- This topic has 21 replies, 2 voices, and was last updated 2 years, 7 months ago by
.
-
Posts
-
When I make an ICC profile with displaycal of my wide gamut monitor (so no 3DLUT, no decklink, no resolve, simply displaycal straight to monitor in wide gamut mode), I obtain 99.8% sRGB, so with the 3DLUT I know I can’t go up to 100%, but I should be able to arrive at that same 99.8%, am I wrong? Instead I have 98.7, 1.1% less.
It depends on mesh resolution. Irrelevant.
If that’s what you’re referring to, I tried with both 33x33x33 and 65x65x65 resolution, same coverage result…
That is LUT3D resolution, but
a ) if display profile is idealized (matrix) it does not match display (to certain amount)
b ) if is is detailed, a 17x17x17 mesh is ~5000 patches.. and even with 5k patches resolution is not close to LUT3D mesh resolution.
anyway, not relevant.
For the display profile I always use XYZ LUT + matrix, not only matrix.
Anyway, so you’re telling me that 3DLUT resolution is so big that it inevitably results in inferior coverage?
But even if it were, what I don’t explain is that before using the mini converter I had reached a coverage of 99.8% sRGB, this 1.1 lower for no apparent reason trigger me.
It depends on mesh resolution. Irrelevant.
Explained before, nodes in LUT3D are interpolated from whatever source & destination colorspace meshes are.
And it is not relevant.
I think you should go to start point and understand what a LUT3D is, where it comes from, and what is native gamut and colorspace coverage.
Vincent, I thank you but if I’m here is because I’m trying to understand what’s going on and didn’t find any easy accessible informative documentation elsewhere.
And honestly I always end up not to get the answer I’m searching, continuing to struggle by myself.
For example, you named mesh resolution, if I try to search from google “mesh resolution 3d lut” nothing useful appear, there are stuff from 3d modeling softwares and their mesh, nothing about color grading or calibration, or profiling, so instead of sending me in search who knows where, would it not be more useful to write it directly here in simply words and solve the doubt? So that also future people like me have an easy answers too?
Draw it.
Get an A4 sheet, draw a curve. Use n equally spaced points to characterize that curve, draw straight lines between them (easy interpolation for sample purposes), call it MESH A. Now use m equally spaced points m>n and not multiple (non coincidents points) to characterize MESH A (not the original curve, which is now unknown to you). Draw straight lines between neighbor points in m set and you get MESH B.
That is what is happening there is a mesh aproximation on profilling, on SW interpolation, on LUT3D calculation, and on LUTD execution on your HW external lutbox… and none of these errors seem to be relevant at all (by magnitude).
