Computer Graphics - Outlook - Hendrik Lensch Computer Graphics - - PowerPoint PPT Presentation

▶
computer graphics
SMART_READER_LITE
LIVE PREVIEW

Computer Graphics - Outlook - Hendrik Lensch Computer Graphics - - PowerPoint PPT Presentation

Oct 17, 2022 293 likes •729 views

Computer Graphics - Outlook - Hendrik Lensch Computer Graphics WS07/08 Outlook Overview Last Lectures Current trends Current work in AG4 MPI Current work at UdS Exam Monday, 18 th please be there at 8:00

slide-1
SLIDE 1

Computer Graphics WS07/08 – Outlook

Computer Graphics

  • Outlook -

Hendrik Lensch

slide-2
SLIDE 2

Computer Graphics WS07/08 – Outlook 2

Overview

  • Last Lectures

– Current trends – Current work in AG4 – MPI – Current work at UdS

  • Exam

– Monday, 18th

  • please be there at 8:00 sharp
  • starts at 8:15 will end at 10:00.

– bring a ruler – no other devices!

slide-3
SLIDE 3

Computer Graphics WS07/08 – Outlook

Highlight you should not have missed!

a non-exclusive list of relevant topics of this lecture

slide-4
SLIDE 4

Computer Graphics WS07/08 – Outlook

Topics (1)

  • ray tracing vs rasterization
  • recursive ray tracing
  • ray surface intersections
  • spatial acceleration structures (dynamics)
  • shading, reflection, refraction, BRDF, …
  • radiometry
  • rendering equation
  • texture mapping (mip-maps, … )
  • sampling theory
  • antialiasing
  • HDR, contrast, tonemapping
  • transformations!
  • rasterization (Bresenham, polygons)
slide-5
SLIDE 5

Computer Graphics WS07/08 – Outlook

Topics (2)

  • OpenGL, Cg (basics)
  • plenoptic function, light fields, panoramas
  • splines (evaluation)
  • volume rendering
slide-6
SLIDE 6

Computer Graphics WS07/08 – Outlook

Lots of topics not touched in this lecture

(have a look at the SIGGRAPH proceedings)

slide-7
SLIDE 7

Computer Graphics WS07/08 – Outlook

Level of Detail in Shading

  • Frequency Domain Normal Map Filtering

[Han et al. – SIGGRAPH 2007]

  • rendering of surface detail depends on viewing distance
slide-8
SLIDE 8

Computer Graphics WS07/08 – Outlook

Level of Detail in Shading

  • correct mip-mapping of bump maps
  • multiple normal will integrate to some super-pixel
  • consider the normal distribution function
  • approximate NDF with a set of spherical functions
slide-9
SLIDE 9

Computer Graphics WS07/08 – Outlook

Level of Detail in Shading

slide-10
SLIDE 10

Computer Graphics WS07/08 – Outlook

Normal Mapping

slide-11
SLIDE 11

Computer Graphics WS07/08 – Outlook

Non-Photo Realistic Rendering

  • Line Drawing Via Abstracted Shading

[Markosian – SIGGRAPH 2007]

  • NPR creates a more abstract image
  • removal of unnecessary detail
slide-12
SLIDE 12

Computer Graphics WS07/08 – Outlook

Ridge Detection in Tone Image

  • compute a tone image
  • detect lines in the tone image -> abstraction

movie

slide-13
SLIDE 13

Computer Graphics WS07/08 – Outlook

Interactive Global Illumination

  • Implicit Visibility and Antiradiance for Interactive Global

Illumination [Dachsbacher et al. – SIGGRAPH 2007]

slide-14
SLIDE 14

Computer Graphics WS07/08 – Outlook

Antiradiance

  • do not use any shadow rays
  • simply transport negative light
  • solve for global illumination through iteration
slide-15
SLIDE 15

Computer Graphics WS07/08 – Outlook

Antiradiance

slide-16
SLIDE 16

Computer Graphics WS07/08 – Outlook

Image-Resizing

  • “Seam Carving for Content-Aware Image Resizing”

[Avidan and Shamir, SIGGRAPH 07]

slide-17
SLIDE 17

Computer Graphics WS07/08 – Outlook

Image-Resizing

  • “Seam Carving for Content-Aware Image Resizing”

[Avidan and Shamir, SIGGRAPH 07]

slide-18
SLIDE 18

Computer Graphics WS07/08 – Outlook

Energy Minimization

magnitude of gradient accumulated path cost: horizontal, vertical

slide-19
SLIDE 19

Computer Graphics WS07/08 – Outlook

Algorithm

  • Find minimum path from top to bottom

(left to right)

  • Dijkstra’s Algorithm O(n log n )
  • For shrinking:

– iteratively remove individual seams – increases the energy in every step

  • For growing:

– iteratively insert individual seams – decreases the energy in every step – make sure not to insert at the same place over and over again, use the seams suggested for shrinking

movie

slide-20
SLIDE 20

Computer Graphics WS07/08 – Outlook

Large Images

  • Capturing and Viewing Gigapixel Images

[Kopf et al. – SIGGRAPH 2007] movie

slide-21
SLIDE 21

Computer Graphics WS07/08 – Outlook

Physics Simulation

  • Efficient Simulation of Inextensible Cloth

[Goldenthal et al. – SIGGRAPH 2007]

  • this motion is governed by an augmented Lagrange

equation with velocity v(t), mass matrix M, bending, shear and gravity V(x), extensibility constraints C

