Ad
Ad
Ad
Pages: [1] 2 3 »   Bottom of Page
Print
Author Topic: RawTherapee 4.1 is out, with greatly improved medium format support.  (Read 6226 times)
torger
Sr. Member
****
Offline Offline

Posts: 1588


« on: May 22, 2014, 07:14:25 AM »
ReplyReply

The free and open-source raw converter RawTherapee has now been released in version 4.1.

I contribute myself to this project, and as I have a special interest in medium format I've contributed patches to support various MF cameras. Leaf MOS and Phase IIQ has good support, Hasselblad 3FR should be quite okay too but it has been much more difficult to get Hasselblad test files so it's not so much tested. Leica and Sinar DNGs should work of course, and various older MFD formats.

There's also native support for Adobe (Lightroom/Camera Raw) DCP profiles and Capture One ICC profiles, so if you own any of those programs you can use profiles from there if you want to.

There's now also a Mac version in addition to the usual Windows and Linux versions, but note it's built with a cross-platform GUI toolkit so it does not have a streamlined native look and feel.

I use RawTherapee a lot myself, for all my fine-art work as I like to have control. However I'm not going to lie to you and say that it's a super-smooth user-friendly software. We're doing stepwise improvements in user-friendliness, but currently I'd say that it's a quite high threshold to get into it. A key thing that makes it a bit difficult is that it's possible to develop a file in many different ways, with different color models. You can work in RGB space, in Lab space, and in CIECAM02 space, and you have many different types of tone curves to play with. Not exactly ideal to make a user-friendly software, but quite nice to have for an advanced user interested in color.

In addition to being a great raw converter once you learn it, it's also a good learning tool concerning how digital photography works. It doesn't try hard to simulate film behavior like the commercial converters, meaning that the "Neutral" profile is really neutral, and +1 on the exposure slider really is +1.

If you use a technical camera like me and need color cast correction you can do this in RawTherapee, it's called "flatfield correction". Normally this is done on a blurred LCC shot, but if you need to kill dust spots reduce "Blur Radius" to 4 (default is 32), and to kill potential tile lines reduce to 0 (or experiment with vertical/horizontal filtering). If you have a tech wide angle shot  plagued with crosstalk so you get demosaicing artifacts, switch to the VNG4 demosaicer. The default demosaicer (Amaze) is very good at extracting detail, one of the best around, with the disadvantage that it can become unstable in crosstalk situations, which can be seen on some tech wides and sensor combinations.

If you work with Capture One ICC profiles note that they expect that there is a RGB curve applied with lots of highlight compression. You can provide that by applying a standard tonecurve manually.

The default profile applied when opening a file first time applies auto-tone and various fixups, and according to me it's a disaster, I never use it. Hopefully we'll make a better default to the next version, I'll fight for it Smiley. Anyway, today I start off with the neutral profile and work from there, and I'd recommend you to do so as well.

http://www.rawtherapee.com/
« Last Edit: May 22, 2014, 07:38:41 AM by torger » Logged
torger
Sr. Member
****
Offline Offline

Posts: 1588


« Reply #1 on: May 22, 2014, 07:38:03 AM »
ReplyReply

Here's a quick workflow for a MF tech cam picture.

1. Open the main file and apply the Neutral profile. It will look kind of dull and flat.
2. Goto the color tab and apply an input profile if you have one, either Lightroom DCP or Capture One ICC. Otherwise use "camera standard" which is matrix only, which in many cases are good, especially for manually processed landscape photography
3. Goto the raw tab, and enable hot/dead pixel filter, and apply flatfield correction if you have an LCC shot, set blur radius to 4 to kill dust, 0 to kill tile lines, or experiment with vertical or vertical+horizontal filtering. Best all-around setting for suppressing most things without introducing much additional noise is blur radius 4 and vertical+horizontal filtering.
4. Go back to the exposure tab and apply a tone curve to your liking. Start with "standard" (RGB, Capture One style) or "film curve" (Adobe style) to put in a contrast and highlight compression you like, and use exposure slider if you need exposure correction.
5. Now you should have a starting point, start playing with all sliders and checkboxes you can find.
Logged
jerome_m
Sr. Member
****
Online Online

Posts: 622


« Reply #2 on: May 22, 2014, 02:05:38 PM »
ReplyReply

Hasselblad 3FR should be quite okay too but it has been much more difficult to get Hasselblad test files so it's not so much tested.

