Re: [DCP-o-matic] XYZ image sequence colour problems

On Jun 17, 2017, at 06:14 000AM, Carsten Kurz via DCPomatic &lt;dcpomatic(a)carlh.net&gt; wrote: Another ‚indicator‘ - I took your rec709.TIFF, fed it through OpenDCP to make an XYZ-transformed J2C from it. Then converted to TIFF (with no color transform). That is, to see how a ‚proper‘ XYZ-TIFF of that content should look like, compared to the version that the post house delivered. See attached picture (clockrec709_xyz.tiff) This XYZ-TIFF converted from the rec709.TIFF looks very different from the post house version. Of course, I also fed this XYZ-TIFF through DCP-o-matic, applied ‚no color conversion‘ - and the result in the DCP looks identical to the rec709 version (aside from some minor gamma changes, but that shouldn’t bother us for now (rec709, 16Bit/8Bit, etc.). I also attached three snapshots from the same DCP, converted from all these sources with DOM: - first image (DCP_snap_TIFF_XYZ_org.png) is what you know - the original posthouse XYZ-TIFF converted to DCP with ‚none‘ - greenish clock - second image (DCP_snap_TIFF_rec709_org.png) is your original rec709.TIFF, converted to DCP with the rec709/sRGB conversion - white clock - third image (DCP_snap_TIFF_rec709_XYZ.png) is the rec709.TIFF, converted to XYZ->J2C->XYZ-TIFF, and then again converted to DCP, again with ‚none‘ - white clock. You see that this XYZ-TIFF I created from the rec709.TIFF, with no color conversion applied in DOM now matches the rec709 and thus your preferred look very closely. Now, I am not a color scientist, but to me this seems to prove that whatever the post house gave you was not made to transfer 1:1 into a DCI J2C/MXF… So I’d get back to them and try to solve it with them. - Carsten <clockrec709_xyz.tiff><DCP_snap_TIFF_XYZ_org.png> <DCP_snap_TIFF_rec709_org.png><DCP_snap_TIFF_rec709_XYZ.png> _______________________________________________ DCPomatic mailing list DCPomatic(a)carlh.net http://main.carlh.net/cgi-bin/mailman/listinfo/dcpomatic