YARC ARF vs. Linear conversion

I have had my D60 for about 6 weeks now and have been experimenting. I have found that converting to non linear 16 bitt tiff with YARC's ARF yields superior results to converting to linear tiff and using either Fred's or Colin's linear actions. The YARC ARF conversions give deeper colors and a more pleasing image overall. This is in contrast to my experience with the D30 where I thought that linear conversion was superior (using Pekka's LinearSharpen).

Any comments appreciated.

BDF said:

I have had my D60 for about 6 weeks now and have been
experimenting. I have found that converting to non linear 16 bitt
tiff with YARC's ARF yields superior results to converting to
linear tiff and using either Fred's or Colin's linear actions. The
YARC ARF conversions give deeper colors and a more pleasing image
overall. This is in contrast to my experience with the D30 where I
thought that linear conversion was superior (using Pekka's
LinearSharpen).

Any comments appreciated.

Click to expand...

I also prefer YarcPlus's nonlinear conversion, with the artifact reduction as well. BreezeBrowser has a mode which combines aspects of linear and nonlinear but I haven't tried it yet. I do own both programs.

--
Walter K

I'm using the 1D currently (D30 previously) and have found ARF to be an excellent technique to getting nice, clean images. Personally I have never found linear conversions to provide any real benefit, so have just used non-linear with ARF, and adjust curves/levels/USM in Photoshop as needed.

Just to clarify what ARF does/doesn't do - it removes artifactual pixels. It will not do anything differently as far as color depth, balance, etc. It will convert, using Canon dll's, according to the conversion parameters selected. It just adds ARF on top of that to clean up the file.

What you're seeing in the color depth is probably just coming from working with the non-linear files to begin with, rather than having to go through a lot of Photoshop work to get the linear files anywhere near useable.

My $.02 on the matter.

Don
http://www.dlcphotography.net

...For instance, the method used AFTER converting from RAW to TIFF would also play into the final result. (The curves used, the degree of saturation, the converted profiles of the action used), would all have an effect on the final product.

As an example, when I convert a 16bit linear file using Breezebrowser, then apply Fred Miranda's LPBatch action, and compare it to a straight 16bit (non-linear) conversion, I usually prefer the overall "look" of the non-linear conversion. I believe that it has to do with the curves that Fred has incorporated into his action.

However, maybe your observation IS correct; it's just that I haven't heard anyone else mention this before.

BTW, the next release of YarcPlus is supposed to offer ARF for 16bit linear tiffs, so that will really offer us some choices, yes?

BDF said:

I have had my D60 for about 6 weeks now and have been
experimenting. I have found that converting to non linear 16 bitt
tiff with YARC's ARF yields superior results to converting to
linear tiff and using either Fred's or Colin's linear actions. The
YARC ARF conversions give deeper colors and a more pleasing image
overall. This is in contrast to my experience with the D30 where I
thought that linear conversion was superior (using Pekka's
LinearSharpen).

Any comments appreciated.

Click to expand...

--
Gary Shepard
Foreside PhotoGraphics
Maine, USA

Hi Don -

I do not own a 1D, so I don't know anything about its features, but I do have a D30 and a D60.

What 16bit linear tiff conversions do for my cameras, is allow my converted TIFFs to be free of any in-camera processing, the most important of which (to me), is sharpening. There is no provision in the D30/D60 for turning off sharpening completely, only minimizing it.

So, any conversion method for RAWS (except 16bit linear), introduces some amount of sharpening, contrast, and saturation into my images.

In the documentation for NIK Sharpener Pro, (my preferred method of sharpening), it warns that images will have unpredictable/unreliable results if they are sharpened twice. (regardless of the method of sharpening used). In tests which I have run, I have found this to be the case.

DonCohen said:

I'm using the 1D currently (D30 previously) and have found ARF to
be an excellent technique to getting nice, clean images.
Personally I have never found linear conversions to provide any
real benefit, so have just used non-linear with ARF, and adjust
curves/levels/USM in Photoshop as needed.

Just to clarify what ARF does/doesn't do - it removes artifactual
pixels. It will not do anything differently as far as color depth,
balance, etc. It will convert, using Canon dll's, according to the
conversion parameters selected. It just adds ARF on top of that to
clean up the file.

What you're seeing in the color depth is probably just coming from
working with the non-linear files to begin with, rather than having
to go through a lot of Photoshop work to get the linear files
anywhere near useable.