I can send you 3FR files from an H3D-31 and from an H3DII-50 if you need them. Do you need a particular test picture or would anything do?
Logged
torger
Sr. Member
****
Offline Offline

Posts: 1588


« Reply #3 on: May 22, 2014, 02:23:47 PM »
ReplyReply

A base ISO file of anything real is good (so one can figure out if the color rendering is okay or not), a clipped highlight doesn't hurt (to see where clip levels are).

Usually the file can be read (as the general 3FR format has been reverse engineered), and what needs fixing is just things like get the proper sensor crop, clip levels and possibly a color matrix.

For IIQ we have some problems with high ISO files (maybe some different calibration data used then), but we don't prioritize high ISO that much as most MFD shooters shot at low ISOs. Thus having test files at low ISO is good enough for now.

The only test files I have now from Hassie cameras is for the CFV-50 and lots from CF-22 (as I've shot some with it myself), so yes it's surely welcome with more. My guess is that the 3FR format does not need much hacking to support various models, maybe nothing at all, but I don't know because I don't have the files to test with.

You can send me a private message. I won't publish the files (only keep in my personal archive for developing and testing raw format support). Any format support will be public though of course as it's an open source project. The IIQ patches to support calibration data I made has now recently become available in dcraw, which means that other small third-party software both commercial and open-source will get better support, and I hope the same will happen with any Hasselblad patches I make.

For the P+ and IQ series I've got files from most models, as well as Leaf Aptus and Credo. I haven't prioritized getting for Sinar as it's DNG and is supposed to work as DNG is an open documented format.
Logged
FranciscoDisilvestro
Sr. Member
****
Offline Offline

Posts: 554


WWW
« Reply #4 on: May 22, 2014, 05:53:30 PM »
ReplyReply

Is there a compiled Windows 64 bit version available? I can't find it in the download page

Regards
Logged

MichaelEzra
Sr. Member
****
Offline Offline

Posts: 659



WWW
« Reply #5 on: May 22, 2014, 10:37:00 PM »
ReplyReply

Win x64 build version 4.1.1 is now on the download page
Logged

torger
Sr. Member
****
Offline Offline

Posts: 1588


« Reply #6 on: May 24, 2014, 08:57:38 AM »
ReplyReply

I've now got some more Hasselblad test files from kind people on the forum, and the fff/3fr format handling turned out to need more work than I thought. It's generally about constant fixing (ie crop masks etc), so it takes 10 minutes per model to fix if I just have a file to test with. As many backs use the same sensor it's generally only needed to have a test file from one back with that sensor. Ie if I got H4D-60 working, H5D-60 should work too.

I'm working on a patch (not yet available) that hopefully will make files from all the 22, 31, 40, 50 and 60 megapixel Hasselblad backs load properly. 16 and 39 is still missing plus the special H5D-50c as I don't have any test files, and I'm a bit uncertain about 40 crop.

That is, I still need files for the following:

  • CFV or CFV II (aka CFV-16)
  • CF-39 or CFV-39 or H3D-39 or H3DII-39
  • H4D-40 or H5D-40
  • H5D-50c

So if anyone can provide a .fff or a .3FR for any of these models I would be most grateful (and so will the rest of the independent raw developer community that relies on open and shared format support efforts.)

If you don't have a drop-box or similar you can upload to my web server directly, just send me a private message and I give you the web address and username/password to upload.
« Last Edit: May 24, 2014, 10:16:52 AM by torger » Logged
gerald.d
Sr. Member
****
Offline Offline

Posts: 392


« Reply #7 on: May 24, 2014, 10:34:56 AM »
ReplyReply

torger - does the software work with Achromatic+ (P45+) files?

