Creating and Editing Texture Coordinates (UVs) - An Introduction to Computer Graphics for Artists - page 152

Table 10.1 Power of two sizes and linear equivalents in centimeters
Pixels 2 4 8 16 32 64 128 256 512 1,024

cm 0.78 1.56 3.13 6.25 12.5 25 50 100 200 400

Fig. 10.9 When a square texture is used to fi ll a non-rectangular polygon, the effi ciency of the
map is reduced due to lost and duplicated pixels

10.2.5.1 Map Size Limitations

Most game and fi lm studios require that all texture maps are power of two dimensions.
The reason for using power of two sizes is that renderers search for maps in memory
more effi ciently when they meet this requirement. Objects rarely fi t into power of
two dimensions though, leaving artists with a choice of stretching their maps (which
increases the pixel count and distorts the pixels) or not using all of the available
texture space (Wang et al. 2008 ; Dong et al. 2009 ). Either way, an average of about
40 % of the texture map is wasted because of this limitation. See Fig. 10.9 for an
example of this.
Square maps limit an artist's options even more, by forcing an aspect ratio that
will not match the majority of the objects in a scene. Multiple objects can use different
parts of the map, but this reduces the overall resolution of the scene by fi tting more
polygons into the same UV space.
For some renderers, square or power of two-sized maps are not optional, but
for most, they are. This requirement is normally the default setting in game engines,
but it is a switch and can be turned off. Some offl ine renderers use a similar switch,
but again, it can be turned off. The argument in favor of the limitation is that it speeds
up render time. The argument against is that it reduces the quality of texture maps.
The way it reduces the quality of texture maps is by making it very diffi cult to use a
consistent scale (see Sect. 10.2.5.2 and Fig. 10.10 ) .
When dealing with map size limitations, quality will suffer unless your work
takes the limitation into account when you make your maps. The idea is to use maps
that are all exactly the same scale relative to each other, and map sizes that are the
same aspect ratio as the unfolded object the map covers. When this cannot be done,
the next best option is to use UV layouts that combine many objects on one map (to
make any distortion affect the group equally) or to leave as much unused space on a
map as necessary to avoid distortion of pixels.