My $.02 on the matter.

Don
http://www.dlcphotography.net

Click to expand...

--
Gary Shepard
Foreside PhotoGraphics
Maine, USA

Hi Gary,

Thanks for the explanation of how/why you use Linear files. I certainly understand the logic, as I convert my Raw files (prior D30 and current 1D) using the lowest settings for contrast, sharpening, etc. to provide the least processed image to work with in Photoshop.

I have never done any controlled studies to more definitively answer this question, but I'm inclined to think that a good photographer, with good Photoshop skills, will have pretty much comparable results using either workflow method, in terms of brightness, contrast, color accuracy, etc.

My main interest in possibly using linear files would be to theoretically reclaim blown highlights in images I have that are otherwise good, but where I messed up the exposure a bit.

Do you have any examples that demonstrate true retention of detail in areas of blown highlights using linear vs non-linear conversion? I have played around with this some, and my impression (not a firm conclusion) is that the linear gave me gray areas, rather than white areas, but otherwise held no other true image detail. But I do remain open-minded about this, which is why I ask if you have any reasonably controlled examples that demonstrate this.

Thanks!

Don
http://www.dlcphotography.net

I understand that YARC Plus will offer ARF with linear conversions soon...

I have been using lin conversions processed in PS with Colin Walker's latest V3 profile/action an been getting very accurate results.

John

BDF said:

I have had my D60 for about 6 weeks now and have been
experimenting. I have found that converting to non linear 16 bitt
tiff with YARC's ARF yields superior results to converting to
linear tiff and using either Fred's or Colin's linear actions. The
YARC ARF conversions give deeper colors and a more pleasing image
overall. This is in contrast to my experience with the D30 where I
thought that linear conversion was superior (using Pekka's
LinearSharpen).

Any comments appreciated.

Click to expand...

If you want to retreive blown highlights, the best way I've found is to convert the RAW file twice, once for shadows and once for highlights and put these two images in one Photoshop file as two layers. Then I use the eraser tool at a low percentage to reveal the better highlight exposure in the lower layer. I used this recently for an annual report portrait shoot; white shirts are often blown out if faces are correctly exposed. It was very straightforward to bring the shirts back.

If you've got an image in which the highlights only blow out once you've used levels or curves to get the main subject looking correct, you can use the History Brush to restore the newly blown out areas (again, the brush can be set at a low percentage to slowly build up the effect). In the History Palette you'll need to click the History Brush box at the state just prior to the Levels or Curves change.

Regards

Hi Richard,

I use that History Brush technique all the time, and agree it is quite effective. It's great for restoring details in bright areas that would get blown out when levels/curves are adjusted to brighten up the rest of the image.

My question, though, is whether or not linear conversion actually maintains true image detail in those blown out areas, or just gives those areas a darker color (rather than being white). Any examples of actual image detail being present in those blown-out areas in linear vs non-linear conversion?

I guess I can do it myself, but was hoping somebody had done some basic controlled comparisons.

Regards,

Don
http://www.dlcphotography.net

Don:

Yes, lin conversion/processing has a wider dynamic range. I am getting very good results with Colin Walker's profile/action for the D60. I'm not familiar with what's available on the 1D. You need to find a linear processing action that does a good job with the 1D files.

John

DonCohen said:

Hi Richard,

I use that History Brush technique all the time, and agree it is
quite effective. It's great for restoring details in bright areas
that would get blown out when levels/curves are adjusted to
brighten up the rest of the image.

My question, though, is whether or not linear conversion actually
maintains true image detail in those blown out areas, or just gives
those areas a darker color (rather than being white). Any examples
of actual image detail being present in those blown-out areas in
linear vs non-linear conversion?

I guess I can do it myself, but was hoping somebody had done some
basic controlled comparisons.

Regards,

Don
http://www.dlcphotography.net

Click to expand...

Hi, Don -

I do not convert to 16 bit linear TIFFs for the purpose of restoring highlight detail, only to avoid in-camera processing. Therefore, I haven't performed any tests to determine whether or not linear conversion does indeed help in this area (by itself).

I do not find that I have many images which suffer from burned out highlight detail, but when I do have one, I have used the technique described in this thread by Richard L.

I convert the RAW to a 16bit linear TIFF, then convert the same RAW again as a 16bit non-linear TIFF. The linear TIFF becomes the background layer in Photoshop, I then paste the 16bit non-linear TIFF as a layer above the linear TIFF layer. After that, I use the eraser tool to erase the burned out non-linear layer details, revealing the (hopefully) non burned out linear details beneath.

Of course, this method causes the resulting image to suffer the effects of in-camera processing, (because the non-linear layer becomes the resulting final layer after flattening the image), but whaddya gonna do?

That's all I know...

DonCohen said:

Hi Gary,

Thanks for the explanation of how/why you use Linear files. I
certainly understand the logic, as I convert my Raw files (prior
D30 and current 1D) using the lowest settings for contrast,
sharpening, etc. to provide the least processed image to work with
in Photoshop.

I have never done any controlled studies to more definitively
answer this question, but I'm inclined to think that a good
photographer, with good Photoshop skills, will have pretty much
comparable results using either workflow method, in terms of
brightness, contrast, color accuracy, etc.

My main interest in possibly using linear files would be to
theoretically reclaim blown highlights in images I have that are
otherwise good, but where I messed up the exposure a bit.

Do you have any examples that demonstrate true retention of detail
in areas of blown highlights using linear vs non-linear conversion?
I have played around with this some, and my impression (not a firm
conclusion) is that the linear gave me gray areas, rather than
white areas, but otherwise held no other true image detail. But I
do remain open-minded about this, which is why I ask if you have
any reasonably controlled examples that demonstrate this.

Thanks!

Don
http://www.dlcphotography.net

Click to expand...

--
Gary Shepard
Foreside PhotoGraphics
Maine, USA

John/All -

Where can I find Colin Walker's profile/action for the D60? I've been using Fred Miranda's but find myself tweaking color balance afterwards a lot of the time.

Thanks -

John Davis said:

Don:

Yes, lin conversion/processing has a wider dynamic range. I am
getting very good results with Colin Walker's profile/action for
the D60. I'm not familiar with what's available on the 1D. You need
to find a linear processing action that does a good job with the 1D
files.

John

Click to expand...

--
Gary Shepard
Foreside PhotoGraphics
Maine, USA

gary shepard said:

Hi, Don -

I do not convert to 16 bit linear TIFFs for the purpose of
restoring highlight detail, only to avoid in-camera processing.
Therefore, I haven't performed any tests to determine whether or
not linear conversion does indeed help in this area (by itself).

I do not find that I have many images which suffer from burned out
highlight detail, but when I do have one, I have used the technique
described in this thread by Richard L.

I convert the RAW to a 16bit linear TIFF, then convert the same RAW
again as a 16bit non-linear TIFF. The linear TIFF becomes the
background layer in Photoshop, I then paste the 16bit non-linear
TIFF as a layer above the linear TIFF layer. After that, I use the
eraser tool to erase the burned out non-linear layer details,
revealing the (hopefully) non burned out linear details beneath.

Of course, this method causes the resulting image to suffer the
effects of in-camera processing, (because the non-linear layer
becomes the resulting final layer after flattening the image), but
whaddya gonna do?

Click to expand...

Gary,

could you not use a linear for what you are calling the "non-linear layer" but run, for example, Pekka's LinearSharpen across it first?

That would get you the two "exposures" but without any in-camera processing, wouldn't it?

KRs
Chris

Hi John,

John Davis said:

Yes, lin conversion/processing has a wider dynamic range. I am
getting very good results with Colin Walker's profile/action for
the D60. I'm not familiar with what's available on the 1D. You need
to find a linear processing action that does a good job with the 1D
files.

Click to expand...

I guess what I'm after is the objective basis for coming to this conclusion. Many time, "conventional wisdom" tells us this or that about any particular subject, but when someone takes the time to try to answer the question, in an objective fashion, that can be duplicated by others (the basis of the scientific method), those ideas often are proved invalid.

I'm not saying this or isn't the case with linear conversion, as I haven't done the homework myself, and haven't seen it demonstrated empirically by anybody else one way or the other.

It's in this context that I'm asking if anybody has or can objectively demonstrated that linear conversion does in fact maintain highlight detail that is lost in a non-linear conversion. I know the theory behind this, and I'm not saying it's incorrect. I'm just saying I haven't seen anybody demonstrate this in a scientific fashion.

Thank for your input.

Don
http://www.dlcphotography.net

DonCohen said:

Member said:

Yes, lin conversion/processing has a wider dynamic range. I am
getting very good results with Colin Walker's profile/action for
the D60. I'm not familiar with what's available on the 1D. You need
to find a linear processing action that does a good job with the 1D
files.

Click to expand...

I guess what I'm after is the objective basis for coming to this
conclusion. Many time, "conventional wisdom" tells us this or that
about any particular subject, but when someone takes the time to
try to answer the question, in an objective fashion, that can be
duplicated by others (the basis of the scientific method), those
ideas often are proved invalid.

I'm not saying this or isn't the case with linear conversion, as I
haven't done the homework myself, and haven't seen it demonstrated
empirically by anybody else one way or the other.

It's in this context that I'm asking if anybody has or can
objectively demonstrated that linear conversion does in fact
maintain highlight detail that is lost in a non-linear conversion.
I know the theory behind this, and I'm not saying it's incorrect.
I'm just saying I haven't seen anybody demonstrate this in a
scientific fashion.

Thank for your input.

Don
http://www.dlcphotography.net

Doctor,

Click to expand...

I love your thirst for the scientific method, it's refreshing. Enough of "he said", or "she thought"...just show me the facts! It's amazing, though, some minds just don't work this way.
Don (also)

no text

Don:

This is far from a proof. I just took a RAW image (courtesy David Eppstein) and converted/processed it both ways. I think that there is more detail in the shadows of the lin image - wider dynamic range. The only processing was to apply levels (conservative stretch of the 3 color componets) and resizing, followed by modest sharpening. The lin image can be sharpened further than the non-linear.

Non-linear conversion (converted w/ YARC Plus, Manual Levels, Resize to 800X533, USM with 125%, 0.5, 0)



Linear Conversion (converted w/ YARC Plus, processed w/ Colin Walker's Profile/action V.3, Manual Levels, Resize to 800X533, USM with 125%, 0.5, 0)



You asked for highlghts, and this is not a good example of those. I call your attention, though to the blown highlights in the out-of-focus blue flower on the left side.

I feel that the lin-processed image is more natural looking, more 3-dimensional, particularly on the left-side, out-of-focus, elements.

John

DonCohen said:

Member said:

Yes, lin conversion/processing has a wider dynamic range. I am
getting very good results with Colin Walker's profile/action for
the D60. I'm not familiar with what's available on the 1D. You need
to find a linear processing action that does a good job with the 1D
files.

Click to expand...

I guess what I'm after is the objective basis for coming to this
conclusion. Many time, "conventional wisdom" tells us this or that
about any particular subject, but when someone takes the time to
try to answer the question, in an objective fashion, that can be
duplicated by others (the basis of the scientific method), those
ideas often are proved invalid.

I'm not saying this or isn't the case with linear conversion, as I
haven't done the homework myself, and haven't seen it demonstrated
empirically by anybody else one way or the other.

It's in this context that I'm asking if anybody has or can
objectively demonstrated that linear conversion does in fact
maintain highlight detail that is lost in a non-linear conversion.
I know the theory behind this, and I'm not saying it's incorrect.
I'm just saying I haven't seen anybody demonstrate this in a
scientific fashion.

Thank for your input.

Don
http://www.dlcphotography.net

Click to expand...

John Davis said:

This is far from a proof. I just took a RAW image (courtesy David
Eppstein) and converted/processed it both ways. I think that there
is more detail in the shadows of the lin image - wider dynamic
range.
...
You asked for highlghts, and this is not a good example of those. I
call your attention, though to the blown highlights in the
out-of-focus blue flower on the left side.

Click to expand...

John --

You asked in another thread whether I mind this use of my photo -- no, no problem.

Re highlights and shadows, I've also found that this is one of the big advantages of linear processing (along with generally more accurate colors despite my difficulties in the other thread). An example on the highlight side is below -- I'd rather not post the .crw for it, but with a nonlinear conversion the foreground just looked white with little black specks. With linear workflow I was able to produce a lot more color and detail while still keeping that area very light. The same photo also has better shadow detail (e.g. on the animal's face and neck) than the nonlinear conversion.

This was processed with an old version of Colin's converter (1.7), with the only other touch-up being a curve on the L channel in Lab mode and a little sharpening.

BDF said:

I have had my D60 for about 6 weeks now and have been
experimenting. I have found that converting to non linear 16 bitt
tiff with YARC's ARF yields superior results to converting to
linear tiff and using either Fred's or Colin's linear actions. The
YARC ARF conversions give deeper colors and a more pleasing image
overall. This is in contrast to my experience with the D30 where I
thought that linear conversion was superior (using Pekka's
LinearSharpen).

Any comments appreciated.

Click to expand...

BDF,

I am glad that you are so pleased with the results thst ARF has provided to you. While I won't get into the Linear vs. Non-Linear debate, I will comment that I receive mail all the time from D60 (and 1D) users who reports that ARF seems to do more than just remove artifacts. While on a technical level, what it does do is just remove artifacts, it is the result of losing this "junk" (as we like to call it) that is so interesting.

May D60 users report more 3-dimentionality or more depth. In my opinion, the removal of the artifacts gives the images an apparent change, analogous to removing tape hiss from a recording (remember tape hiss). It changes the appranet sound of the recording when in fact only the hiss has been removed. You can hear "deeper" into the recording with the artifacts (tape hiss) removed. And if you chose to process the recording (with EQ or reverb or whatever), you could do it more freely, just like you can process an ARF'd image more freely in PS.

The main attribute of ARF is that without the artifacts (to get in the way) ALL post processing of almost any kind, can work better, you can apply more, and there are less side effects based on the processing. Also noise reduction processes such as Fred's ISOR can be more effective because they don't have to "mess" with the artifacts that they are not effective at removing anyway.

So my advice to the readers, is to experiment with ARF and determine how it can best work in each persons workflow. I have found that with rare exception, an ARF'd image will always produce better final results. Our evaluation version of YarcPlus allows full experimentation with ARF (so does the registered version > .

I have many mails from people who were skeptical, but after doing controlled and blind teest, have come to appreciate the beneifts of ARF.

And....

...as has been hinted to elsewhere, I can tell you that ARF will be available for use with Linear conversions in the next version of YarcPlus planned for release this week (along with some other exciting features).

BDF...Once again, I am pleased that ARF has added to the quality of your work. Thanks for sharing your experience.

--
Regards,

Michael Tapes
http://www.YarcPlus.com
http://www.michaeltapes.com

Hi John and David,

I appreciate your thoughts and experience on this matter.

1. With regard to blown-out highlight recovery:

I think the difficulty here, in trying to make truly objective comparisons between the final results from linear vs non-linear conversions, is that it becomes more a test of Photoshop processing techniques. I played around with some images myself, and had great difficulty trying to even define a technique that would allow valid comparisons. I found no way to make direct comparisons in converted images, prior to any Photoshop manipulations. The linear images obviously looked horrible, while the non-linear images contained detail in the blown areas not evident at first, but which could be recovered with adjustment of curves and levels.

I did some extreme manipulations to try and determine if there were actual pixel differences in those blown-out areas. I will acknowledge that there might have been a few pixels of more information contained in the linear samples; but at the same time, these were so minimal/subtle that I didn't think it likely that after correct Photoshop manipulations (for both images), that they would be visible or provide actual image detail or differences.

2. With regard to overall dynamic range:

To me, this is even more a test of Photoshop skills than anything else. The curves tool is extremely powerful when it comes to dealing with this issue, in maintaining shadow detail, etc.

I suspect that if Fred, Pekka, Colin, etc. had spent their considerable talents on tweaking Photoshop actions for non-linear images, rather than linear images, that people would be just as pleased/excited with their results as they are now with linear. But this is just my speculation, not based on hard evidence.

I think that MichaelT's analogy between the hi-end audio world and digital photography imaging techniques is appropriate in this context as well: it is very difficult to sort out subjective impressions from objective differences, and different people will have legitimately different opinions on which is "better" (which is a value judgment based on one's personal preferences).

BOTTOM LINE:

For me, I'd rather spend time working on my photography in general (and Photoshop skills as well) than on the question of whether or not linear conversion provides superior results. (And this would include improving my photo skills so I don't have blown highlights to begin with!).

I'm definitely a pragmatist, and I am convinced that outstanding images can be achieved by a variety of techniques.

I have no definite conclusion as to whether linear conversion provides any theoretical benefit over non-linear conversion, but am inclined to think that whatever theoreticall differences exist, they are of little practical significance in terms of the end result if equally careful and competent Photoshop processing techniques were applied to each.

So, use whatever technique you're comfortable with, don't get obsessed over the minutiae, and get out there and shoot!

Thanks for the interesting discussion.

Don
http://www.dlcphotography.net