Linear Workflow Action - Need Beta Testers

[DMahoney]

Supposedly, this is the bible:
http://www.colorremedies.com/realworldcolor/

Hmm, I'm not a very religous person, is that going to be a problem with this?!

Looks like it might be a good book. I'll have to head down to my local Chapters and see if they have this in stock. It might answer some questions I have about proper, colour-managed workflow. I THINK I have it down well enough. I seem to get decent, predictable results, and I have read TONS or articles and web pages on it, but I am not really 100% SURE I got it yet. Of course, I still have to get that book on Understanding Exposure that everyone keeps recommending. All that has held me back so far is that supposedly there is an updated version that is going to be available in August, which has some sections pertaining to digital cameras. I don't know if this will offer any more "enlightenment" or not, but if I can hold off for a couple more months, I can take a look at both and see what suits my needs the best.

Don

 

[pbleic]

I posted some comparisons in my blog today of EVU vs. Linear EVU vs. DCRAW. The photos are WAY too big to post here, but I will post the text. If you go to the blog you can see all the photos together, rather than going to them one at a time.

http://pbleic.blogspot.com
--
Paul

------------------------------------------------
Pbase supporter
Photographs at: http://www.pbase.com/pbleic
--------------------------------------------------

Unless specified otherwise, all images are Copyright 2003, 2004 All rights reserved.

 

[Victor Engel]

Interesting, but I don't think the last image is superior. There is some significant posterization in the red canoe. Look at the area close to the water. There seem to be bands of varying hue. Could this be a problem with DCRAW near clipped values? Is DCRAW not the software that has problems with sunsets for example? Note that I'm asking because I don't know. I've never used it, so I haven't been following it closely.

I posted some comparisons in my blog today of EVU vs. Linear EVU
vs. DCRAW. The photos are WAY too big to post here, but I will post
the text. If you go to the blog you can see all the photos
together, rather than going to them one at a time.

http://pbleic.blogspot.com
--
Paul

------------------------------------------------
Pbase supporter
Photographs at: http://www.pbase.com/pbleic
--------------------------------------------------
Unless specified otherwise, all images are Copyright 2003, 2004
All rights reserved.

 

[pbleic]

Dcraw is the middle, not the last image in both comparisons.

There does seem to be some bands - this may be an issue with the profile I have been using, I think. I need to test some more.

C1 is the program with the bad sunsets.

Paul

I posted some comparisons in my blog today of EVU vs. Linear EVU
vs. DCRAW. The photos are WAY too big to post here, but I will post
the text. If you go to the blog you can see all the photos
together, rather than going to them one at a time.

http://pbleic.blogspot.com
--
Paul

------------------------------------------------
Pbase supporter
Photographs at: http://www.pbase.com/pbleic
--------------------------------------------------
Unless specified otherwise, all images are Copyright 2003, 2004
All rights reserved.

--
Paul

------------------------------------------------
Pbase supporter
Photographs at: http://www.pbase.com/pbleic
--------------------------------------------------

Unless specified otherwise, all images are Copyright 2003, 2004 All rights reserved.

 

[pbleic]

Turns out I was using Microsoft ICM for conversion, NOT Adobe Conversion engine. Redid the conversion of the DCRAW and the pixelation and linear orange streaks are gone. STUPID.





--
Paul

------------------------------------------------
Pbase supporter
Photographs at: http://www.pbase.com/pbleic
--------------------------------------------------

Unless specified otherwise, all images are Copyright 2003, 2004 All rights reserved.

 

[Al Pacheco]

I was just looking for more info about "your" profile! Is it

intended to be applied to dcraw linear files, or FVU "linear"
files? I notice that the colours don't match what I get from FVU,
is it intentionally off for some reason?

Here's some info on that profile.

When I learned last year that Canon had released a new SDK that had fixed their 10D linear conversion I started to experiment a bit as I had a need for some profiled data. So I used BreezeBrowser with the newer SDK library (FVU wasn't updated as quickly) to do the linear conversions for my targets. That particular profile was made by exposing the target in such a manner as to push the brightest patch on the target up to a level that would be optimized in Photoshop. The reason is that I used Timo Autiokari's XLProfiler to build the profile and it uses a Photoshop plugin to read the target's patches. So dealing with the scaling of the data is important. This step requires experimenting to see how the profiler deals with the data. Could it be made better? You bet...depends upon how much time you want to spend on it.

It is still darker because that is how the profiler renders it. It still requires a levels and/or a curves adjustment to bring it up to par with what you have come to expect from "normal" conversions. And it might need some Hue/Saturation adjustments to bring some colors into line. If you get to that point I can demonstrate a nifty method using a Macbeth chart.

Will it work with DCRaw? Perhaps. But DCRAW is a console application and not very intuitive. I did try extracting the code that was specific to the 10D from DCRAW source code and built a routine I could call from another appplication but it just wasn't worth the hassle when Canon's routines performed things like noise reduction much better. I didn't feel like writing stuff like that.

You'll have to experiment a bit to get the proper value for the "b" (brightness) switch. You want to mimic the same conversion as FVU. Try starting at 3.75 and work around that. It is set to 1.0 by default.
I'd leave the red and blue multipliers to their defaults (1.0) at first.

The author of DCRAW, Dave Coffin, did an absolutely astounding job of reverse engineering the raw format, but he admits that he has made a "risky assumption" in decoding the bayer patterns.

There are a few web sites that draw comparisons between DCRAW and the older "broken" Canon conversion. IMO, these are simply bogus and not worth a second look. Francisco Montilla did such a bogus test and has never upgraded his site... http://www.insflug.org/raw/analysis/dcrawvsfvu/conclusion.php3
Completely false information.

To paraphrase Magne Nilsen from a post above, the DCRAW code decompiles the internal formats to get to the raw non-debayered pixels, but it uses very simplistic attempts of debayering. DCRAW was primarily aimed at decompiling the RAW format. It's built-in debayering methods are included more as an example than anything that has ANY commercial or real life merit. Canons de-bayering, as implemented in FVU and EVU is IMO technically the best there is - so far.
Believe me, Magne's not wet behind the ears when it comes to this stuff.

But when I took the sRGB file and converted it
to the AIM WideGamut linear profile, I was able to use higher USM
settings then doing the USM in sRGB without converting to linear.
to linear and sharpened in linear. Therefore, unless I am reading
it wrong, it seems that being in linear allows for more aggressive
USM before seeing halos. No?

I haven't any experience with the AIM WideGamut colorspace. It is a colorspace, right? Not a camera profile. The linear file should handle greater doses of USM than a non-linear. But I believe you have to watch the noise issues. And do you actually gain that much? I guess it'll be up to you or pbleic (Paul?) to show that.

At about the same time I was working on that profile I discovered that I could use the same methods to profile C1 and started exploring that avenue with some success. And then Magne released his work and saved me many trips to the Advil bottle.

 

[Al Pacheco]

Interesting info concerning the Microsoft ICM vs. the Adobe Conversion engine.

There does seem to be some bands - this may be an issue with the
profile I have been using, I think. I need to test some more.

C1 is the program with the bad sunsets.

Paul

I posted some comparisons in my blog today of EVU vs. Linear EVU
vs. DCRAW. The photos are WAY too big to post here, but I will post
the text. If you go to the blog you can see all the photos
together, rather than going to them one at a time.

http://pbleic.blogspot.com
--
Paul

------------------------------------------------
Pbase supporter
Photographs at: http://www.pbase.com/pbleic
--------------------------------------------------
Unless specified otherwise, all images are Copyright 2003, 2004
All rights reserved.

--
Paul

------------------------------------------------
Pbase supporter
Photographs at: http://www.pbase.com/pbleic
--------------------------------------------------
Unless specified otherwise, all images are Copyright 2003, 2004
All rights reserved.

 

[pbleic]

Not a bad result, though once I got that right.

Paul

There does seem to be some bands - this may be an issue with the
profile I have been using, I think. I need to test some more.

C1 is the program with the bad sunsets.

Paul

I posted some comparisons in my blog today of EVU vs. Linear EVU
vs. DCRAW. The photos are WAY too big to post here, but I will post
the text. If you go to the blog you can see all the photos
together, rather than going to them one at a time.

http://pbleic.blogspot.com
--
Paul

------------------------------------------------
Pbase supporter
Photographs at: http://www.pbase.com/pbleic
--------------------------------------------------
Unless specified otherwise, all images are Copyright 2003, 2004
All rights reserved.

--
Paul

------------------------------------------------
Pbase supporter
Photographs at: http://www.pbase.com/pbleic
--------------------------------------------------
Unless specified otherwise, all images are Copyright 2003, 2004
All rights reserved.

--
Paul

------------------------------------------------
Pbase supporter
Photographs at: http://www.pbase.com/pbleic
--------------------------------------------------

Unless specified otherwise, all images are Copyright 2003, 2004 All rights reserved.

 

[Mike.]

Having used the new EVU and compared it with capture 1 dslr (the old version) I still find it doesnt offers anything I cant do in C1 - the workflow is faster in C1, colour profiles are there etc. C1 does some odd processing of edges most noticeable red edges also for some photos ive noticed pitting.

Anyhow I spent far too long with xlprofiler Q60 targets in an attempt to get the dcraw doing "magical things" - in the end I took the path of least resistance and brought c1.

-Mike.

Really? The jury is stil out on this one, but C1 seems just
slightly better than CS, which is notably better than BB, which (
other than combined conversioons ) is on a par with Canon's D60
conversion software.

Before coming to this conclusion, I suggest you check out the new
EVU. It is getting some pretty good press in the Rob Galbraith
Professional Photography forum. In fact, consensus is that it beats
C1, PSCS and everything else in terms of color and quality.

Really?
consensus: The making of a decision by general agreement
and in the absence of any voiced objection.

• IMHO EVU colors are NOT beating C1, PSCS and everything else.
• IMHO EVU quality is NOT beating C1, PSCS and everything else.

So - there you have it - at least ONE voiced objection, and the
consensus is gone While the EVU is OK - it still has a long way
to go - at least for me...

~~~~
Magne

--
-Mike

 

[pbleic]

After a lot of time spent on this, I have reached some conclusions:

Dcraw is clearly sharper than any other algorithm - it actually has LESS zig zaggy lines in most places. However, it also develops some nice zipper and moire artifacts - apparent in the color of the rope in my example - which should be B&W.

EVU, FVU, and Breeze Browser are all the same. They do have less to no moire or zipper artifacts, though actually a bit more zig zags in many places. They just can't be sharpened as well as dcraw, but they are close.

Magne Nielsen and Al Pacheco, two folks I respect a great deal, tell me in the 10D forum thread that Dcraw is a flawed algorithm. Here is the thread: http://forums.dpreview.com/forums/read.asp?forum=1019&message=8985262

I am giving up on dcraw because of the moire - I couldn't live with it in my photos. However, I DO believe that linear processing is the way to go. The EVU/FVU colors are off; the contrast range is greater, and there is more control.

Good news - I have a pretty easy workflow that I have simplified down to the bare essentials. I am going to make a challenge in another thread.

Paul

--
Paul

------------------------------------------------
Pbase supporter
Photographs at: http://www.pbase.com/pbleic
--------------------------------------------------

Unless specified otherwise, all images are Copyright 2003, 2004 All rights reserved.

 

[FYap]

Paul,

I had the same problems in dcraw output as you did. It still have some merits in my book, mainly for highlight recovery which it somehow manages slightly better than EVU/FVU.

To me at least, the current color profiles still can be improved upon. While in your boat case, they shine, in other cases, EVU/FVU output have better color fidelity IMO.

Beyond that, linear conversion has revealed to me that the normal Canon tone curve really squashes the shadow area (I think to supress noise), and the linear conversions are much gentler here. For some type of photos, that really is an important consideration.

Francis

After a lot of time spent on this, I have reached some conclusions:

Dcraw is clearly sharper than any other algorithm - it actually has
LESS zig zaggy lines in most places. However, it also develops
some nice zipper and moire artifacts - apparent in the color of the
rope in my example - which should be B&W.

EVU, FVU, and Breeze Browser are all the same. They do have less to
no moire or zipper artifacts, though actually a bit more zig zags
in many places. They just can't be sharpened as well as dcraw, but
they are close.

Magne Nielsen and Al Pacheco, two folks I respect a great deal,
tell me in the 10D forum thread that Dcraw is a flawed algorithm.
Here is the thread:
http://forums.dpreview.com/forums/read.asp?forum=1019&message=8985262

I am giving up on dcraw because of the moire - I couldn't live with
it in my photos. However, I DO believe that linear processing is
the way to go. The EVU/FVU colors are off; the contrast range is
greater, and there is more control.

Good news - I have a pretty easy workflow that I have simplified
down to the bare essentials. I am going to make a challenge in
another thread.

Paul

--
Paul

------------------------------------------------
Pbase supporter
Photographs at: http://www.pbase.com/pbleic
--------------------------------------------------
Unless specified otherwise, all images are Copyright 2003, 2004
All rights reserved.

 

[Al Pacheco]

Like I said before, Paul, you are demonstrating a healthy curiosity and desire for knowledge. Those who stick with vanilla all of the time will never know the other flavors. You will eventually establish methods that will be the best for your work. And sharing of information is what this forum is all about. Expect some hot debates now and then and perhaps sometimes a lack of interest. But at least YOU'LL be defining the process and that's what counts.

After a lot of time spent on this, I have reached some conclusions:

Dcraw is clearly sharper than any other algorithm - it actually has
LESS zig zaggy lines in most places. However, it also develops
some nice zipper and moire artifacts - apparent in the color of the
rope in my example - which should be B&W.

EVU, FVU, and Breeze Browser are all the same. They do have less to
no moire or zipper artifacts, though actually a bit more zig zags
in many places. They just can't be sharpened as well as dcraw, but
they are close.

Magne Nielsen and Al Pacheco, two folks I respect a great deal,
tell me in the 10D forum thread that Dcraw is a flawed algorithm.
Here is the thread:
http://forums.dpreview.com/forums/read.asp?forum=1019&message=8985262

I am giving up on dcraw because of the moire - I couldn't live with
it in my photos. However, I DO believe that linear processing is
the way to go. The EVU/FVU colors are off; the contrast range is
greater, and there is more control.

Good news - I have a pretty easy workflow that I have simplified
down to the bare essentials. I am going to make a challenge in
another thread.

Paul

--
Paul

------------------------------------------------
Pbase supporter
Photographs at: http://www.pbase.com/pbleic
--------------------------------------------------
Unless specified otherwise, all images are Copyright 2003, 2004
All rights reserved.

 

[pbleic]

http://forums.dpreview.com/forums/read.asp?forum=1031&message=9029867

 

[DMahoney]

Here's some info on that profile.
When I learned last year that Canon had released a new SDK that had
fixed their 10D linear conversion I started to experiment a bit as
I had a need for some profiled data. So I used BreezeBrowser with
the newer SDK library (FVU wasn't updated as quickly) to do the
linear conversions for my targets. That particular profile was made
by exposing the target in such a manner as to push the brightest
patch on the target up to a level that would be optimized in
Photoshop. The reason is that I used Timo Autiokari's XLProfiler to
build the profile and it uses a Photoshop plugin to read the
target's patches. So dealing with the scaling of the data is
important. This step requires experimenting to see how the
profiler deals with the data. Could it be made better? You
bet...depends upon how much time you want to spend on it.
It is still darker because that is how the profiler renders it. It
still requires a levels and/or a curves adjustment to bring it up
to par with what you have come to expect from "normal" conversions.
And it might need some Hue/Saturation adjustments to bring some
colors into line. If you get to that point I can demonstrate a
nifty method using a Macbeth chart.

Thanks for digging out the info on the file... I have been getting better results with it lately. Is the intent of the profile to match what you get from FVU? Or to try to be more accurate? It seems like a lot of work twiddling with brightening up the image, adjusting saturation and stuff, just to bring it in line with what you would get from FVU. So where is the advantage? Or does that lie in my second suggestion, which was trying to be more accurate?
And I really need to get a Macbeth chart one of these days.

Will it work with DCRaw? Perhaps. But DCRAW is a console
application and not very intuitive.

I still spend a lot of time in consoles, so that is no big deal. And I have it hooked into my directories so that I right click on a dir. full of raws and click on Process, and it goes and does the whole shot at once. Two clicks. It's easier and quicker than using ANY other raw conversion software with this method. Of course, you have no control over the image, and you are stuck with whatever parameters you have set in a text file somewhere. But it's fast. Shame the quality isn't quite there. I've been noticing some moire lately (that I think Paul has commented on too now), and some other things I don't like, but it IS easy to use!

I did try extracting the code
that was specific to the 10D from DCRAW source code and built a
routine I could call from another appplication but it just wasn't
worth the hassle when Canon's routines performed things like noise
reduction much better. I didn't feel like writing stuff like that.

Ya, no point re-writing the wheel I guess!

You'll have to experiment a bit to get the proper value for the "b"
(brightness) switch. You want to mimic the same conversion as FVU.
Try starting at 3.75 and work around that. It is set to 1.0 by
default.
I'd leave the red and blue multipliers to their defaults (1.0) at
first.

Hmm, I'll have to play with it then. I haven't set it that high. I think I am using what I say recommended, which is 3. I'll have to try higher.

To paraphrase Magne Nilsen from a post above, the DCRAW code
decompiles the internal formats to get to the raw non-debayered
pixels, but it uses very simplistic attempts of debayering. DCRAW
was primarily aimed at decompiling the RAW format. It's built-in
debayering methods are included more as an example than anything
that has ANY commercial or real life merit. Canons de-bayering, as
implemented in FVU and EVU is IMO technically the best there is -
so far.
Believe me, Magne's not wet behind the ears when it comes to this
stuff.

I haven't had any real complaints with Canons debayering, that's for sure. I need to look closer at my buddys D100 files that he gets out of the Nikon software, see what that is like! How does C1 compare? Most people seem to be getting good results with that!

I haven't any experience with the AIM WideGamut colorspace. It is a
colorspace, right? Not a camera profile. The linear file should
handle greater doses of USM than a non-linear. But I believe you
have to watch the noise issues. And do you actually gain that much?
I guess it'll be up to you or pbleic (Paul?) to show that.

Yes, AIM WideGamut is a linear colourspace. Or supposed to be anyways. If I understood it correctly! I agree with you on watching the noise issues. And the Threshold value in USM certainly makes more of a difference. In a "normal" file I might use a threshold of 10 to keep noise the sky down. In linear that same threshold leaves a LOT more unsharpened sometimes. Which makes sense once I thought about it.But I haven't had too much of a problem with noise issues. At least not on the pics I have been playing with. As for how much I gain by sharpening in USM, well, starting with a linear file and sharpening in linear, I was able to use much higher settings before I started seeing halos. But even converting sRGB to linear I was able to get less halos with the same settings. This is what I tried to show in my .psd file I posted. Those are extreme examples to try to really show off differences, so in normal useage the differences will be less. But a file that starts out linear does seem to have a nice advantage in sharpening. Too bad it has many DISadvantages too!

At about the same time I was working on that profile I discovered
that I could use the same methods to profile C1 and started
exploring that avenue with some success. And then Magne released
his work and saved me many trips to the Advil bottle.

Not familiar with Magne's work. Mind filling me in? Know the name, don't know the reuptation!

Don

 

[DMahoney]

Like I said before, Paul, you are demonstrating a healthy curiosity
and desire for knowledge. Those who stick with vanilla all of the
time will never know the other flavors. You will eventually
establish methods that will be the best for your work. And sharing
of information is what this forum is all about. Expect some hot
debates now and then and perhaps sometimes a lack of interest. But
at least YOU'LL be defining the process and that's what counts.

EXACTLY why I have been following Paul on his journey. I don't expect to switch over to this process, and if I had something I really wanted to print out, I would use my normal methods right now, because they get good enough results. But if we never try new things, we never learn new things. We might not use the process, but we will probably come away with better insight and better understanding of what we are trying to achieve. In the end, as long as the picture looks good, it doesn't matter what process was used to get it there.

Don

 

[Al Pacheco]

Don wrote... Is the intent of the profile to match what you get from > FVU? Or to try to be more accurate?

I was aiming to satisfy a number of goals, accuracy being somewhat tempered by the limitations of the profiling method itself and of my own knowledge (I'm not a color scientist!). For the 10D, reds look a tad orange (green channel is off a bit) and there were some other issues I wanted to address.

...seems like a lot of work twiddling with brightening up the image,
adjusting saturation and stuff, just to bring it in line with what
you would get from FVU. So where is the advantage? Or does that lie
in my second suggestion, which was trying to be more accurate?

Absolute color accuracy is difficult to attain and probably not required for most images. I decided to get "in the ballpark" with the profile and then do a levels/curve adjustment and finish up with Hue/Saturation to fine tune the colors.

Here's a "sample" of a curve to be applied to the linear file after assigning the profile and converting to Adobe RGB. No levels adjustment is needed. There is also a Hue/Sat adjustment in the zip which was a "preliminary" test I did to correct a few colors.
http://home.comcast.net/~ajpacheco/CurveHueSat.zip

Not familiar with Magne's work. Mind filling me in? Know the name,
don't know the reuptation!

Magne is a long time member on this forum who has been working with color theory for many, many years. He has a very keen understanding of the way profiles function and interact with image data. His normal operating mode is governed by neither myth nor guesswork; instead he has leaned over backwards to assist myself and others trying to get a handle on this stuff. At the moment he is concentrating on Capture One profiling as Capture One REQUIRES a good profile to deliver decent results.
He recently opened a new web site....
http://etcetera.cc/pub/

 

[DMahoney]

I was aiming to satisfy a number of goals, accuracy being somewhat
tempered by the limitations of the profiling method itself and of
my own knowledge (I'm not a color scientist!). For the 10D, reds
look a tad orange (green channel is off a bit) and there were some
other issues I wanted to address.

I take it you never quite met those goals?

Absolute color accuracy is difficult to attain and probably not
required for most images. I decided to get "in the ballpark" with
the profile and then do a levels/curve adjustment and finish up
with Hue/Saturation to fine tune the colors.

Absolute colour accuracy would probably be nice sometimes, but a lot of times I find myself prefering a slightly warmer white balance, for example. It might be just a "phase" I am going through right now, but I try and make a pic look as close to how I remember it, and then end up trying to make it look "better". Sometimes even with success! But like you say, if you can get in the ballpark, and be "close enough" it goes a long way to cutting down your time and making a good pic.

Here's a "sample" of a curve to be applied to the linear file after
assigning the profile and converting to Adobe RGB. No levels
adjustment is needed. There is also a Hue/Sat adjustment in the zip
which was a "preliminary" test I did to correct a few colors.
http://home.comcast.net/~ajpacheco/CurveHueSat.zip

Paul has already mentioned this and I tried it, but I didn't realize that I was supposed to convert to Adobe RGB FIRST, and then running them. Works pretty good. So I am curious. After doing all this work, and getting decent results, it seems like you ended up not using this method. How come? What did you eventually settle on?

Magne is a long time member on this forum who has been working with
color theory for many, many years. He has a very keen
understanding of the way profiles function and interact with image
data. His normal operating mode is governed by neither myth nor
guesswork;

He's NOT governed by myth or guesswork? NOOO!! And here I thought I was doing pretty good... you mean someone can actually get a HANDLE on all this stuff?!

instead he has leaned over backwards to assist myself
and others trying to get a handle on this stuff. At the moment he
is concentrating on Capture One profiling as Capture One REQUIRES a
good profile to deliver decent results.
He recently opened a new web site....
http://etcetera.cc/pub/

Great, even MORE money to spend on software! DOES IT EVER END???!!!! Heh, seriously though, I take it that it makes a good difference? I don't see any examples on his site, so I don't know what that $30 (american?) would do for me.

Don

 

[Al Pacheco]

Works pretty good. So I am curious. After doing all
this work, and getting decent results, it seems like you ended up
not using this method. How come? What did you eventually settle on?

I did attain some fair degree of success towards meeting some of my goals.

When I first started investigating the process I thought it would be simple and painless. No such luck.

My main objective was to be as aware as possible as to what adjustments would be needed when I finally opened my images in my editor. I did not want to just stumble around. My intention was to identify what I considered problem areas and see if a good profile could get me off to a good start in fixing them.

But!!! I had to accept the fact that no matter how good a conversion I could attain using the linear format, I would NOT have a decent workflow.

No way to preview the images...no way to judge the effects of editing until I loaded the images into Photoshop. And when I was away on business I needed an easier and quicker method.

So I started using Capture One because of the instant "color-managed" feedback you get.

Capture One absolutely requires a decent profile and when I started experimenting I was pleasantly surprised as to the increased accuracy I could get with my own custom profile using some of the same methods I investigated with the linear work. So now I don't need a levels/curves adjustment in Photoshop or even a Hue/Sat adjustment. I have some specific C1 curves I use and several good profiles and I really doubt you could see any drastic improvement by using linear.

Is C1 perfect? Far from it. But it is improving and it has something that is unique these days...Phase One is listening to the user base. Can't say the same for Canon.

I'll be away for a week or so. Good luck with your efforts.

 

[Al Pacheco]

One last thing....(still reading through my notes!).

I mentioned that I was seeing more and more problems with the Canon conversion. Some of these issues are sadly inherited by the linear conversion. With the 10D, at least, there is a green channel "weirdness" that can show up in odd places. It has to do with the white balance routines. Capture One is extremely clean in this regard. It's been widely reported.

Here's an example:

A) I tool an image an converted via FVU using normal contrast, normal saturation, Adobe RGB,... everything set to "normal" except sharpening set to the lowest point so I could add that in Photoshop.

B) I then converted using FVU linear mode, 1) applied my profile, 2) converted to Adobe RGB, 3) applied my curve and then 4) applied the Hue/Sat adjustments as I mentioned above.

C) Lastly, I did a conversion in Capture One LE with my custom profile with Adobe RGB as the target space.

I sharpened all the same...USM =500, 0.2, 2. I did this three times.

I resized & tiled the images (in sRGB) ...so you can see that they all seem similar. And the histograms do show that the three images are extremely similar.

(This proves that the linear file is brought up to the level of the non-linear file fairly well by my profile/curves adjustments.)



OK, I next cropped out a small portion at 100% to show that the linear is surely capable of bringing out some more detail than the non-linear. Compare the window at the far right.

At this point, I think Capture One does just about as well as the linear file. That's my opinion, of course, and the linear file will hold up better with further sharpening while the C1 file will start to fall apart. This I've tested and believe firmly.

But, but, but.... more importantly, Capture One doesn't have the little green/blue "uglies" that you can see in the Canon conversions. Look at the clapboards right above the center window in the second image. Is this unique to the 10D? I don't know. It surely doesn't show in D30, D60, or 1D files I've seen.

Does the 300D show the same thing? Don't know.

Is it important? Only you can decide.

 

[Magne Nilsen]

Magne is a long time member on this forum who has been working with
color theory for many, many years. He has a very keen
understanding of the way profiles function and interact with image
data. His normal operating mode is governed by neither myth nor
guesswork;

He's NOT governed by myth or guesswork? NOOO!! And here I thought I
was doing pretty good... you mean someone can actually get a HANDLE
on all this stuff?!

Well, actually I am governed by myth and guesswork!. Truly. Yes! Please?

instead he has leaned over backwards to assist myself
and others trying to get a handle on this stuff. At the moment he
is concentrating on Capture One profiling as Capture One REQUIRES a
good profile to deliver decent results.
He recently opened a new web site....
http://etcetera.cc/pub/

Great, even MORE money to spend on software! DOES IT EVER END???!!!

Nope, never...

Heh, seriously though, I take it that it makes a good
difference? I don't see any examples on his site, so I don't know
what that $30 (american?) would do for me.

$30 US. If you select the "ETC profiles" section on the left menu, and then open the article about your camera, you will find samples for most profiles, several for some. The 300D and 10D are extremely identical, and in sRGB you can't even measure any difference between them.

~~~~
Magne