2007-12-06 10:02:13 +00:00
|
|
|
#summary Mipmap Generation
|
|
|
|
|
|
|
|
== References ==
|
|
|
|
|
2007-12-06 10:02:27 +00:00
|
|
|
* Jon Blow articles in Game Developer Magazine:
|
2007-12-06 10:03:11 +00:00
|
|
|
* [http://number-none.com/product/Mipmapping,%20Part%201/index.html Mipmapping, Part 1]
|
2007-12-07 03:23:10 +00:00
|
|
|
* [http://number-none.com/product/Mipmapping,%20Part%202/index.html Mipmapping, Part 2]
|
|
|
|
|
|
|
|
* [http://developer.nvidia.com/object/np2_mipmapping.html Non-Power-of-Two Mipmapping]
|
|
|
|
|
|
|
|
== Conclusions ==
|
|
|
|
|
2007-12-07 03:26:35 +00:00
|
|
|
Here are some of my findings:
|
2007-12-07 03:23:10 +00:00
|
|
|
|
2007-12-07 03:26:35 +00:00
|
|
|
* Windowed sinc filters with radius greater than 3 produce too many ringing artifacts.
|
2007-12-07 03:23:10 +00:00
|
|
|
* For good results with non power of two textures or arbitrary scales, you must use polyphase filters.
|
|
|
|
* Sampling the polyphase filter with a box filter greatly improves the quality.
|
|
|
|
* Using a triangle filter looks even better, but the difference is small.
|
|
|
|
* For most cases box does a great job.
|
2007-12-07 03:26:35 +00:00
|
|
|
* The Lanczos and Kaiser windows produce indistinguishable results.
|
|
|
|
* Gamma correction produces more correct results, but sometimes enhances the ringing artifacts.
|