[Tig] RGB to XYZ on Resolve

Rogério Moraes rogermamo at gmail.com
Fri Feb 17 00:23:55 GMT 2012


Humm.... This adresses a lot of issues around here (namely, inside my
head). The gamma, that I was pretty sure was being ignored or incorrectly
interpreted and specially the white point, wich people seems to forget that
is different at a P3 calibration/ LUT, to emulate (correct me if I´'m
wrong) the one at a DLP projector or xenon lamp film projector (although,I
know, this varies a lot...).. I now have a starting point to test some
other paths to a better workflow. All other inputs are welcomed still, but
for now: Thank you very much, Mike!

Em 16 de fevereiro de 2012 21:33, Michael Chenery
<michael at chenery.com.au>escreveu:

> I am not sure what Resolve is doing when doing internal conversions but I
> am pretty sure the rgb2xyz LUT is designed to convert DCI-RGB (P3, gamma
> 1/2.6) data to X'Y'Z' (XYZ with a gamma of 1/2.6).
>
> This means the X'Y'Z' data is created using math that assumes the RGB data
> has a P3 white point (x=0.314,y=0.351). This is why if this LUT is used to
> convert Rec.709 material to X'Y'Z' the resulting data looks greenish when
> viewed on a DCI compliant projector.
>
> A similar problem will also be occurring with the gamma. the rgb2xyz LUT
> is expecting RGB data encoded with a gamma of 1/2.6. Rec.709 data is
> encoded with a gamma of 1/2.4 so the resulting X'Y'Z' data will look
> 'heavier' with mid to low levels pulled down.
>
> Any RGB-to-XYZ conversion needs to make assumptions about how the RGB data
> is defined: what color are the primaries and white point what is the
> transfer function or gamma curve (if not linear to light). If any of these
> parameters change then a new RGB-to-XYZ transform (LUT) is required. To
> convert Rec.709 RGB data to XYZ requires a transform that uses math based
> on a RGB white point of D65 (x=0.3127,y=0.3290), the Rec.709 primaries and
> a encoding gamma of 1/2.4.
>
> Cheers,
> Mike
>
>   *_________________________________________________________________
>
> Michael Chenery
> Senior Color Scientist
> THX Ltd.
> *
> *
> *
> *
> *
> On 16/02/2012, at 10:59 PM, Rogério Moraes wrote:
>
> Sohonet http://www.sohonet.co.uk sponsors the TIG.
> Support from Nucoda www.imagesystems.tv
> Support from Blackmagic Design http://www.blackmagic-design.com/
> ====
>
> Yes! That´s some o f my problme plus I get darker images. Wich, for m, goes
> up to the question : in the RGB>XYZexporting the Resolve applies a 2.6
> gamma or preserves the native RGB gamma form the unrendered material, the
> one in the timeline (let´s say REC 709, a log-to-lin with a baked in LUT).
> the color conversion is pure martrix mathematics, and the numbers can be a
> little off in one or another software, thus the resulting greenish tone.
> But what about the Gamma? I use easyDCP and if I do bypass the color
> convertion I (and thye software) are assuming that the image is already DCI
> compliant in terms of color sapce (XYZ, and gamma). XYZ, easy to spot the
> difference, but what about gamma? Who can say for sure that the input image
> is already with the required gamma? All coments are welcome at this stage.
> Thanks all
>
> Yes! That´s some o f my problme plus I get darker images. Wich for me goes
> up toi the alley os the question : in the RGB correction the Resolve
> applies a 2,6 gamma or preserves the native RGB gamma form the unrendered
> material (let´s say REC 709, a log-to-lin Lut). the cocolor conversion is
> pure Martrix mathematics, and the numbers can be a little off in one or
> another software, thus the resulting greenish tone. But what abou the
> Gamma? I use easyDCP and if I do bypáss the color convertion I (and thye
> software) are assuming that the image is already DCI compliand in terms of
> color sapce (XYZ, and gamma). SXY, easy to spot the difference, but what
> about gamma? All coments are welcome at this stage.
