I can’t get a good AA when there is a bright point near some lower value pixels. I tried everything including subpixel mapping and clamping output but only the workaround was to fake it in Photoshop for me.
Today I discovered that AA problem mostly appears when “Dont Affect Colors-Adaptation Only” is checked.
Below are 2 renders: for both of them
subpixel mapping enabled
exponential color mapping (so didnt enable clamp output)
As you see below,
The one with “Dont Affect Colors-Adaptation Only” has AA problems.
Normally clamping to 1 or slightly above one should fix this as the renderer has problems to sample proper if the difference between the pixels is too large. Works for me all the time.
It’s normal - the exponential color mapping and the subpixel mapping in the first case smooth out the AA, but if you don’t apply them to the image (i.e. the “Don’t affect colors” option is enabled), then obviously their effect is gone.
Why should they? They still affect the way V-Ray performs adaptive sampling. This was the purpose for adding the “don’t affect colors” option - people wanted the color mapping to affect the sampling (i.e. the noise levels), but not to influence the final colors.
sorry, I probably misunderstood something but your lines above sounded different
a further clearification could be useful because in the helpfile I can read:
Don’t affect colors (adaptation only) - when this parameter is on, the color mapping will not be applied to the final image, however V-Ray will proceed with all its calculations as though color mapping is applied (e.g. the noise levels will be corrected accordingly). This can be useful, for example, if you know that you will apply some color correction to the image later on, but wish to keep the rendering itself in linear space for compositing purposes. Note that the Clamp output option will have an effect regardless of the value of the Don’t affect colors option.
there is nothing that talks about subpixel mapping
that said, I understand that disable them it’s not a good idea
So if I am understanding you, it’s good for when we are rendering 32bit (say exr’s) that we want to output at gamma 1.0 for the render elements, and do the corrections in post? (But the noise sampling knows that we are correcting in post.)
