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 5454 times)
torger
Sr. Member
****
Offline Offline

Posts: 1468


« Reply #40 on: July 25, 2014, 04:45:43 AM »
ReplyReply

I'm a little late to the news of the RT upgrade but does the latest version work with the old Mamiya ZD files? The previous vesrion opened them but they were all green and 'orrible.

The only way we can support a camera is either that dcraw has builtin support or someone sends us a file so we can tune support for it. I have got a lot of MFD test files, but so far none from ZD.

So likely it still doesn't work. However, if you send me a test file I can probably make it work. If the file opens it's generally only some minor things that needs to be tuned.
Logged
MichaelEzra
Sr. Member
****
Offline Offline

Posts: 655



WWW
« Reply #41 on: July 25, 2014, 08:45:25 AM »
ReplyReply

I work with zd files without any issues. You do need to use flat field correction to compensate for the lens color cast.
Logged

ppmax2
Newbie
*
Offline Offline

Posts: 40


« Reply #42 on: July 25, 2014, 09:41:04 AM »
ReplyReply

Quote
RT is more of an experiment box for interested image algorithm developers and digital photography interested users rather than a fast to learn efficient production tool.

Personally I find it to be a great tool for artistic low volume work and I do all my medium format landscape work in it (with some aid of other post-processing tools when needed), but I fire up Lightroom when I have 1500 images from the latest sports shoot to deal with.

We in the developer group know about the issue that RT allows to do things in multiple ways, RT is not at all as user-friendly as Lightroom. We do strive to make it more user-friendly and take one step at a time, but it will never be as stream-lined as say Lightroom. Developers come and go based on time and interest, each developer has typically interest in some smaller subset functionality and put an effort into that.

Simply put we don't have the resources and continuity to make it a streamlined application as user-friendly as the commercial apps.

I also find it quite interesting from an educational perspective to for example have four different curve types. Few actually know how a contrast curve affects color, but RT can demonstrate this really well. We have standard RGB curve (ie Capture One way), we have Adobe's film-like curve (close but not same to RGB), a more color-neutral weigthed curve and the theoretically neutral luminance curve (but will actually produce a perceptually desaturated result). Likewise we allow selecting different demosaicers, which in a streamlined user-friendly app would be an auto-choice of course.

So if you ask me I actually prefer if we don't make it too user-friendly but instead leave options for the user to experiment with various processing methods. User-friendly means cutting away all but one tool for making a specific task, reduce slider ranges to not include "crazy range", and make many settings automatic. While this would make it more competitive with Lightroom and C1, it would also rob the user of understanding of how digital photography works. RT is "rawer", while commercial tools try to immitate film behaviour.

That said RT can be used quite efficiently, with your own custom profiles and reducing the number of tools you use.

My own pet hate of RT is the default profile which enables auto-levels ("auto tone"), typically works well for high contrast landscape images but sucks for everything else (there's an ongoing debate of replacing it). You can replace that default profile with your own though, which is what I have done, with just a DCP and a curve.

As a Lightroom and C1 owner I enjoy the possibilty to use both Lightroom DCPs and C1 ICCs in RT. For my MF back I use a custom DCP, but when I casually process files from my compact I generally use ICCs from C1 (which I think has better color than Adobe). Note that C1 ICC are designed to be used with a standard (RGB) curve, and Lightroom DCPs with a film-curve if you want to replicate their looks.

Thanks for your comments Torger--I hope my comments did not appear too critical. I've only been using RT for about a month (many hours each night after work!) and I love it because it does things no other tool does, and in some areas does them better than any other tool can. I wouldn't trade all the features/benefits you mention above so that RT becomes a LR or C1 clone. FWIW I haven't used Aperture or C1 much recently...RT is pretty cool!

My interest in RT began upon hearing the news about Aperture. I bought a copy of C1 and downloaded RT within a week of hearing the news.

For the record, I definitely appreciate the opportunity to use a tool like RT, which (due to it's non-commercial status) doesn't need to be "a product" with all the baggage that that entails. Your description that RT is an "experiment" for image processing enthusiasts makes me appreciate it even more--it doesn't feel like one Wink

Regarding my comments about "gouging my eyes out" trying to duplicate in RT what I did in another tool...Now that I've been able to sleep on it I think my outburst can be distilled to this:
For some of my RAW 5D3 files RT seems to render a yellow/orange or orange/magenta cast that is very difficult for me to eliminate...it's not just color temp/tint... Strangely, this may be coupled with a general "flatness" or lack of contrast/punch between tones. Generally speaking, I always zero out the Neutral profile for the reasons you state above, and agree with your "pet hate" Wink

To tackle flatness I have been using both tone curves in the Exposure set, the Lab Lightness curve, Lab CH, and Lab CL curves, as well as Contrast by Level.