Kind regards,

Gerald.
Logged
torger
Sr. Member
****
Offline Offline

Posts: 1588


« Reply #8 on: May 24, 2014, 11:00:53 AM »
ReplyReply

torger - does the software work with Achromatic+ (P45+) files?

Oh, that's about the only Phase One back I don't have a test file for. I'd be glad to receive one Wink. I know you have one but I haven't asked before as we don't have a real monochrome rendering pipeline in there yet, so I cannot really show any true results until that is in place.

IQ260 Achromatic works anyway though, so probably P45+ will work too, but it's not really optimal since it makes a dummy demosaicing. For true monochrome support there's some tweaks required in the processing pipeline. It's mostly about disabling stuff so it's a lot simpler to fix than supporting say foveon so at some point we'll do it.
Logged
gerald.d
Sr. Member
****
Offline Offline

Posts: 392


« Reply #9 on: May 24, 2014, 04:55:57 PM »
ReplyReply

Sounds good.

I'm travelling at the moment. Let me know what you need and I'll get you files in early June.
Logged
Paul2660
Sr. Member
****
Offline Offline

Posts: 1919


WWW
« Reply #10 on: May 24, 2014, 05:32:38 PM »
ReplyReply

Torger

I have not downloaded the software yet. However can you elaborate a bit on the tech camera processing with LCC.  I am coming from a C1 background where you process the LCC the save it an apply it to the particular raw file. With this software do you process the LCC and save it?  How does the LCC get applied to the raw file?

Will be pulling down the code later this weekend

Thanks
Paul



Logged

Paul Caldwell
Little Rock, Arkansas U.S.
Photography > http://photosofarkansas.com
Blog> http://paulcaldwellphotography.com
torger
Sr. Member
****
Offline Offline

Posts: 1588


« Reply #11 on: May 25, 2014, 02:09:12 AM »
ReplyReply

Torger

I have not downloaded the software yet. However can you elaborate a bit on the tech camera processing with LCC.  I am coming from a C1 background where you process the LCC the save it an apply it to the particular raw file. With this software do you process the LCC and save it?  How does the LCC get applied to the raw file?

Will be pulling down the code later this weekend

With RawTherapee you don't process the LCC shot separately, in fact you don't even open the LCC shot (unless you want to look for it for some reason to investigate exposure for example). You just go to the raw processing tab, to the Flat-Field toolbox, click on the "File" button and select the LCC shot. You need to keep the LCC shot, ie you can't throw it away like with C1.

RawTherapee will then load and apply the LCC shot, according to the LCC settings. If you have a 6um Dalsa sensor so you can get microlens ripple and tiling you should change blur type to "horizontal+vertical" or reduce blur radius to zero.

The RT flat-field is basic, it's quite effective but has a few weaknesses:

  • If you have a clipped highlight in a area with low vignetting that highlight may become pink (compensate with raw scaling or pushing exposure)
  • If you have heavy microlens ripple that cannot be cleaned up as good as C1 does (because it takes more than just flatfield correction to do so due to crosstalk)

I have not contributed myself to the flat-field algorithm and I will not do that as it conflicts with my commercial software Lumariver HDR for which I'm in progress to develop the best LCC you can get Wink. RT's basic algorithm will however work well in most situations, and sometimes better than C1 as the C1's current algorithm seems a bit unstable concerning tiling suppression.

If you have an image with lots of crosstalk (say a shifted SK35 or SK28) the Amaze demosaicer will break and you will see mazing and poor color (due to green channel separation), and then you need to switch to the VNG4 demosaicer. The VNG4 handles crosstalk even better than C1's demosaicer does, but you lose some fine detail.

