Anders,
This is not a matter of adding or not the feature. It would be easy
for me just to promote the tables and use them in floating point mode.
The issue is, there are profiles around using the limited
precision as a trick to do other things, and doing what you propose
would break color rendition on those tricky yet legal profiles.
Thinking a little bit more on that, I believe it could be done by
using a plug-in. If I understand correctly, you want to limit this
to matrix shaper profiles, like the old sRGB, and only in floating
point mode. Is that right?
Using a optimization plug-in, you could capture the cases where
input and output profiles a both matrix shaper and the CMM is
operating in foating point. The actual code does something similar
wwhen using 8 bits, to provide high performance. You could do
some similar for high accuracy.
In the documentation is explained how to build such plug-ins.
Look for "optimization plug in". The "myOptimize" function
described in the manual should be changed to something like that
if (*InputFormat == TYPE_RGB_FLT &&
*OutputFormat == TYPE_RGB_FLT &&
cmsPipelineCheckAndRetreiveStages(Src, 4,
cmsSigCurveSetElemType, cmsSigMatrixElemType,
cmsSigMatrixElemType, cmsSigCurveSetElemType,
&Curve1, &Matrix1, &Matrix2, &Curve2))
{
- Store the matrixes and the curves elsewere
_cmsPipelineSetOptimizationParameters(*Lut,
MyEvaluator, NULL, NULL, NULL);
}
Then in MyEvaluator() do whatever interpolation you wish
with matrix and curves. This approach would capture only
matrix-shaper pipelines, so the huge rest of other
profiles and formats will be handled transparently by the
CMM.
Sorry, I really cannot break compatibility because other
users, but I think plug-ins gives you a clear path forward.
Please let me know if you need any additional help with
the plug-in.
Regards
Marti
Quoting Anders Torger <tor...@ludd.ltu.se>:
> Ok, I was afraid that this would be the response. People are using
> legacy icc profiles all the time so having poor support for them is not
> really a great idea, at least not for our project. As a first step we
> can implement matrix shaper conversions ourselves though as it's quite
> easy to do, but it would be great if we could get floating point
> support throughout *including* legacy icc profiles.
>
> /Anders
>
> On 11/24/2013 01:25:34 AM, marti.ma...@littlecms.com wrote:
>>
>> Quoting Anders Torger <tor...@ludd.ltu.se>:
>>
>>
>> Hello,
>> > with the gamma curve specified as an array of numbers
>>
>> That's the whole point. If the gamma is specified as an array of 16
>> bits numbers, then the interpolation happens in 16 bits. This is
>> because the profile, not the CMM. In fact, if the profile uses
>> multiprofile elements and a table of floats, then the interpolation
>> happens in float precision.
>>
>> I would recommend to use modern (V4) profiles, with parametric curves
>>
>> or much better with multiprofile elements. This runs well in
>> unbounded
>>
>> mode.
>>
>>
>> Regards
>> Marti
>>
>> > Hello,
>> >
>> > I'm one of the developers of RawTherapee, a photo editing software
>> with
>> > a floating point engine and which uses lcms2 for color management.
>> >
>> > Recently I discovered that when making a transform from float to
>> float
>> > (or float to 16 bit) using a basic icc with the gamma curve
>> secified
>> as
>> > an array of numbers the output contains fewer distinct values than
>> the
>> > input. If the gamma is specified with a single number (ie pow(x,
>> > g/256)) but otherwise the same icc (not LUT, just matrix) the
>> output
>> > is the expected, ie about the same amount of distinct numbers as
>> the
>> > input.
>> >
>> > Thus there is some precision lost in the curve interpolation code.
>> >
>> > If I've managed to navigate the code correctly it happens in
>> > cmsEvalToneCurveFloat() where a 16 bit integer path is taken
>> despite
>> > it's perfectly possible to make a floating point interpolation.
>> >
>> > As we want to maintain precision with widely used legacy icc files
>> we'd
>> > like to see this fixed.
>> >
>> > /Anders Torger
>> >
>> ------------------------------------------------------------------------------
>> > Shape the Mobile Experience: Free Subscription
>> > Software experts and developers: Be at the forefront of tech
>> innovation.
>> > Intel(R) Software Adrenaline delivers strategic insight and
>> game-changing
>> > conversations that shape the rapidly evolving mobile landscape.
>> Sign
>> up now.
>> > http://pubads.g.doubleclick.net/gampad/clk?id=63431311&iu=/4140/
>> ostg.clktrk
>> > _______________________________________________
>> > Lcms-user mailing list
>> > Lcms-user@lists.sourceforge.net
>> > https://lists.sourceforge.net/lists/listinfo/lcms-user
>>
>>
>>
>>
>
>
>
> ------------------------------------------------------------------------------
> Shape the Mobile Experience: Free Subscription
> Software experts and developers: Be at the forefront of tech innovation.
> Intel(R) Software Adrenaline delivers strategic insight and game-changing
> conversations that shape the rapidly evolving mobile landscape. Sign up now.
> http://pubads.g.doubleclick.net/gampad/clk?id=63431311&iu=/4140/ostg.clktrk
> _______________________________________________
> Lcms-user mailing list
> Lcms-user@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/lcms-user
------------------------------------------------------------------------------
Shape the Mobile Experience: Free Subscription
Software experts and developers: Be at the forefront of tech innovation.
Intel(R) Software Adrenaline delivers strategic insight and game-changing
conversations that shape the rapidly evolving mobile landscape. Sign up now.
http://pubads.g.doubleclick.net/gampad/clk?id=63431311&iu=/4140/ostg.clktrk
_______________________________________________
Lcms-user mailing list
Lcms-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/lcms-user
