Second Life landscape by Vania Lahtoh, CC-BY-SA

Image Files

Image data is often saved in or loaded from a file

There are many common image formats: JPEG, TIFF, PNG, GIF, etc

Features which may vary between formats are:

• Color channels: greyscale, RGB, RGBA
• Bit resolution
• Compression: lossy / lossless

Texturing

Textures vs. Images

Drawing an image (glDrawPixels in OpenGL)

Applies transformations to the starting point

After that, skips the 3D rendering pipeline and just copies pixels directly from the image to the framebuffer

Graphics.Blit in Unity3d is roughly similar

Texturing

Is a stage in the 3D rendering pipeline, for polygon data

Determines the color of each pixel while filling in a polygon

                  

Texturing

Applying textures involves the following steps:

• provide a texture image
  
  
• set modes (repeating, filtering, etc)
  
  
• load a pixel shader that uses the texture
  
  
• assign texture coordinates for each vertex

Texture Objects

Textures usually originate as image data in main memory (often from a file)

Data is passed to the graphics card for rendering

A "texture object" holds the data in graphics card memory for repeated use

Texture size

Note: Historically, width & height had to be powers of 2 - i.e., 2, 4, 8, 16, 32, 64, 128, 256, 512, etc.
This is often still a good idea.

Maximum size is typically 4096 x 4096

Texture Coordinates

Objects need texture coordinates to define how a texture is applied

Texture coordinates form a coordinate system on the texture image, ranging from 0 to 1 in each direction

They are referred to as U & V (Unity3D, Maya, etc) or S & T (OpenGL)
(to distinguish them from X & Y geometric coordinates)

Texture Coordinates

One texture coordinate should be assigned for each vertex

This defines what part of the texture should be used at that point

When a polygon is filled in, the vertices' texture coordinates are interpolated across its surface, to find the texture data for each pixel

Texture Coordinates

A texture does not have to be evenly applied to a polygon.

The image can appear distorted, either as a result of the geometry or the texture coordinates used.

Texture Tiling

Texture coordinates outside the range 0 - 1 can be used to tile a texture

The "wrap" mode can select between repeating or clamping the texture

Textured Mesh

Filtering

Applying a texture involves sampling - for each pixel drawn on the screen, OpenGL must compute a color from the texture image.

Texels (texture pixels) rarely match up exactly 1-to-1 with the pixels on the screen .

Filtering modes select what to do when texels are larger than screen pixels (magnification), or when texels are smaller than screen pixels (minification).

    

Common modes:

• NEAREST - choose the texel nearest the center of the pixel
• LINEAR - average the four nearest texels
• ANISOTROPIC - average multiple texels, accounting for angle of surface
• MIPMAP - (see next page)

Mipmapping

Mipmapping uses versions of the texture at smaller & smaller sizes image, to effectively pre-compute the average of many texels.

As a texture gets more and more minified, many texels will correspond to a single screen pixel, and using a "nearest" or "linear" filter will not compute an accurate result. The texture will appear to flicker as the object moves.

A mipmapped texture avoids the flickering problem when minified. As a result, it can look a bit fuzzier.

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