  • efficient solution by simplification

movie

slide-22
SLIDE 22

Computer Graphics WS07/08 – Outlook

Fluid Flow

  • Multiple Interacting Liquids

[Losasso et al. – SIGGRAPH 2006]

slide-23
SLIDE 23

Computer Graphics WS07/08 – Outlook

Fluid Flow

  • Multiple Interacting Liquids

[Losasso et al. – SIGGRAPH 2006]

slide-24
SLIDE 24

Computer Graphics WS07/08 – Outlook

Crowd Simulation

  • Continuum Crowds [Treuiller et al. – SIGGRAPH 2006]
  • simulated using continuous dynamics
  • implicit collision avoidance
slide-25
SLIDE 25

Computer Graphics WS07/08 – Outlook

Crowd Simulation

slide-26
SLIDE 26

Computer Graphics WS07/08 – Outlook

Collision Detection

  • BD-Tree: Output-Sensitive Collision Detection for Reduced

Deformable Models [James and Pai – SIGGRAPH 2004]

  • collision handling necessary for correct physics simulation
  • spatial acceleration structure to perform collision detection

even of deformable objects

slide-27
SLIDE 27

Computer Graphics WS07/08 – Outlook

Collision Detection

slide-28
SLIDE 28

Computer Graphics WS07/08 – Outlook

Animation of Flexible Bodies

  • FastLSM: Fast Lattice Shape Matching for Robust Real-

Time Deformation [Rivers&James – SIGGRAPH 2007]

slide-29
SLIDE 29

Computer Graphics WS07/08 – Outlook

Current Topics at MPI Department 4 Computer Graphics

slide-30
SLIDE 30

Computer Graphics WS07/08 – Outlook

Motion Retrieval

  • given a small motion sequence determine segments

that perform a similar motion

Meinhard Mueller www.mpi-inf.mpg.de/~mmueller

slide-31
SLIDE 31

Computer Graphics WS07/08 – Outlook

Multi-Touch-Display

  • 3D modeling application

tracking: 100Hz using Cuda

slide-32
SLIDE 32

Computer Graphics WS07/08 – Outlook

  • Image gradients (color difference between neighboring

pixels)‏

  • Heavy-tail distribution function on a global scale

Global Image Statistics

Boris Ajdin www.mpi-inf.mpg.de/~bajdin

slide-33
SLIDE 33

Computer Graphics WS07/08 – Outlook

Local Image Statistics

  • Global statistics do not allow subtle local changes

within the images

  • We have developed a novel
  • bjective function which

should correspond to naturally looking images

  • It is defined on a small pixel

neighborhood, allowing for easy pixel reconstruction.

  • Applied to image demosaicing.

Boris Ajdin www.mpi-inf.mpg.de/~bajdin

slide-34
SLIDE 34

Computer Graphics WS07/08 – Outlook

Proposed work

  • Acquiring a large high quality full RGB image datasets

(internet, our cameras).

  • Further testing the objective function – possible

adjustments.

  • Code optimization (implementing CUDA)
  • Applications to image denoising, upsampling, ...

Boris Ajdin www.mpi-inf.mpg.de/~bajdin

slide-35
SLIDE 35

Computer Graphics WS07/08 – Outlook

Scientific Computing with GPUs

Accelerator Node with Accelerator Node with PCI (Express) Connection PCI (Express) Connection Heat distribution on Heat distribution on a chip under load a chip under load Robert Strzodka www.mpi-inf.mpg.de/~strzodka

slide-36
SLIDE 36

Computer Graphics WS07/08 – Outlook

GPU Results: Conjugate Gradient (CG) and Multigrid (MG)

Smaller is better

5e-7 5e-6 5e-5 5e-4 6 7 8 9 10 Seconds per grid node Data level Performance of double precision CPU and mixed precision CPU-GPU solvers CG CPU CG GPU MG2+2 CPU MG2+2 GPU

Robert Strzodka www.mpi-inf.mpg.de/~strzodka

slide-37
SLIDE 37

Computer Graphics WS07/08 – Outlook

Distance Transforms and Skeletons

  • riginal

boundary trimmed skeleton fine skeleton distance transform Robert Strzodka www.mpi-inf.mpg.de/~strzodka

slide-38
SLIDE 38

Computer Graphics WS07/08 – Outlook

paint features compute DT simplify simplify

Feature-Preserving Simplification

Robert Strzodka www.mpi-inf.mpg.de/~strzodka

slide-39
SLIDE 39

Computer Graphics WS07/08 – Outlook

Distance Transforms and Voronoi Diagrams

Generalized weighted Voronoi diagram Robert Strzodka www.mpi-inf.mpg.de/~strzodka

slide-40
SLIDE 40

Computer Graphics WS07/08 – Outlook

Motion Estimation – Feature Extraction

Fig1a.mov Fig1d.mov Fig1e.mov Fig1g.mov

[Strzodka and Garbe, Visualization 2004]

Robert Strzodka www.mpi-inf.mpg.de/~strzodka

slide-41
SLIDE 41

Computer Graphics WS07/08 – Outlook

Motion Estimation – Blood Visualization

Fig8ppre.mov

Fig8abc.mov

Robert Strzodka www.mpi-inf.mpg.de/~strzodka

slide-42
SLIDE 42

Computer Graphics WS07/08 – Outlook

3D-Reconstruction from Video

Thorsten Thormaehlen www.mpi-inf.mpg.de/~tormaehlen

slide-43
SLIDE 43

Computer Graphics WS07/08 – Outlook

Differential Photon Mapping

  • Special version of photon

mapping for augmenting photographs

– Simulates the difference in illumination introduced by virtual

  • bjects
  • Bachelor Thesis

– Include Final Gathering in the existing photon mapper – Thorsten Grosch – (tgrosch@mpi-inf.mpg.de) Thorsten Grosch www.mpi-inf.mpg.de/~tgrosch