Transcript Digital Media - University System of Georgia

Digital Media
Dr. Jim Rowan
ITEC 2110
Color
COLOR
• Is a mess
• It’s a subjective sensation PRODUCED in the
brain
• Color differs for light and paint/ink
• Printing is different than viewing a monitor
– a monitor EMITS light of a specific wavelength (or a
combination of them)
– print is like paint... it absorbs all the colors EXCEPT
the color that you see which is reflected by the paint
– a ball that is painted yellow and is viewed in a room
that is lit by a light that is completely blue will look
black... Why?
Light
• Is electromagnetic radiation (just like
microwaves and radio signals... just
different wavelengths)
• Visible light has a wavelength that is
between 400 and 700 nanoMeters
• A nanoMeter is 1 billionth of a meter...
– HINT: It’s a very short wave
http://en.wikipedia.org
/wiki/Electromagnetic_
spectrum
Light
• Visible light is a mix of different
wavelengths of light at different
intensities
• Light is composed of different intensities
of different frequencies
– Spectral Power Distribution
Color
• We need to reproduce it electronically
and manipulate it digitally
• So… we need a way to model color.
• i.e. we need a way to convert a
subjective sensation to a reproducible
physical phenomenon
But first… Bird eye trivia
• Eagles have 600,000 cones per square
mm of retina, humans have 150,000
• The chestral (a bird) has cones that can
see UV light (we don’t)
• Owls “see” with their ears
– they have flaps that are offset in front of
the ear openings to detect vertical
positioning
One model of color:
roughly based on the eye
• Rods(night vision, B&W)
• Cones (3 kinds, one for red, one for
blue and one for green)
==> RGB
tri-stimulus theory: the theory that states any
color can be completely specified with just
3 values
RGB
• Color is captured as 3 numbers
– one for red
– one for green
– one for blue
• Color is displayed on a monitor by generating
3 different colors
– one for red
– one for green
– one for blue
The 3 colored things
• Phosphor for a CRT and some Flat panel
displays
• Pockets of fluorescent gas for Plasma panel
• …plus a bunch of other varieties...
• All of them have the ability to adjust the
intensity of each of the three colored things
resulting in the display of most (Not All!) of
the visible colors
• Why not all colors ==>
http://en.wikipedia.org/wiki/Computer_display
RGB...
Good but not all visible colors
• In truth, the 3 different cones in the eye
are cross connected in very complex
ways
– The firing of one receptor can inhibit or
accentuate the firing of another
• The model we use assumes (wrongly)
that each receptor is strictly sensing R
or G or B
• ==> RGB cannot completely reproduce
the visual stimulus
Gamut: The complete range or scope of
something
RGB
• Pure red
– (100%, 0%, 0%)
– (255,0,0)
• Pure green
– (0%, 100%, ,0%)
– (0,255,0)
• Pure blue
– (0%, 0%, 100%)
– (0,0,255)
• Gray? R = G = B
– (25,25,25)
– (150,150,150)
Mixing the Color of Light
• …is an additive process
– monitors emit light
• …is not like mixing paint
– mixing paint is a subtractive process
– paint absorbs light
How many colors?
• Different cultures have different ideas about
when 2 colors differ
• People individually differ in their ability to
distinguish between two colors
• With the range of 0-255
– which can be encoded in 1 byte (8 bits)
• The combinations of Red, Green and Blue
results in 16.8 million possibilities
16**24 = 16,777,216
Color Depth
• Usually expressed in bits
• One byte for each of the RGB => 24 bits
• Back to binary...
–
–
–
–
1 bit => 21 => 2 choices
2 bits => 22 => 4 choices
4 bits => 24 => 16 choices
8 bits => 28 => 256 choices
Black, White and Grey
• RGB (0, 0, 0) is black
– Macs and PCs are opposite… go figure!
• RGB (255, 255, 255) is white
• When R = G = B you get grey
– RGB (25, 25, 25) is dark grey
– RGB (200, 200, 200) is light grey
Color at 16 bit Color Depth
16 bits total to represent a color
• RGB with 24 bit color depth
– 24 bits => 3 bytes
– 3 bytes, 3 colors => one byte per color
• RGB with 16 bit color depth
–
–
–
–
–
16 bits => 2 bytes
2 bytes, 3 colors...
16/3 = 5 bits with one left over...
HMMMmmmm...
What to do?
…16 bit color depth
• 16/3 = 5 bits with one bit left over...
• What to do with the extra bit?
• Go back to human perception
– Humans do not discriminate Blue as well as they
do Green
– Evolutionary roots?
• Our environment is green
• Lots of green to discriminate
• Assign 5 bits to R & B, and 6 bits to G
– allows twice (how?) as many greens as blues
Why would you want more
than 16.8 million?
• 24 bits color depth is plenty for human
vision...
• 30 and 48 bit color are WAY more than
needed for human vision...
• If you scan at 48 bit color there is a lot of
information buried in the image than we can
see BUT...
• This information can be used by the program
to make extremely fine distinctions during
image manipulation (edge finding for
example)
– (Failed rocket engine example)
Why worry about color depth?
• One reason: file size
• A 100X100 RGB image
– at 24 bit color => 30,000 bytes uncompressed
– at 16 bit color => 20,000 bytes uncompressed
– 1 byte => 1/3 reduction of size
Why is big bad?
• All that data needs to be shuttled around
• It wastes valuable computer resources
– hard drive disk space
– vram space
– data transit time
internet
monitor
main
memory
VRAM
hard drive
FYI
• Indexed color…
– Same as a color table representation
– Same as a dictionary table representation
• The color table is aka the color palate
Indexed (indirect) color
vs.
8 bit (direct) color
• 8 bit (direct) color defines only 256 colors
– red through blue
– 256 choices, whether they are used or not
• Indexed color allows 256 different colors
– colors that actually exist in the image
Indexed color vs.
8 bit direct color
• But images in nature have a narrower range of
colors... a palate
– 8 bit direct color only allows 256 choices
• With indirect color you can store 256 different
colors that are actually found in the image
– results in an image that more closely mimics the image
Indexed (indirect) color
with 256 colors in palate
• Even though it allows for a closer-toreal-life image natural images will have
more than 256 different colors…
• What to do about this?
– use the nearest color
– optical mixing... dithering
Nearest color
• Loss of some detail
• Distorted color
• Generates artifacts
– Banding or posterization
– Example ===>
Optical mixing: Dithering
• Dithering uses a group of colors to
approximate the desired color
• Works well for high resolution images
– (why?)
• Works poorly for low resolution images
– (why?)
Questions?