Demonstrating Anisotropic Surface Reflectance with Example-Based Microfacet Synthesis Jiaping Wang 1 , Shuang Zhao 2 , Xin Tong 1 John Snyder 3 , Baining Guo 1 Microsoft Research Asia 1 Shanghai Jiao Tong University 2 Microsoft Research 3Slide 2
Surface Reflectance glossy silk metal woodSlide 3
Anisotropic Surface Reflectance isotropic anisotropicSlide 4
Our Goal displaying spatially-differing anisotropic reflectan ceSlide 5
Surface Reflectance in CG 4D BRDF ρ ( o , i ) B idirectional R eflectance D istribution F unction what amount of light reflected wrt in/out bearings o iSlide 6
Surface Reflectance in CG 4D BRDF ρ ( o , i ) B idirectional R eflectance D istribution F unction what amount of light reflected wrt in/out headings 6D Spatially-Varying BRDF: SVBRDF ρ ( x , o , i ) BRDF at every surface point xSlide 7
Related Work I parametric BRDF models minimized representation simple procurement and fitting need reasonable subtle elements ground truth parametric model [Ward 92]Slide 8
Related Work II arranged SVBRDF sensible substantial information set hard to catch long process costly equipment picture enlistment light vault [Gu et al 06]Slide 9
Related Work II organized SVBRDF practical expansive information set hard to catch protracted procedure costly equipment picture enrollment light arch [Gu et al 06]Slide 10
Microfacet BRDF Model surface displayed by little reflect features [Cook & Torrance 82]Slide 11
Microfacet BRDF Model surface demonstrated by modest mirror aspects [Cook & Torrance 82] fresnel term typical dissemination shadow termSlide 12
Microfacet BRDF Model in light of Normal Distribution Function (NDF) NDF D is 2D capacity of the midway vector h rules surface appearanceSlide 13
Challenge: Partial Domains tests from a solitary review course i cover just a sub-district h Ω of NDF How to acquire the full NDF? ? fractional NDF complete NDF incomplete districtSlide 14
Solution: Exploit Spatial Redundancy discover surface focuses with comparative yet contrastingly turned NDFs material specimen halfway NDF at every surface pointSlide 15
Example-Based Microfacet Synthesis incomplete NDFs from other surface focuses Align + = fractional NDF to finish pivoted incomplete NDFs finished NDFSlide 16
Comparison ground truth our model isotropic Ward model anisotropic Ward modelSlide 17
Overall Pipeline BRDF Slice Capture Partial NDF Recovery Microfacet SynthesisSlide 18
Overall Pipeline BRDF Slice Capture Partial NDF Recovery Microfacet SynthesisSlide 19
Device Setup Camera-LED framework, in light of [Gardner et al 03]Slide 20
Capturing ProcessSlide 21
Overall Pipeline BRDF Slice Capture Partial NDF Recovery Microfacet SynthesisSlide 22
NDF Recovery modify the microfacet BRDF model Measured BRDF Unknown NDF Shadow Term , [Ashikhmin et al 00]Slide 23
NDF Recovery (con\'t) iterative methodology [ Ngan et al 05 ] settle for NDF, then shadow term works for complete 4D BRDF information [ Ngan et al 05 ] 1. , [Ashikhmin et al 00] 2.Slide 24
Partial NDF Recovery one-sided result on deficient BRDF information [Ngan et al. 05] ground truth NDF shadow term shadow term NDFSlide 25
Partial NDF Recovery (con\'t) minimize the predisposition isotropically compel shadow term in every emphasis after requirement before limitationSlide 26
Recovered Partial NDF [Ngan et al. 05] ground truth our outcomeSlide 27
Overall Pipeline Capture BRDF cut Partial NDF Recovery Microfacet SynthesisSlide 28
Microfacet Synthesis fractional NDF to finish finished NDF Merged incomplete NDFsSlide 29
Microfacet Synthesis (con\'t) clear usage: For N NDFs at every surface point Match against ( N - 1) NDFs at different focuses In M pivot plots for arrangement number of revolutions/examinations: N 2 *M ≈ 5 × 10 11 ( N ≈ 640k , M ≈ 1k )Slide 30
