1 / 40

Blue Noise Sampling via Delaunay Triangulation

Zoltan Szego †* , Yoshihiro Kanamori ‡ , Tomoyuki Nishita † † The University of Tokyo, *Google Japan Inc . , ‡ University of Tsukuba. Blue Noise Sampling via Delaunay Triangulation. Contents. Background Related Work Our Method Results Conclusions and Future Work. Contents.

abie
Download Presentation

Blue Noise Sampling via Delaunay Triangulation

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. ZoltanSzego†*, Yoshihiro Kanamori‡, TomoyukiNishita† †The University of Tokyo, *Google Japan Inc., ‡University of Tsukuba Blue Noise Sampling via Delaunay Triangulation

  2. Contents • Background • Related Work • Our Method • Results • Conclusions and Future Work

  3. Contents • Background • Related Work • Our Method • Results • Conclusions and Future Work

  4. Background • Sampling is essential in CG • rendering, image processing, object placement etc. Light sampling on HDR environment maps Halftoning

  5. Background • Desired sampling patterns • Equally distant samples … e.g. Poisson disk • Low energy in low frequency of the Fourier spectrum … Blue noise Equally distant → Blue noise cf. Totally random → White noise

  6. Background • Blue noise property • Observed in natural objects • Considered optimal for human eyes Layout of human eye photoreceptors [Yellott, 1983]

  7. Background • Quality measures for blue noise spectra • Radial average power spectrum • The larger the central ring, the better • Anisotropy • The lower and flatter, the better ring Radial averagepower spectrum ring Spectrum Anisotropy

  8. Our Goal • Efficient, high-quality blue noise sampling • Adaptive sampling should be supported Uniform Adaptive

  9. Our Goal • Support for sampling in various domains • 2D • 3D (volumetric sampling) • On curved surfaces (spheres, polygonal meshes) 2D 3D On curved surfaces

  10. Contents • Background • Related Work • Our Method • Results • Conclusions and Future Work

  11. Related Work • Two major approaches • Dart throwing • Random sampling of equidistant samples • Tiling • Tiling of precomputed samples

  12. Related Work • Dart throwing[Cook, 1986] • Used for distributed ray tracing • High computational cost • Quality improvement: Lloyd’s relaxation … more costly • Parallel Poisson disk [Wei, 2008] • GPU-based acceleration • # of samples cannot be determined • Only supports 2D and 3D • Our method • # of samples can be specified • Supports 2D, 3D, and curved surfaces

  13. Related Work • Wang tiles [Kopf et al., 2006] • Requires precomputation • Low quality • Polyominoes [Ostromoukhov, 2007] • Requires complicated precomputation • Our method • High quality • No precomputation

  14. Contents • Background • Related Work • Our Method • Results • Conclusions and Future Work

  15. Overview • Input: seed points • Given by the user • Output: blue noise samples • Features: • Deterministic (reproducible with the same seeds) • No precomputation • Supports various sampling domains

  16. Overview • Sequentially sample atthe most sparse region • The largest emptycircle problem[Okabe et al., 2000] • Can be solved using Delaunay triangulation • Correspond to finding the largest circumcircle in Delaunay triangles 2D example

  17. Basic Algorithm • Loop: • Find the largest empty circle • Add a sample at the center 2D example

  18. Basic Algorithm • Loop: • Find the largest empty circle • Add a sample at the center • UpdateDelaunay triangles 2D example

  19. Basic Algorithm • Acceleration for search: Use of heap • To find the largest circumcirclein O(1) • Costs for insert / delete: O(log N) • Support for adaptive sampling • Scale the radii stored in the heapusing density functions • The greater the density, the higher the priority Heap of circumcircles’ radii Density function

  20. Artifact #1 • Regular patterns peaks in the spectrum

  21. Modification #1 • Reason of the artifacts • Iterative subdivisions of equilateral triangles • Our solution: • Detect an equilateral triangle • Displace the new samplefrom the center of its circumcircle(see our paper for details)

  22. Artifact #2 • Sparse samplesat boundaries • Reason • Very thin trianglesaround boundaries • Our solution: • Use of periodic boundaries Tiled samples(tiled just for illustration)

  23. Modification #2 • Periodic boundaries • Toroidal (torus-like) domain

  24. Modification #2 • Pros: • Sparse regions disappear • Edge lengths of triangles become balanced • Overall centers of circumcircles lie within their triangles • Allows us to specify the position of the new sample in O(1) • Cons: • A little additional cost for modifying coordinates

  25. Parallelization • Exploit multi-core CPUs • Uniform subdivision of 2D domain • Further subdivision • Costs: O(N log N)4 M log M < N log N (if M = N/4) • 4x4 subdivision is the fastest for a 4-core CPU • 1.69 times faster for 100K samples 1 2 3 4 1 1 1 1 2 2 2 2 3 3 3 3 4 4 4 4

  26. Sampling in 3D • 3D domain: [0, 1)3 • 2D → 3D • Triangles → Tetrahedra(Delaunay Tetrahedralization) • Circumcircles→ Circumspheres • Similar to 2D algorithm Delaunay tetrahedralization

  27. Sampling on Curved Surfaces • Sampling domain: • Spherical surfaces • Polygonal mesh surfaces • Initial seeds: • Vertices of simplified mesh • Similar to 2D • New samples are projectedonto the surface Simplified Given mesh Initial seeds Samples on a sphere

  28. Contents • Background • Related Work • Our Method • Results • Conclusions and Future Work

  29. Results • Uniform sampling # of samples: 20KTime: 92 ms Experimental environment:Intel Core 2 Quad Q6700 2.66GHz,2GB RAM

  30. Comparison – 50,000 samples – Radial average Anisotropy Radial average Anisotropy Our method: 378 msec Wang tiles [2006]: 1.35 msec

  31. Comparison – 50,000 samples – Radial average Anisotropy Radial average Anisotropy ours Our method: 378 msec Dart throwing [2007]: 420 msec

  32. Results • 20K samples in 3D

  33. Results • Spectra for 10K samples in 3D Low energy spheres in the center → blue noise property

  34. Results • Sampling on a sphere • Initial mesh: an equilateral octahedron Dense Sparse Density function

  35. Results • Sampling on HDR environment maps • Blighter region → denser samples

  36. Contents • Background • Related Work • Our Method • Results • Conclusions and Future Work

  37. Conclusions • High-quality blue noise samplingusing Delaunay triangulation • Find centers of largest circumcirclesof Delaunay triangles • Adaptive sampling by scaling circumcircles’ radii • Support for sampling on various domains:2D, 3D, and curved surfaces

  38. Future Work • GPU acceleration using CUDA • Fast Lloyd’s relaxation using the connectivity of Delaunay triangles

  39. Thank you

More Related