Content-Type: text/plain; charset=utf-8
Mime-Version: 1.0 (Mac OS X Mail 9.2 \(3112\))
From: Dave Rice <dave@dericed.com>
In-Reply-To: <568EE224.8020101@das-werkstatt.com>
Date: Wed, 25 May 2016 12:19:19 -0400
Content-Transfer-Encoding: quoted-printable
Message-Id: <CF71F9AD-1B30-46B5-BA1A-1495C48C5E0A@dericed.com>
References: <CAO7v-1TCNOcqhv=JR5dMWhibD=kLyQwLztk8RG3v-rtegn7s7Q@mail.gmail.com>
 <E89E5459-836A-4FD7-BA52-6F75AFD67083@dericed.com>
 <20160105163439.GA13213@nb4> <568BF59B.6050106@mediaarea.net>
 <20160105174445.GC13213@nb4> <568C0516.3040605@mediaarea.net>
 <CAO7v-1R44SDSZWZ_55SBT0vE+ka7e+s5V9suiW3emM-Z-ViS2A@mail.gmail.com>
 <568EA7D2.50503@das-werkstatt.com> <568EAFC3.1000202@mediaarea.net>
 <568EE224.8020101@das-werkstatt.com>
To: Peter Bubestinger <pb@das-werkstatt.com>
Archived-At: <http://mailarchive.ietf.org/arch/msg/cellar/tITaToTTxpuTgoAYeIO7YR5YYvk>
Cc: cellar@ietf.org
Subject: Re: [Cellar] FFV1 Version 4
Precedence: list


> On Jan 7, 2016, at 5:09 PM, Peter B. <pb@das-werkstatt.com> wrote:
>=20
> On 01/07/2016 07:34 PM, Jerome Martinez wrote:
>> PS: just to be clear, when you speak about "logarithmic", you speak
>> about DPX transfer characteristic value 3 "Logarithmic [to be defined
>> by SMPTE I23 Technology Committee,
>> sub-group on =E2=80=9CTransfer Characteristics=E2=80=9D]", right?
>> (note: the "to be defined" is in the spec from 2003 as well from the
>> spec from 2014 :), without further description)

The poster of the DPX file at =
https://github.com/epiil/dpxdepot/tree/master/test_files/dpx_frame =
claims that it is log. But the Transfer characteristic value is 0x01 aka =
"Printing density", rather than value 0x03 as Jerome supposes. Is =
"Printing density" also considered as log.

Is this data maintainable in Matroska with the new =
TransferCharacteristics element set to option 9 aka "Log" or should FFV1 =
contain this type of data.

> Well, personally I'd say I'm speaking about "color values that are
> non-linear".
> In practice, I've only encountered "quasi-log(TM)" (aka "logarithmic) =
-
> and DPX being the most popular case for these non-linear color values.
>=20
> So in order to stick to real-world use-cases I mentioned
> "logarithmic-DPX-to-FFV1-and-back".
> Could also be "Non-linear-Foo-to-FFV1-and-back" ;)
>=20
>=20
> Maybe it would make sense to collect what kind of non-linear color
> interpretations appear in the real world and derive some way of =
handling
> them.
> At least the most prominent ones?
>=20
> For example, like the ones Reto mentioned:
> // -------------------
> - log RGB (=3D "quasi-log", sometimes also called "pseudo-log") =
encodings:
> FilmStream (log60), ARRI log F, Panalog, Sony-log (=3D S-log), REDLOG, =
SI
> log90
> - log neg encodings: Cineon printing density (CPD), "DPX", ARRI log C
> // -------------------
>=20
>=20
> Regards,
> Pb
>=20
>=20
> PS: Maybe something like a CLUT could be stored as a self-contained =
way
> of interpreting the non-linearity of the color values?


For reference here is the list for DPX transfer characteristics which is =
quite different from the Matroska list.

0 User defined
1 Printing density
2 Linear
3 Logarithmic [to be defined by SMPTE I23 Technology Committee, =
sub-group on =E2=80=9CTransfer Characteristics=E2=80=9D]
4 Unspecified video
5 SMPTE 274M
6 ITU-R 709-4
7 ITU-R 601-5 system B or G (625)
8 ITU-R 601-5 system M (525)
9 Composite video (NTSC); see SMPTE 170M
10 Composite video (PAL); see ITU-R 624-4
11 Not applicable
12 Not applicable
13 =E2=80=93 254 Reserved for future use

Dave Rice