Images

• Continuing with chapter 1
• He states that images created in a graphics system are synthetic
• We will do an overview of the "graphics model"
• The system is composed of Objects and Viewers
  • Humans:
    • You are the viewer
    • You are looking at real physical objects.
    • Light from the objects
    • Passes through your cornea and lens
    • And strikes your retina where an image is formed.
    • This is an upside down two dimensional image.
    • And it is discrete, estimated at 576 megapixels.
    • Your brain takes the two images, slightly offset and "constructs" a 3d image.
    • By the way, the number of colors you can see is estimated to be around 10 million.
    • Look at the Color vision table on Wikipedia.
  • For a camera
    • The model is very close to the same.
    • But the biological items are replace
    • And the image is formed on film or a sensor
      • CCD: Charge-coupled device, or a CMOS based system
      • Thank monitor/framebuffer in reverse
  • In a graphics system
    • Objects are specified as
      • Primitives:
        • What to draw
        • Usually a vertex or a set of vertexes
        • A "type" for the primitive, or how to connect the vertexes
      • Attributes:
        • How to draw it
        • Colors, widths, ...
      • The viewer is somewhat akin to a photographic plate or CCD
        • It has a position in space
        • It has a resolution
        • It has a range of colors supported.
      • The trick is to fill in the right pixels for the primitives defined.
        • We will spend quite a bit of time discussing how to do this.
    • In the end, almost everything we deal with will be triangles specified as vertices.
  • Light
    • We tend to simplify light in computer graphics systems.
    • Think of it as an RGB value.
    • Think of it as a point source
      • A single infinately small element
      • Emitting "light" in all directions.
      • We control the color, intensity and other parameters
    • Light is tough and an entire chapter later on.
• Image Models
  • Light strikes an object
    • Based on the (physical properities/attributes) of the object, light is
      • Absorbed (stealth object, story over)
      • Reflected (bounces off where the angle of incidence and algle of reflection are equal)
      • Transmitted (acording to Snell's Law
      • Or all three.
  • Ray Casting
    • There is a light source emitting light rays.
    • Follow all of the rays to see what light strikes the viewer
    • Computationally impossible.
    • But Radiosity is an attempt.
      • Objects are divided into a number of patches.
      • Connections between patches, known as view factors, are described mathematically.
      • Every patch interacts with every other patch.
      • These are used to compute the light that is transfered from one patch to anohter.
      • The process is applied iteratively.
        • Initially only the light sources have any "light"
        • Later others get "reflected" illumination
      • 
      • Great for diffues light, soft glow.
      • Good for any position in the "area"
  • Ray Tracing
    • From each pixel in the viewer
    • Trace a ray out along the viewing direction.
    • Each ray will either
      • Strike a light source
        • This will give us a color for the pixel (or at least a start)
        • The process ends (sort of)
      • Strike an object (see above)
        • Continue the process, but accumulate results of traced ray(s).
      • Go to infinity, so no color from this ray.
      • The ray has "too many iterations"
        • handle this somehow
    • (An image I stole from online)
    • Great for light and shadows, shiny, translucient ...
    • Only good for one position.
  • Neither of these tehcniques are acceptable (yet) for real time graphics.