Synthesis Acceleration a direct execution: For N NDFs in every surface point Match with ( N - 1) NDFs in other area In M turn plots for arrangement times of circular capacity turn and correlation N 2 * M ≈ 5 × 10 11 ( N ≈ 640k ) Clustering [Matusik et al 03] complete delegate NDFs just (1% of full set) N\' 2 * M ≈ 5 × 10 7 ( N\' ≈ 6.4k )Slide 31
Synthesis Acceleration a direct execution: For N NDFs in every surface point Match with ( N - 1) NDFs in other area In M revolution plots for arrangement times of round capacity pivot and correlation N 2 * M ≈ 5 × 10 11 ( N ≈ 640k ) Clustering [Matusik et al 03] complete agent NDFs just (1% of full set) Search Pruning precompute every pivoted competitor prune by means of various leveled seeking N\' 2 * M ≈ 5 × 10 7 ( N\' ≈ 6.4k ) N\'* log( N\'* M ) ≈ 5 × 10 5Slide 32
Performance Summary 5-10 hours for BRDF cut procurement in HDR 1 Hour for obtaining in LDR 2-4 hours for picture preparing 2-3 hours for halfway NDF recuperation 2-4 hours for a ccelerated microfacet combination On a PC with Intel Core TM 2 Quad 2.13GHz CPU and 4GB memorySlide 33
Model Validation full SVBRDF dataset [Lawrence et al. 06] information from one perspective for demonstrating information from different perspectives for acceptanceSlide 34
Validation ResultSlide 35
Limitations visual displaying, not physical precision single-skip microfacet model retro-reflection not took care of spatial repetition of turned NDFs simple fix by pivoting the specimenSlide 36
Rendering Result: SatinSlide 37
Rendering Result: WoodSlide 38
Rendering Result: Brushed MetalSlide 39
Conclusions model surface reflectance by means of microfacet amalgamation general and smaller representation high determination (spatial & rakish), sensible result less demanding procurement: single-perspective catch shoddy gadget shorter catching timeSlide 40
Future Work execution streamlining catching and information handling expansion to non-level articles augmentation to various light ricochetSlide 41
Acknowledgments Le Ma for hardware of the LED exhibit Qiang Dai for catching gadget setup Steve Lin, Dong Xu for profitable talks Paul Debevec for HDR symbolism Anonymous analysts for their accommodating recommendations and remarksSlide 42
Picture separating operation that specially smoothes one piece of a picture ... This operation i ...
The workspace of an obliged robot has less that six degrees of opportunity. ... Double quaternio ...
Christopher Alexander has his own standards (see A Pattern Language and A New Theory of Urban De ...
Joint Prosody Prediction. what's more, Unit Selection for. Concatenative Speech Synthesis ... De ...
Glycerolipid Biosynthesis. Essential for the union of film lipids and triacylglycerolSynthesis h ...
Isotropic versus Anisotropic. Anisotropic minerals contrast from isotropic minerals because:the ...
Headache medicine: some foundation. Protected by Bayer in 1893One of the most seasoned drugsOne ...
Aqueous SYNTHESIS AND CRYSTALLIZATION OF ZEOLITES . Nax(AlO2)x(SiO2)y.zH2O(crystals)Extraframewo ...
Indicatrix. Imaginary figure, but very useful The figures show and/or define: Location of o ...
Surface Plasmon Resonance: antigen-antibody interactions. Vamsi K. Mudhivarthi. Analytical T ...
Eukaryotic Protein Synthesis. See Figure 33.22 for the structure of the typical mRNA transcri ...
27.4 Synthesis of Amino Acids. O. O. CH 3 CHCOH. CH 3 CHCO. Br. N H 3. From a -Halo Ca ...
2. Februrary 2006 . UK Cochrane Center. Destinations. Talk about the reason and attributes of me ...
Imines in putative prebiotic blend of histidine:. . . Strecker amalgamation. Interestingly, AA\' ...
The Purpose. Nanocrystals of the attractive metals and metal oxides are utilized as: - recording ...
Dr. Ronald FediwDepartment of Computer Science, Stanford UniversityConference continuing at ACM ...
About Optronic Laboratories, Inc.. In 1998, Optronic Laboratories, Inc. moved into its new 25,00 ...
Strong state NMR spectra. 2. Strong state NMR. Anisotropic InteractionsWhat are they, what do th ...