More documentation on the flat-field correction function exists at http://50.87.144.65/~rt/w/index.php?title=Flat_Field
Logged
kirkt
Full Member
***
Offline Offline

Posts: 185


« Reply #12 on: May 28, 2014, 03:11:36 PM »
ReplyReply

torger - have you been able to open LumaRiverHDR DNGFloat images in RT successfully?  On my 2011 MacBook Pro running Mavericks, trying to open a DNG Float exported from LumaRiverHDR crashes RT.  The build on the download page says that it is for OS X 10.7 but that it should work with 10.8 and 10.9.

Thanks,

kirk
Logged
torger
Sr. Member
****
Offline Offline

Posts: 1588


« Reply #13 on: May 29, 2014, 10:11:07 AM »
ReplyReply


Uhh haven't tested. I'll do it after the weekend, I'm travelling for the moment. I'm one of the minor contributors though, my time is spread over too many projects already so if it's hard to fix someone else's gotta do it.  Undecided

torger - have you been able to open LumaRiverHDR DNGFloat images in RT successfully?  On my 2011 MacBook Pro running Mavericks, trying to open a DNG Float exported from LumaRiverHDR crashes RT.  The build on the download page says that it is for OS X 10.7 but that it should work with 10.8 and 10.9.

Thanks,

kirk
Logged
ondebanks
Sr. Member
****
Offline Offline

Posts: 833



« Reply #14 on: May 29, 2014, 11:40:24 AM »
ReplyReply

Hi Torger,

RAWTherapee was never quite what I wanted before, but this new version has some of the things I had feature-requested:

(1) Option for no de-Bayering
(2) View RAW pixels in R/G/B alone (a fantastic diagnostic when used with (1) above!)
(3) More responsive rendering...and it no longer crashes when it has to make thumbnails of Kodak .DCR files!

It's still missing one important thing I feature-requested though!

(1) The "Hot/Dead pixel filter" under RAW (and the similar "Impulse Noise Reduction" filter under Detail) should be split into separate controls for dark/dead and bright/hot noise pixels.
There is hardly ever a natural (photographic subject) cause for single dark pixels - in my experience they are always due to oversubtraction of hot pixels in long exposures, or the occasional completely dead pixel. So this needs an aggressive control, to interpolate across them. Single bright pixels, OTOH, are not uncommon in photos from fat-pixel sensors without AA filters [MFDBs!] - they are a sign of sharp rendering of the photographic subject. But sometimes they are due to thermal noise. So this needs a separate, subtle control.
It should be possible to zap the dark pixels, while leaving the bright ones alone or hardly modifying them.

Overall, I've enjoyed trying out the new version. Needless to say, the flat-fielding and dark-frame subtraction features, which have been there for some time, are standout attractions too.

Ray


Logged
kirkt
Full Member
***
Offline Offline

Posts: 185


« Reply #15 on: May 29, 2014, 07:58:15 PM »
ReplyReply

No problem - I reported the bug on the RT forum.

kirk
Logged
ondebanks
Sr. Member
****
Offline Offline

Posts: 833



« Reply #16 on: May 30, 2014, 09:48:11 AM »
ReplyReply

In addition to being a great raw converter once you learn it, it's also a good learning tool concerning how digital photography works. It doesn't try hard to simulate film behavior like the commercial converters, meaning that the "Neutral" profile is really neutral, and +1 on the exposure slider really is +1.

Hi Torger,

Hmmm...I wonder if that is really the case. Here's a situation I just encountered with the new version, where I zeroed everything in RT, but I get very different results to when in zero everything in Dcraw:

First Dcraw - no de-bayering, no channel intensity scaling, 400% view:



Now RT - no de-bayering, settings applied for no channel intensity scaling - but it does still seem to be applying a strong scaling:



The image is a small section of an IR > 780nm shot with a Kodak CCD, where all 3 channels respond identically in quantum efficiency from 820nm longwards, and very closely as 820nm is approached. So this is almost the same response as an Achromatic back (pure mono sensor) in this wavelength regime. Therefore I think that de-bayering is not actually needed and a sharper image could be obtained without this interpolation. Which is why I turned it off [- which in turn is why I was so excited about the new RT being able to turn it off!].