To tackle the color cast I have been using RGB curves, and the assorted Lab Chromaticity/Hue related curves.

As you can see, this particular image has me in futzing with just about every curve and curve type in RT Wink I have three UI suggestions that I believe will make a big difference:
1. Eliminate the Flat tone curve in all curve drop downs. Instead, make one of the various curve types the default (doesn't matter which) and make the curve flat. This will eliminate two mouse clicks anytime someone wants to work with a curve. If people want to save their curve settings they can save a history snapshot.
2. Consolidate RGB/HSV type adjustments into a single control, perhaps similar to C1's neat little color wheel or Aperture's Color adjustment block. These UI's enable someone to pick a color and then adjust hue, saturation, level, and range all in one spot. This does not mean I want to eliminate RGB curves...
3. RT, LR, C1, DarkTable...all these tools offer some form of a tabbed interface for organizing adjustments like exposure, color, sharpness/noise, etc. IMO these UI's are horrible because the user has no way of seeing all the adjustments that have been made to an image. In addition, while this categorical "workflow-driven" interface makes sense on paper (...get your exposure right, then do sharpen/noise...then do color...etc) this disciplined workflow paradigm breaks down in practice, and the user is constantly switching back and forth between different tabs to "get it right." If they haven't already, I highly suggest that the RT UI team take a look at how Aperture does this, which provides a single tab for all adjustments...which only need be exposed within the UI if they are actually being used. Note in the attached image all "adjustments" are centralized under one tab. Scrolling through a list of adjustments is far easier than clicking, scrolling, collapsing...etc.

Anyways--thanks for your contributions to the project and for compiling the latest for us Mac users; I appreciate it!

PP

Logged
Fine_Art
Sr. Member
****
Offline Offline

Posts: 1087


« Reply #43 on: July 25, 2014, 08:23:13 PM »
ReplyReply

"RT seems to render a yellow/orange or orange/magenta cast that is very difficult for me to eliminate...it's not just color temp/tint..."

Hi Paul,

You might want to try the lab a, b curves. One shifts yellow/blue, the other green/magenta.
Logged
ppmax2
Newbie
*
Offline Offline

Posts: 40


« Reply #44 on: July 27, 2014, 03:04:13 AM »
ReplyReply

Fine_Art--great suggestion, that did the trick. Those were the only curves I didn't touch. It took me a few tries and several hours, but I was able to recreate the look I achieved in C1 and Aperture...but the RT image is cleaner with better detail and less noise.

Thx
pp
Logged
Justinr
Sr. Member
****
Offline Offline

Posts: 1009


WWW
« Reply #45 on: July 27, 2014, 04:25:28 AM »
ReplyReply

I work with zd files without any issues. You do need to use flat field correction to compensate for the lens color cast.

When you say flat field correction I wonder if you could enlarge upon that as I'm not quite sure what it is.
Logged

torger
Sr. Member
****
Offline Offline

Posts: 1468


« Reply #46 on: July 27, 2014, 07:05:13 AM »
ReplyReply

When you say flat field correction I wonder if you could enlarge upon that as I'm not quite sure what it is.

Flatfield correction is the formal name of what is popularly called "LCC correction" in Capture One. In RT this is like said called flatfield correction, which is a better term as it just like C1's LCC corrects more than just color cast.

More elaborate docs you can find on the rawpedia: http://50.87.144.65/~rt/w/index.php?title=Flat_Field
Logged
BartvanderWolf
Sr. Member
****
Offline Offline

Posts: 3630


« Reply #47 on: July 27, 2014, 07:37:28 AM »
ReplyReply

Flatfield correction is the formal name of what is popularly called "LCC correction" in Capture One. In RT this is like said called flatfield correction, which is a better term as it just like C1's LCC corrects more than just color cast.

Hi,

The RT implementation works fine. The only improvement I could think of, is the selectable combined operation of both dust removal and flat-fielding (perhaps with an automatically adapting blur radius, or two passes which are combined) from the same file. One currently has to choose between either dust (small radius) or flat-field (larger radius), being able to use 2 sliders would also make it clearer that dust removal is a similar operation as Flat-fielding, it just requires different radii.

Cheers,
Bart
Logged
Fine_Art
Sr. Member
****
Offline Offline

Posts: 1087


« Reply #48 on: July 27, 2014, 10:45:24 AM »
ReplyReply

Flatfield correction is the formal name of what is popularly called "LCC correction" in Capture One. In RT this is like said called flatfield correction, which is a better term as it just like C1's LCC corrects more than just color cast.

More elaborate docs you can find on the rawpedia: http://50.87.144.65/~rt/w/index.php?title=Flat_Field

Thanks for the rawpedia link.

