based on frequency• Order 1: Order 1: t I amy, vin. id wht omanly heay atuss n t I amy, vin. id wht omanly heay atuss n macon aresethe hired boutwhe t, tl, ad torurest t plur I macon aresethe hired boutwhe t, tl, ad torurest t plur I wit hengamind tarer-plarody thishand. wit hengamind tarer-plarody thishand. Examples: Bentley, Programming Pearls, Addison Wesley, 2000Markov Texture SynthesisMarkov Texture SynthesisState: 2D neighborhood filled so farProbability based on similarity of blocks in source textureState: 2D neighborhood filled so farProbability based on similarity of blocks in source textureImage: Wei & Levoy, Fast Texture Synthesis Using Tree-Structured Quantization, SIGGRAPH 2000QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.5x5 NeighborhoodQuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.7x7 NeighborhoodQuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.9x9 NeighborhoodMultiresolution SynthesisMultiresolution SynthesisBlur and sample sourceImage pyramidBlur and sample sourceImage pyramidState at level n includes n & n+1State at level n includes n & n+1Some SamplesSome SamplesImage: Wei & Levoy, Fast Texture Synthesis Using Tree-Structured Quantization, SIGGRAPH 2000QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.Some SamplesSome SamplesQuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.Image: Wei & Levoy, Fast Texture Synthesis Using Tree-Structured Quantization, SIGGRAPH 2000Some SamplesSome SamplesQuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.Image: Wei & Levoy, Fast Texture Synthesis Using Tree-Structured Quantization, SIGGRAPH 2000Image AnalogiesImage AnalogiesAnalogyA : A’ :: B : B’Synthesize B’ from sources A, A’ & BAnalogyA : A’ :: B : B’Synthesize B’ from sources A, A’ & BQuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.:QuickTime™ and aTIFF (LZW) decompressorare needed