Anaglyph Shader Matrix Values

[jdeboi]

Is your feature request related to a problem? Please describe.
I noticed that the anaglyph effect really didn't look good with my red cyan glasses-lots of ghosting that doesn't appear when using an anaglyph Processing library based on the Dubois red/cyan algorithms. I did some digging, and (I think, but totally correct me if I'm wrong) the current three.js anaglyph shader uses right/left matrices that prioritize color purity over 3D.

I'm pretty sure the angler matrices this.colorLeft and this.colorRight (at least in the linked shader) are Dubois. I replaced the three.js matrices in the AnaglyphShader with these angler values (below), which helped with the ghosting of my models.

this.colorMatrixLeft = new Matrix3().fromArray( [
     0.456100, -0.0400822, -0.0152161,
     0.500484, -0.0378246, -0.0205971,
     0.176381, -0.0157589, -0.00546856
] );

this.colorMatrixRight = new Matrix3().fromArray( [
     -0.0434706, 0.378476, -0.0721527,
     -0.0879388, 0.73364, -0.112961,
     -0.00155529, -0.0184503, 1.2264
] );

Describe the solution you'd like
I think a simple comment in the AnaglyphEffect.js mentioning what type of anaglyph matrix / algorithm is being used (e.g. red/cyan or red/amber? Dubois or something else?). A more complete and complex solution might be an extra parameter to the anaglyph effect to set the type of matrices used in the shader based upon glasses color or desired effect.

Describe alternatives you've considered
As mentioned, I was able to copy the Dubois angler matrix values here, so maybe this isn't a feature request as much as a documentation request.

I've been trying to find a good resource w/ the computed matrix values. This is the best I've been able to find so far:
Processing lib and optimized anaglyph.

Anyway, thanks!

[Mugen87]

I think a simple comment in the AnaglyphEffect.js mentioning what type of anaglyph matrix / algorithm is being used (e.g. red/cyan or red/amber? Dubois or something else?).

Improving the inline documentation is a good start.

@tschw Do you mind clarifying the differences between the matrices used in AnaglyphEffect.js and the matrices from angler.js?

[makc]

@Mugen87 the AnaglyphShadertoyPlayer matrix seems to be coming from this paper where they took Dubois method and applied it to LCD spectral distribution instead of CRT. To quote their conclusion:

An anaglyph produced by the [Dubois] LS method is normally darker with less detail and requires brightening or gamma correction but appears to have no ghosting.

[Mugen87]

Good to know. But the theoretical background of the matrices used in AnaglyphEffect.js are still not clear, right?

[makc]

Those seem to be the values dumped to the console by tschw script under default settings:

tschw comment about them is here (but does not really explain much :~)

[Mugen87]

What do you think about replacing the matrices with the one from the paper? This is essentially what the OP did in his first post.

The example will look different with the new matrices but at least we have a clear resource we can link to.