Home | Lehre | Videos | Texte | Vorträge | Software | Person | Impressum, Datenschutzerklärung | Blog
Visual Computing in A Nutshell
Survey
-
Visual computing
-
Computer graphics aka image synthesis
-
Rendering [Wiedergabe in jeglichem Sinne: 3D, 2D, Fonts, ...]
-
Modeling
-
Texturing
-
Animation, simulation, motion capture
-
Virtual Reality (i.e., games ;-) ), Augmented/Mixed Reality
-
Interaction, GUIs, devices
-
Data visualization
-
Image manipulation [Bildbearbeitung]
-
...
-
Image processing aka image analysis [Bildverarbeitung]
-
Segmentation, detection, ..., OCR, machine vision
-
Computer vision: "understanding" 3D scenes
-
Data compression [Datenreduktion und/oder -kompression]
-
Noise reduction, image restoration
-
Color management
-
Content-based image retrieval
-
Data hiding, watermarking
-
...
Historical Perspective
Some milestones in the history of computer graphics:
-
1963: Ivan Sutherland: Sketchpad
(drawing with a light pen)
-
1975: Ed Catmull: z-Buffer
(described one year earlier by Wolfgang Straßer in his PhD thesis)
-
1980: Turner Whitted: Ray
Tracing (described 1968 by Arthur Appel)
-
1986: Pixar: "Luxo,
Jr."
-
1987 to date: Pixar: RenderMan
and RenderMan Interface
Things to be Covered in this Course
A look at the syllabus:
-
Event-based programming (standard for windows-based user interfaces)
-
A bit of 2D computer graphics
-
Image manipulation
-
Vector graphics
-
Lots of real-time 3D programming
Organizational Stuff
-
Tools: .NET, C#, XNA
Game Studio Express
-
C# for Java programmers: 1,
2
-
Lab groups
-
Mailing list
The Human Visual System (HVS)
Images and further reading: Webvision
Optics
-
Cornea [Hornhaut]
-
Iris [Iris], cf. f-stop setting [Blendeneinstellung] of a camera
-
Pupil [Pupille], a hole, not an actual part
-
Lense [Linse]
-
Vitreous body [Glaskörper]
-
Retina [Netzhaut], cf. red-eye effect on flash photographs
-
Optic nerve [Sehnerv]
Receptors
-
The receptor cells (size: some mm) measure the
spatial distribution of light and---somewhat limited---its spectral distribution,
too
-
Each eye contains 100 million rod cells [Stäbchenzellen] for night
vision. They can detect single photons, but produce only a grayscale image.
Night vision does not constitute a mainstream topic of computer graphics
-
Each eye contains 6 million cone cells [Zapfenzellen] for daylight vision
-
Three different types for three primary colors [Grundfarben]: red (long
wavelength, L type), green (medium wavelength, M type), blue (short wavelength,
S type)
-
Blue-sensitive cones are much less abundant than the other two types (1/40
of red). Hence, it is difficult to make out details in images composed
of blue and black or yellow and white or red and violet or ...
-
Color-to-grayscale conversion: Green contributes most, blue least. Typical
weights are approximately 0.3*red + 0.6*green + 0.1*blue (Demo with Gimp).
-
Red-green color blindness: A genetic defect wipes out the difference between
red and green cone cells (Demo with Gimp).
Color Mixing, Color Spaces
-
Additive mixing [additive Farbmischung]: Start from black and switch on
lights: green + blue = cyan, blue + red = magenta, red + green = yellow
(somewhat surprising), equal parts of red + green + blue = grey or white.
(Demo with Gimp)
-
There can be a bluish green but no bluish yellow etc.
-
Color naming anomaly: There is no dark shade of yellow, only brown. (Demo
with Gimp)
-
In printing, we subtract color (hence subtractive mixing) from paper white.
RGB as primary colors for a printer won't allow mixing. We rather use cyan
(absorbs red = long wavelengths), magenta (absorbs green), and yellow (absorbs
blue). Y+M=R, Y+C=G, C+M=B (only theoretically; with existing printing
colors it's violet), equal parts of C+M+Y = gray or black (only theoretically;
in practice it's brownish and never becomes really dark). In practice,
black is almost always used in addition to CMY. (Demo with Gimp)
-
Physics views color as a distribution of power over the spectrum of visible
light (short/medium/long wavelengths), corresponding to MRYGCBM, the colors
of the rainbow. Humans, however, think of color in terms of the color wheel
[Farbkreis] with parameters such as Hue [Farbton], Saturation [Sättigung],
and Brightness [Helligkeit]. There are some simple methods (such as HSV
and HSL) to map these parameters to RGB and some sophisticated methods
(such as Lab, which officially is called L*a*b*) that are based on precise
psycho-visual experiments.
-
Most display devices [Anzeigegeräte] behave in a non-linear manner.
RGB = (255, 255, 255) will typically be about four times as bright
as RGB = (127, 127, 127). This relationship is represented through a power
law [Potenzgesetz] of the form output = inputg
with g ranging from 1.8 to 3.0. This g
is called the "gamma value" of the display.
-
The color of a glowing piece of material of a certain temperature can be
used as a reference for yellowish, reddish and bluish colors. In physics,
temperature is measured in Kelvins (symbol K), which is degrees Centigrade
[Grad Celsius] plus 273.15. Standard light bulbs have both an thermal temperature
and a color temperature
of about 2,000 to 3,000 K. Direct sunlight is around 6,500 K, with an overcast
sky having a higher color temperature. Note that a heated material
glows red, yellow and finally blue (!) with increasing temperature.
Built-In "Intelligence" and Visual Illusions
-
There are no receptor cells where the optic nerve connects to the retina.
This "blind spot" [blinder Fleck] is hidden from consciousness. (Experiment)
-
Only the Fovea [gelber Fleck], a small central region of the retina, contains
a large concentration of cone cells. Only here, daylight (i.e., color)
vision yields sharp images. (Experiment) To compensate for that, the eye
scans the focused object with small leaps (saccades). The completely sharp
image that we perceive is only an illusion.
-
Color balance aka white balance [Weißabgleich]: Color perception
adapts to the color of the light. For instance, we perceive a sheet of
paper as white even when it is lit by a candle. A photograph taken without
white balancing reveals the yellow tint [Farbstich].
-
3D vision = perception of depth: to be covered later
-
Visual illusions: In uncommon situations, the built-in "intelligence" of
the visual system may fail. (Demonstrations) We are even blind
for seemingly obvious changes in an image. (Demonstrations)