Intro to Textures

• This material is from Angel
• Textures are patterns
  • Textures provide depth cues (size and location)
  • And shape information
  • The red book says "Glue" a picture onto a surface.
• Texture mapping is looking up colors for a pixel in a table.
• One, two and three dimensions
  • One dimensional, a stipple for a line
  • Two dimensional a surface covering
  • Three dimensional, a block of wood, as you carve it
• Textures are often images
  • But can be computed in the cpu
  • Or even in the GPU on the fly
  • Reflections and shadows have been handled in this manner.
• The basic unit of a texture is a texel or a texture element.
• Angel says that texture mapping involves four types of coordinates
  • Object coordinates
  • Screen coordinates
  • Texture coordinates - usually mapped to object coordinates but used when filling a pixel in screen coordinates
  • Parametric coordinates, for filling as we have seen before.
• Let us consider a two dimensional texture (an image or pattern) mapped onto a surface.
  • Stored as an array of values in memory.
  • But think of it as a continuous patter T(s,t)
  • s and t are texture coordinates, 0 ≤ s ≤ 1, 0 ≤ t ≤ 1
• A texture map associates a texel with each point in a geometric object.
  • One challenge is to find a mapping from a point in object space (x,y,z,w) to a point in the texture (s,t)
  • This is usually done through a parametric coordinate mapping in terms of (u,v).
• For a pixel
  1. Find the portion of an object which maps to a pixel.
  2. Find the corresponding area of the texture that maps to that pixel.
  3. Modify the color of the texture at that point by the lighting equations
  4. Assign the new color to the frame buffer.
• Step 2 is difficult, and may result in a non-linear mapping (distortion)
  • Textures are usually rectangles.
  • These don't map the best to spheres or other curved surfaces.
  • This also requires an inverse computation from pixel to object to texture coordinates.
  • And in general, we are interested in mapping areas to areas, not points to points.
• 
• The above image indicates some of the difficulties
  • The preimage of a pixel on a surface might not be square
  • And it might map to multiple "values" in the texture.
  • Just selecting the "center of the pixel" value leads to serious antialiasing problems.
  • But "averaging" the corners and center leads to other imaging problems, but is a better solution.
• In the end, a section of the texture is mapped to a section of a surface
•