> Thanks all
>
> ---------- Mensagem encaminhada ----------
> De: "Susiluoto Jarni" <Jarni.Susiluoto at kava.fi>
> Data: 16/02/2012 07:35
> Assunto: VS: [Tig] RGB to XYZ on Resolve
> Para: "Rogério Moraes" <rogermamo at gmail.com>
>
> Hello
>
> I've tried to use Resolve's own rgb2xyz LUT and render out both 10bit DPX
> and 16bit XYZ tiff. DCP were made with Curator and bypassing color
> conversion.  BW originals appeared to get greenish tone in projection...
>
> Have anyone any idea for what that Resolve's LUT is supposed to use for?
>
> Jarni Susiluoto
>
> Film Technician
> National Audiovisual Archive, Finland
>
> -----Alkuperäinen viesti-----
> Lähettäjä: tig-bounces at colorist.org [mailto:tig-bounces at colorist.org]
> Puolesta Rogério Moraes
> Lähetetty: 15. helmikuuta 2012 18:33
> Vastaanottaja: gold at duplicorp.com
> Kopio: TIG
> Aihe: Re: [Tig] RGB to XYZ on Resolve
>
> Sohonet http://www.sohonet.co.uk sponsors the TIG.
> Support from Nucoda www.imagesystems.tv
> Support from Blackmagic Design http://www.blackmagic-design.com/ ====
>
> No responses yet. It appears that the 8.2 Resolve deals with some DCI/XYZ
> issues( found a thread here http://www.knuterikevensen.com/?p=711 ). Since
> I do not have this version, and at 8.1 Resolve does not output .j2c files
> (basically, I'm trying to improve a DCP workflow) the gamma (or degamma)
> question still is the main point for me in here. Where in Resolve should I
> tell it to output the image (even if a tiff... my  DCP software will do the
> .j2c/mxf wrapping, without color converting to XYZ and the required gamma
> later, if the image is already DCI compliant) with a 2.6 gamma? Will I have
> to create a 3d LUT for this?
> Thank you all
>
> 2012/2/15 gold at duplicorp.com <gold at duplicorp.com>
>
> Hi Rogério,
>
> I didn´t find  a response to your question and I am very intersted in it.
>
> Did you receive a response or found anything new?
>
> Thank you in advance for any info you can give me.
>
> Daniel Goldschlager
>
>
> DRG New Media
>
> Caracas, Venezuela
>
>
> Original Message:
>
> -----------------
>
> From: Rogério Moraes rogermamo at gmail.com
>
> Date: Mon, 13 Feb 2012 12:15:18 -0200
>
> To: tig at colorist.org
>
> Subject: [Tig] RGB to XYZ on Resolve
>
>
>
> Sohonet http://www.sohonet.co.uk sponsors the TIG.
>
> Support from Nucoda www.imagesystems.tv Support from Blackmagic Design
>
> http://www.blackmagic-design.com/ Have a beer in Hollywood on 9 Feb.
>
> Details:
>
> http://tig.colorist.org/pipermail/tig/2012-January/019841.html
>
> ====
>
>
> Hello
>
>
>     In an experiment, I was trying to do a color conversion from a
>
> REC
>
> 709 colorspace to an XYZ, for DCI compliant images on a Resolve and as a "
>
> control group", did the same on AE. Got two different images as output....
>
> my question: does anyone knows what kind of conversion Davinci Resolve
>
> does when exporting (rendering) a REC 709 graded material (or in any
>
> RGB color
>
> space) to an XYZ .tiff? Does it apply a gamma of 2.6 for the XYZ color
>
> space or just converts the color gamut matrix for it's calculations,
>
> resulting in an XYZ colored image, but with a different gamma (2.2 is
>
> my guess)?
>
> Thanks for the attention
>
> --
>
> Rogério Moraes
>
> _______________________________________________
>
> http://reels.colorist.org
>
> http://tig.colorist.org/wiki3
>
>
>
> --------------------------------------------------------------------
>
> mail2web.com - Enhanced email for the mobile individual based on
>
> Microsoft® Exchange - http://link.mail2web.com/Personal/EnhancedEmail
>
>
>
>
>
>
> --
> Rogério Moraes
> _______________________________________________
> http://reels.colorist.org
> http://tig.colorist.org/wiki3
> _______________________________________________
> http://reels.colorist.org
> http://tig.colorist.org/wiki3
>
>
>


-- 
Rogério Moraes


More information about the Tig mailing list