Photorealism






Perception






Color

Colors in the physical world can be any wavelength, or combination of wavelengths, of light


ColorWavelength
Violet 420 nm
Blue 470 nm
Green 530 nm
Yellow 580 nm
Orange 620 nm
Red 700 nm





The Eye

Rods & cones absorb light, send signal to brain






Color Perception

Any visible wavelength is perceived the same as some combination of 3 basic colors
(roughly blue, green, and red)






RGB Color

RGB = Red , Green , Blue
Each component (R, G, or B), ranges from a minimum (no intensity) to a maximum (full intensity), typically 0.0 to 1.0.






Color on the Computer

glColor3f(1.0, 0.5, 0.0)

Computer numbers have a finite resolution - how many distinct values can be represented

24 bit color = 8 bits red + 8 bits green + 8 bits blue
(a.k.a. 8 bits per component)
8 bits = 256 possible values

32 bit color usually means 8 bits red + 8 bits green + 8 bits blue + 8 bits alpha
16 bit color can be 5 bits red + 6 bits green + 5 bits blue


HDRI: High Dynamic Range Imaging - uses 16 or 32 bits per component






RGB Photography


Prokudin-Gorskii, ca. 1910





RGB Photography






CMY Color

CMY = Cyan , Magenta , Yellow
    C = 1.0 - R
    M = 1.0 - G
    Y = 1.0 - B

CMYK = Cyan , Magenta , Yellow , Black






HSV Color

HSV = Hue , Saturation , Value






Color Gamuts






Luminance

The "brightness" of a color.

Formula, used in NTSC television standard, based on human perception:

    0.30 * R + 0.59 * G + 0.11 * B






Real-world Luminance

Background Luminance
Moonless overcast night sky 0.00003 cd/m^2
Moonlit clear night sky 0.03
Twighlight sky 3
Overcast day sky 300
Day sky with sunlit clouds 30,000


Rods & cones adapt to average level of illumination

Rods most sensitive at low levels (scotopic vision)

Cones more sensitive at higher levels (photopic vision)






Vector vs Raster

Historically, displays have been divided between vector and raster

vector - pictures are drawn as a set of precise lines, connecting arbitrary points on the screen



raster - pictures are drawn by scanning the screen in a discrete sequence of rows




Vector vs Raster

A similar distinction continues - graphical objects can be described by images or geometry

    





Images

     

A digital image is a 2 dimensional array of pixel colors

Pixel = "picture element"

Each pixel is a sample of a continuous, analog image






Pixels



Pointillism can be considered to take a similar approach - breaking an image down into discrete samples






Image Data

Basic image data in computer memory is a stream of numbers


1 239 120 1 1 37 94 8 92 31 80 92 134 89 2 3 50 9 3 10 93 109 134  ...

Important additional information is:


Total memory needed for a simple image:

    width * height * components * bytes_per_component

e.g. a 512x512, RGB, 8-bit image requires 768 kilobytes






Frame Buffer

The frame buffer is a chunk of graphics card memory that contains what is displayed on the screen.

Like an image, but for each pixel there can be additional data besides color - depth, masking, etc.

OpenGL renders shapes, images, etc. into pixels of the frame buffer.
rasterizing it - converting it into raster form in the frame buffer.






Visual Acuity

Resolution of the eye - approximately 0.5 arc-minute, at the center of vision

fovea: 1 - 2 degree area at center of retina with maximum density of cones

1280x1024, 17" monitor viewed from 2 feet : 1 pixel = ~ 1.4 arc-minutes

Creative Commons License
This document is by Dave Pape, and is released under a Creative Commons License.