Is it possible to have the wavelet sharpen/soften (contrast by detail) have user set radiuses rather than 1,2,4,8,16 ?
Logged
torger
Sr. Member
****
Offline Offline

Posts: 1468


« Reply #49 on: July 28, 2014, 10:29:25 AM »
ReplyReply

Broader suggestions is better posted in the RT forum or as feature request in RT issue tracker on google code, I personally mainly make smaller contributions now and then, there are others that are better at responding to specific feature requests.

Your suggestions on FF improvements are great, but as I also work on commercial software which has flat-field correction (Lumariver HDR) there's a conflict of interest so you have to ask some other developer to improve RT's implementation :-). Personally I use RT's FF for the basic cases and Lumariver HDR's flat-field correction in more difficult cases (export to DNG from Lumariver HDR and process the FF-corrected raw file in RT).

RT's FF correction works alright for most cases but when you have issues like microlens ripple or crosstalk (mostly tech cam specific issues) it's not up for the task, then C1 does a better job and Lumariver HDR even better.
« Last Edit: July 28, 2014, 10:32:14 AM by torger » Logged
ppmax2
Newbie
*
Offline Offline

Posts: 40


« Reply #50 on: August 15, 2014, 10:51:59 AM »
ReplyReply

Hello Torger, if you are still around Wink

There was a fix committed recently for improper scaling when using the Lab a* b* curves...which apparently is available in 4.1.44 onwards. Any chance you would be interested in posting another build for OS X?

Also, are there any OS X specific build instructions? I'd be up for taking a whack at it, but the RawTherapee OS X build instructions page is devoid of any info.

thx!
pp
Logged
torger
Sr. Member
****
Offline Offline

Posts: 1468


« Reply #51 on: August 19, 2014, 09:19:43 AM »
ReplyReply

Hello Torger, if you are still around Wink

There was a fix committed recently for improper scaling when using the Lab a* b* curves...which apparently is available in 4.1.44 onwards. Any chance you would be interested in posting another build for OS X?

Also, are there any OS X specific build instructions? I'd be up for taking a whack at it, but the RawTherapee OS X build instructions page is devoid of any info.

It's a mess to build on OS X. I build on 10.7 currently as it's a lot less messier than 10.9. The problem with 10.9 is that you can't build with GCC in the same effortless way as in 10.7, the reason is that macports are built with the Clang compiler and links to libc++, while GCC links to libstdc++. The solution I came up with to build on 10.9 is to first build a custom Clang compiler with OpenMP support (which RT requires, but the standard Clang doesn't have it) and then build RT with this custom Clang instead of GCC.

However, instead of doing that I found that the current best solution is to build on 10.7 (those builds work on 10.9 too) and hopefully we can do that until Clang gets OpenMP integrated so one can build on a standard 10.9 setup.

Here's a build I just made: http://torger.dyndns.org/rt-bugs/RawTherapee_OSX_10.7_64_4.1.51.zip (edit changed from 50 to 51, added a P65+/P40+ bug fix I got reported during my vacation)

I've attached a draft compile instruction for 10.9 I did back in March. I have not built on 10.9 since then so I don't know how up to date it is today.
« Last Edit: August 19, 2014, 09:39:02 AM by torger » Logged
ppmax2
Newbie
*
Offline Offline

Posts: 40


« Reply #52 on: August 19, 2014, 10:33:11 PM »
ReplyReply

Hello Torger and thanks for your reply and the new build.

I'm on the road right now and won't have a chance to download or digest the build instructions you sent...but I will this weekend.

Thanks again and have a great day
PP
Logged
torger
Sr. Member
****
Offline Offline

Posts: 1468


« Reply #53 on: September 08, 2014, 07:33:32 AM »
ReplyReply

Leaf Credo 50 got released today, and through the Capture Integration raws at https://captureintegration.com/introducing-the-leaf-credo-50/ I added a basic color matrix support for it in RawTherapee, here's the new build (it's a quick and dirty release, haven't test run this OS X version yet as my main test box is Linux, but it should work):

http://torger.dyndns.org/rt-bugs/RawTherapee_OSX_10.7_64_4.1.68.zip

If you don't load any separate color profile (both standard and Adobe-style DCPs and Capture One-style ICCs are supported) the color matrix will be the same for Hasselblad H5D-50c, Phase One IQ250, Pentax 645z and now Leaf Credo 50 in this version of RawTherapee. I haven't been able to verify 100% sure through side-by-side shots that the raw colors indeed are exactly the same, but they are highly similar for sure. Even with exact same sensor and the digital readout CMOS provides I would expect at least some very slight measurable difference in raw colors, as IR filters in front of the sensor are probably not the same. Most will come down to your ICC or DCP profile though.
Logged
Pages: « 1 2 [3]   Top of Page
Print
Jump to:  

Ad
Ad
Ad