All that should be needed, I hope, is a tiny bit of channel scaling so as to even out the intensity differences coming from the slightly difference responses below 820nm. Or denoising might even take care of that faint mottling in the Dcraw rendering.

But RT is applying some major scaling because the R, G and B pixels come out at strongly and systematically different intensities in the jpeg - you can clearly see the CFA grids - even though the RAW histograms showed the 3 channels as being almost identical.

Any idea what's wrong with RT here? It is not leaving RAW stay RAW, that's for sure.

Ray
« Last Edit: May 30, 2014, 09:50:39 AM by ondebanks » Logged
torger
Sr. Member
****
Offline Offline

Posts: 1588


« Reply #17 on: May 31, 2014, 02:53:07 AM »
ReplyReply

Hmmm...I wonder if that is really the case. Here's a situation I just encountered with the new version, where I zeroed everything in RT, but I get very different results to when in zero everything in Dcraw:

You saw through my claim Wink that neutral is totally neutral is not 100% true, there is some processing/scaling related to white balance. What I meant with neutral is neutral is that there is no curve applied, there's no sharpening or noise reduction etc, all those things that commercial programs usually apply.

I don't know exactly what the problem is but I do know that the b/w pipeline needs more work, there is no real support for B&W imagery, turning off demosaicing is only a initial step. My guess is that what you see is the white balance processing that's still active.

Few have B&W cameras so it has not been prioritized, but you could file a new issue or vote for this one I made a while ago:
http://code.google.com/p/rawtherapee/issues/detail?id=2017

Unfortunately I'm in a period now when I cannot contribute as much myself (contributed a lot fall 2013, but not much now), so I've focused my efforts on smaller 10-minutes-at-a-time things like adding support for another MF camera if I get a new test file etc (soon a new patch release will be coming with lots of new Hasselblad camera support). Like many open-source projects RT is a bit understaffed in periods, so if you do programming yourself any contributions are welcome. Reporting bugs and suggesting features are also valuable contributions of course.
« Last Edit: May 31, 2014, 03:02:28 AM by torger » Logged
ondebanks
Sr. Member
****
Offline Offline

Posts: 833



« Reply #18 on: June 19, 2014, 09:40:50 AM »
ReplyReply

You saw through my claim Wink that neutral is totally neutral is not 100% true, there is some processing/scaling related to white balance. What I meant with neutral is neutral is that there is no curve applied, there's no sharpening or noise reduction etc, all those things that commercial programs usually apply.

I don't know exactly what the problem is but I do know that the b/w pipeline needs more work, there is no real support for B&W imagery, turning off demosaicing is only a initial step. My guess is that what you see is the white balance processing that's still active.

Few have B&W cameras so it has not been prioritized, but you could file a new issue or vote for this one I made a while ago:
http://code.google.com/p/rawtherapee/issues/detail?id=2017

Unfortunately I'm in a period now when I cannot contribute as much myself (contributed a lot fall 2013, but not much now), so I've focused my efforts on smaller 10-minutes-at-a-time things like adding support for another MF camera if I get a new test file etc (soon a new patch release will be coming with lots of new Hasselblad camera support). Like many open-source projects RT is a bit understaffed in periods, so if you do programming yourself any contributions are welcome. Reporting bugs and suggesting features are also valuable contributions of course.

Thanks, I voted on that issue and supplied a link to a sample IR RAW file.

Yes, I think you're right that it's "the white balance processing that's still active".

Ray
Logged
torger
Sr. Member
****
Offline Offline

Posts: 1588


« Reply #19 on: June 19, 2014, 05:38:35 PM »
ReplyReply

Thanks, I voted on that issue and supplied a link to a sample IR RAW file.

Yes, I think you're right that it's "the white balance processing that's still active".

Thanks, got some time to do make a patch, seems to work. Thanks for providing the file. It's not committed yet but I expect to do during the weekend.

The patch also adds supports for other monochrome cameras like Leica Monochrome.
« Last Edit: June 19, 2014, 05:55:45 PM by torger » Logged
Pages: [1] 2 3 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad