New profile making software: Lumariver Profile Designer

[daicehawk]

It's all three axes (the "tint" is the hue axis), you can make modification in a gamut sphere around the anchor color, just not that many DE. If the patch is several DEs off then there is something really wrong though.

I am sorry, but Hue is warm\cool (sun\sky or b Lab axis) plus tint magenta\green (wet\dry or a Lab axis). Saturation and Lightness caanot make up the b part of the Hue.

[narikin]

I simply using an ISIS XL and x-rite software to build my own Epson printer profiles - sorry for being daft, but which version should I be using, and is there any way to carry on using the Isis automated reading with Lumariver?

(great name btw!)

[torger]

I simply using an ISIS XL and x-rite software to build my own Epson printer profiles - sorry for being daft, but which version should I be using, and is there any way to carry on using the Isis automated reading with Lumariver?

Lumariver Profile Designer is for input profiles for cameras and scanners, that is you can't make printer profiles with it 

[torger]

I am sorry, but Hue is warm\cool (sun\sky or b Lab axis) plus tint magenta\green (wet\dry or a Lab axis). Saturation and Lightness caanot make up the b part of the Hue.

Hmmm... I see your point. Maybe I should change to Jab before people start getting used to JCh. It will be an incompatible behavior change, but I don't think many have started using those adjustments yet. I'll think about it. Or I simply add A/B sliders on top, it may make it more user friendly overall. Here's an example showing both sliders side by side:

http://bl.ocks.org/connorgr/0a299fe77d5c7feccd22e02f2ac5d69b

C/h affects a/b simultaneously. The shape of the local volume you can move around with the sliders will differ though, Jab will be more uniform in all directions, while JCh is more pie-shaped, and with the sliderlimited range on h you can't move effectively near the neutral axis, so I guess it's anyhow a good idea to have ab sliders.

In all, early adopters may need to be prepared that the software is moving a little bit here in the start before everything settles...

The fun thing with a lot of interest is that there is a lot of interest. The drawback is that I'm getting this backlog that grows faster than I'm reducing it... time to scrap my weekend plans...

[torger]

Done some more thinking and there will be some A+B stuff added, maybe not all at once though. Chroma+hue can technically address all positions A+B can, but when you have the range limited to locally around an anchor color that's not the case. Still adjusting chroma isolated from the other dimensions are important and hue+chroma is considered by many as a more user-friendly color addressing, so A+B won't replace the current, there will be some sort of additional control...

[Alexey.Danilchenko]

My current view on SSF is that it's good for research, testing targets, how large impact different color filter shapes has etc, but that it seems to often suffer from quite large measurement errors.

Really? I would have said the same about targets. Why would there be even a commercial products (albeit with strange approach and costing a fortune) to take sensor spectral sensitivity curves?

To the profiler SSF always looks like "perfect", there's never glare or uneven light or those kind of non-linear measurement errors. When shooting targets the profiler gets to see the measurement errors, with SSF there are no errors as seen by the profiler -- but is the SSF provided actually matching the camera's real SSF or is it differing substantially due to measurement error when making the SSF?

What errors would those be? I take it you have not spend long actually looking how we do the measurements and what is involved.

The monochromators are easily calibrated and unlike targets do not have a habits of changing too much over time. Integrating spheres taking care of the uneven lighting. Measuring spectrum simultaneously with taking shot at each wavelength takes care of the light source fluctuations and allows to even out the response to equal-energy source. Moreover the spectrum measurements and monochromator in pair allow verifying each other. The spactral measurement that I have done with Hamamatsu sensor are a lot more prcise and stable than i1Pro (not to say they caprured with more resolution and less noise on a borad that cost less than 100$ to build). With this approach you can take measuerments of spectral curves with any light really - even tungsten halogen (xenon works best though and Iliah is looking at high power LEDs).

I have no gear to measure SSF myself, so I have just tested the stuff found on the net from some academic projects, and for my old 5DMk2 I say that shooting targets leads to a more accurate profile than using those SSF, even if you have moderate measurement errors on the target. But the theory is if you have proper accurate SSF, they should be great... I just don't know how easy it is to get accurate SSF measurements...

Do you know how those were obtained? Have you factored in the lens and the fact that your very 5Dm2 may have slightly different curves?

My experience is dramatically different from yours - the very first spectral curves I took manually using just monochromator and sphere and adjusting the light source intensity crudely gave me much better matrix profile that I never succeeded building with targets.

[torger]

I'm glad SSF is working out for you, and it's a super-cool DIY project (I'm a bit interested in doing it myself!), and you will be able to use the measurements in the upcoming patch release, although I don't see that big value using LRPD when making matrix-only profiles and already used to programming and command line tools -- then just use DCamProf instead. It's not like SSF matrix-only users is going to be a big customer group...

Measurement errors I'm concerned about would be sensitivity at a specific wavelength (too high or low), not that the wavelength in the monochromator would be off. That it is transmissive light is an advantage though. Flare can be an issue, but not glare.

I once asked Hasselblad's Ove B if they were using SSF for making their profiles. They don't. The thing is that making a great general-purpose profile is not so much about target or measurement method if you just have decent precision, the rest is about tone operators, highlight handling, gamut handling, possibly manual tuning of individual colors etc.

Even if you have 100% precision in the measurement, the camera can't make a 100% match, and then the profiler needs to make an approximation. It makes a good job automatically, but if you're really picky you may want to adjust those approximations by eye individually, which LRPD allows. And as soon as you are doing manual adjustments by eye, that 100% exactness in measurement no longer matters.

[howardm]

Sorry to pollute the thread but if you are successfully running the v1.0.1 version of the tool under Yosemite (OSX 10.10), pls give a shout out.  It looks like there may be a Mac OSX 'codesigning' problem w/ v1.0.1.

[torger]

Sorry to pollute the thread but if you are successfully running the v1.0.1 version of the tool under Yosemite (OSX 10.10), pls give a shout out.  It looks like there may be a Mac OSX 'codesigning' problem w/ v1.0.1.

I think the link you sent me in PM points to the problem, a technical build/compile issue with the third party-library libarchive (used by LRPD to archive projects). It will work on 10.11 and 10.12 (which I'm running myself), but the software won't start on 10.10. I'll look into the issue as soon as I get time. Meanwhile a work-around is to update to the latest MacOS.

[howardm]

I think you may end up having to manually compile (and set the min version) of the libarchive stuff, unless you can convince 'brew' to change it's default min_version at install time or force a rebuild.

updating OSX is a big deal for me and not desirable.

I"m not scared by command-line

[Alexey.Danilchenko]

I'm glad SSF is working out for you, and it's a super-cool DIY project (I'm a bit interested in doing it myself!), and you will be able to use the measurements in the upcoming patch release, although I don't see that big value using LRPD when making matrix-only profiles and already used to programming and command line tools -- then just use DCamProf instead. It's not like SSF matrix-only users is going to be a big customer group...

And why are you making the assumption that it is usable for matrix only? Iliah was building camera profiles using monochromator approach for quite some time.

Measurement errors I'm concerned about would be sensitivity at a specific wavelength (too high or low), not that the wavelength in the monochromator would be off. That it is transmissive light is an advantage though. Flare can be an issue, but not glare.

I am sorry but that sounds a bit nonsensical. I did explain how sensitivity is measured and adjusted at each stage.

I once asked Hasselblad's Ove B if they were using SSF for making their profiles. They don't. The thing is that making a great general-purpose profile is not so much about target or measurement method if you just have decent precision, the rest is about tone operators, highlight handling, gamut handling, possibly manual tuning of individual colors etc.

I would not base my opinion on one company that has little to show for their colour reproduction know-hows. I'd start with Kodak, Hunt books and perhaps talking to former Kodak engineers regarding processes involved etc.

