|WikiProject Computer graphics|
In my opinion, this shall not be merged. Cubemaps and reflection mapping are different things. Notable examples:
- Spherical reflection mapping (reflection mapping not using cubemaps)
- Dual-paraboloid reflection mapping (idem)
- Normalization cubemaps (a cubemap not being used for reflections).
- Weird atlases with fragment shaders (not even sure this could exist but I don't see why it should not - mangles texture coords to access right face: as a pro over traditional atlases, no mip bleeding shall take place).
So, I vote against the merge.
MaxDZ8 talk 07:31, 4 July 2006 (UTC)
Cube mapping is a texture mapping technique with numerous applications. If it is used for environment mapping, it can be used for refractions as well as for reflections. This is done for example in a technique represented in http://www.springerlink.com/content/u518x461461642j1/ In this technique cube mapping is also used for an acceleration structure for geometrical data.
In any case, cube mapping can be used for several different purposes, many of them becoming as important as reflections in future.
My vote goes against the merge as well.
- oppose. Per MaxDZ8. Another use for cube maps is refraction (or at least, I've used them for that). - Rainwarrior 16:57, 3 September 2006 (UTC)
Cube Maps are "Cube Maps", Sphere Maps are "Sphere Maps", both have more applications than just reflection, and Reflection is accomplished via more techniques than either, The best results in reflection come from utilizing a seperate render target. Not a Cube Map or a Sphere Map. Whoever suggested a merge should have his internet privileges suspended for a month. :) —Preceding unsigned comment added by 220.127.116.11 (talk) 01:04, 18 April 2009 (UTC)
Incomplete, Dubious claims
I feel like a parabolic environment mapping should be mentioned! after all you mention some obscure technique that isn't used all that much but neglect this one. kinda awkward. Also, I've never heard of HEALPix being supported on any graphics card I've ever worked on, so the claim of it being faster is highly dubious and Im not an expert by any means on the algorithm but from what I understand it's hierarchal in nature which would imply severe performance losses because of iterative or recursive procedures. Also Spherical harmonics aren't mentioned either, these are used for diffuse reflection mappings. Also, the claim that this is faster then ray tracing is not true for all scenes considering the current state of the art. Although the misconception is a common one.
I agree, parabolic shall be added.
The stuff on healPIX by sure doesn't need its own paragraph, it's just a shader trick (although very valuable) and I believe it shall not be put on the same levels as historical approaches. I'm rather sure no card will ever support this and there's no more need to have it in HW the first place.
Spherical harmonics used for diffuse mappings? Spherical harmonics are used for a ton of stuff but again, they are more like a shader trick than a real reflection mapping method.
As for the ray tracing VS environmental I don't believe it's a problem.
In general however, this article needs more sources.
MaxDZ8 talk 13:23, 27 December 2007 (UTC)
HEALPix performance claim
The claim about HEALPix being the fastest method seems like a false claim to me. After looking up some sources on it, it seems that the partitioning-scheme makes it difficult/slow to perform filtered lookups when four neighboring samples are located in different areas - consumer-grade graphics hardware filtering cannot do this directly. The only performance advantage over cube mapping that I can see is less distortion in some angles, which can allow comparable visual quality with slightly smaller maps. But the fact that with todays graphics hardware it requires separate arithmetic operations, makes me believe this claim to simply be false. Perhaps not if future graphics processors include direct hardware support for it, but I believe this to be unlikely. --Kusmabite (talk) 11:26, 12 April 2008 (UTC)
In defense of the HEALPix performance claim
I thought it was saying that there were fewer operations in an unaccelerated implementation than there would be for other mappings. Even if I'm right, that makes it a moot point when it comes to application, except in environments where hardware acceleration isn't available. Sobeita (talk) 00:42, 22 August 2014 (UTC)
I was going to go ahead and delete everything to do with HEALPix on this page, but then i saw that there was an existing discussion on the topic.
Input information regarding fast shadows and texture mapping
Hello! Disclaimer: I am new here.
I wanted to provide this link: http://dl.acm.org/citation.cfm?id=192193 which describes techniques for reflections specifically and therefore should be relevant to this page, I felt. From the abstract:
"Surface reflections of an environment can be rendered in real time if hardware calculates an unnormalized reflection vector at each pixel. Conventional perspective-correct texture hardware can then be leveraged to draw high-quality reflections of an environment or specular highlights in real time. This fully accommodates area light sources, allows a local viewer to move interactively, and is especially well suited to the inspection of surface orientation and curvature. By emphasizing the richness of the incoming illumination rather than physical surface properties, it represents a new direction for real-time shading hardware."
Unfortunately I do not have membership with the web site hosting the document, but there is some good description of the information that should be valuable for citations. I also have a special connection with one of the authors, I am related to Jim Foran so I can query him directly about this technology.