Even if you have 100% precision in the measurement, the camera can't make a 100% match, and then the profiler needs to make an approximation. It makes a good job automatically, but if you're really picky you may want to adjust those approximations by eye individually, which LRPD allows. And as soon as you are doing manual adjustments by eye, that 100% exactness in measurement no longer matters.

Nobody prevented a check to see how practical measured SSFs are. You need at least illuminant and lens spectral curves to make realistic use of them.

I sense that this is a dead end discussion - I am not going to pollute this thread anymore. I wish you all the luck with the product even though I am not going to use it.

[scyth]

...

ну вот на пустом месте же...

[torger]

Hey Alexey, relax

I guess I missed this part "Measuring spectrum simultaneously with taking shot at each wavelength takes care of the light source fluctuations and allows to even out the response to equal-energy source.", that of course greatly reduces measurements errors. Sorry. Don't need to be mad. I'm already very impressed by your achievement and glad that you share experience and source code so others can try it out.

I wouldn't consider Hasselblad to have little to show about color know-how. The medium format companies live and die by the color reproduction. Not all love it of course, as tastes differ (there's a reason I've made a profile maker...). Phase One doesn't use SSF either (AFAIK), but targets, their look and hand edits (not sure about their repro profiles though). There are just different methods. What I was trying to say is that even if SSF is a superior measurement method, which it surely is when done right, it's not a guarantee for superior profiles as there's so many more aspects to it. And likewise, a profile made from a reasonable good target measurement won't automatically be inferior to one based on SSF.

Good measurements always help though.

And even if you say you won't use the software now, and as said I think if you favor matrix-only profiles DCamProf is probably doing all you need already, I did register your vote on SSF features a couple of pages back in this thread and that was a reason that I'm actually adding in CGATS raw value reading to the next LRPD patch release, so I do hope you give it a look anyway when you get time, could be fun. I try to listen to my users. Daicehawk just got my eyes opened for the value of AB chromaticity adjustments, but again not without a little debate... every feature requires quite some effort to implement so I need to test the arguments a bit.

[torger]

I think you may end up having to manually compile (and set the min version) of the libarchive stuff, unless you can convince 'brew' to change it's default min_version at install time or force a rebuild.

Yeah I'll have a look. Building libarchive myself should not too much of a mess, I'm more worried that it's the same problem with the OpenMP lib (also brew). I have like a mountain of things to do to the next patch release. Or actually not too many things, but I'm a bit stressed by all input coming at the same time, but then again I put myself into this situation so I guess I get what I deserve...

[howardm]

understood.  no worry/rush.

[Alexey.Danilchenko]

ну вот на пустом месте же...

Не на пустом

[Alexey.Danilchenko]

Hey Alexey, relax

"I have never been so relaxed in all my life..." ((c) "How to steal a million")

Just to say that I am never mad - but it is unfortunate when merits of several approaches are discussed without fully understanding what is involved in both.

I am ok with dcamprof and understand your target audience and reasons for new tool so don't worry about me . The tool I am writing for spectral sensitivity curves automation may have a component to build simple profiles for given conditions (probably using external tools like ArgyllCMS or dcamprof) - but it is a bit early to say.

[scyth]

when merits of several approaches are discussed without fully understanding what is involved in both.

a lot of useful discussions start exactly like this

[Alexey.Danilchenko]

a lot of useful discussions start exactly like this

To me discussions start with questions.

[Alexey.Danilchenko]

Yeah I'll have a look. Building libarchive myself should not too much of a mess, I'm more worried that it's the same problem with the OpenMP lib (also brew). I have like a mountain of things to do to the next patch release. Or actually not too many things, but I'm a bit stressed by all input coming at the same time, but then again I put myself into this situation so I guess I get what I deserve...

Not sure if it helps but for cross compilation on Mac I opted for getting the llvm with OpenMP support (from llvm site) and converting it to toolchain. Then I used XCode for building project and added omp dlls to the path in execurable folder with name of that folder in xml resource. That works just fine and has no dependencies on external non standard libraries.