ACR default shadow value

rutt · March 29, 2005

Some people may not have noticed, but the default value for "Shadows" in ACR is 5 (at least for my Canons), which means that some shadow detail is
discarded by default. I always send up moving this value to recover shadow detail and then use curves in subsequent PS editing in order to get a more accurately targeted black point. I want to lose the detail I want to lose and no more. I think that 0 is a better default value for "Shadows". You can make 0 the default by setting it and then using "Set Camera Default".

You want to do this if you are comfortable with curves (or their poor cousins, levels). Even people who are more comfortable with the ACR controls should consider this change. You can always mess with the shadow slider and if you don't intend to use curves or levels later on, you should.

cletus · March 29, 2005

Rutt,

I too adjust the Shadow slider every time I go through Adobe Camera Raw (ACR). However, sometimes I increase it, sometimes I decrease it. I just use the Alt + slide (or Option + slide on a mac) and adjust the Shadow slider until I just start to see a few pixels show up. Yes I am throwing away info in those pixels, but as long as it's only a few isolated pixels I don't think it will make any difference in shadow details. In addition since I run the eyedropper in "5 by 5 Average" mode I don't think those few isolated pixels will have an effect on setting black points inside Photoshop.

FWIW I do the same thing with the Exposure slider as well.

rutt · March 29, 2005

Yeah, that's consistent with the point of my post. I'm much more comfortable setting my shadow and highlight points with curves in PS after conversion and I almost always go through a curves step anyway. I think you lose nothing in terms of quality by getting all the information in the histogram before conversion and then adjusting in PS, but there is a little controversy on this point which I'm researching. There really is no controversy that you can acheive results that are at least as good for all practical purposes by getting entire histogram in bounds and converting to 16 bit format, and then adjusting contrast, shadow, highlights, &etc in PS. So, I guess that means it's just up to you what's more comfortable. Certainly the WB feature of ACR is a lot easier (even for me) to use than using curves later on to correct (say) for tungsten. The result isn't always so perfect that I can leave it alone, but at least it's on the right planet.

pathfinder · March 29, 2005

cletus wrote:

Rutt,

I too adjust the Shadow slider every time I go through Adobe Camera Raw (ACR). However, sometimes I increase it, sometimes I decrease it. I just use the Alt + slide (or Option + slide on a mac) and adjust the Shadow slider until I just start to see a few pixels show up. Yes I am throwing away info in those pixels, but as long as it's only a few isolated pixels I don't think it will make any difference in shadow details. In addition since I run the eyedropper in "5 by 5 Average" mode I don't think those few isolated pixels will have an effect on setting black points inside Photoshop.

FWIW I do the same thing with the Exposure slider as well.

If you move the shadow slider to "0" and the lowest pixel black in your original image is 15, 15, 15 - a dark grey - I think you will find the color saturation of the image is impaired, the darker tones will lack depth. Since no pixels are lost until you start to see black pixels with the Alt-Shadow slider routine, I don't see what you gain by moving the slider to 0 for the Shadow level as a routine thing? Don't you want to use the entire available range of intensities for the data available? By routinely using a Shadow level of 0, aren't you not choosing to not use all the available levels at your disposal???

cletus · March 29, 2005

pathfinder wrote:

Don't you want to use the entire available range of intensities for the data available?

Yes you do. You've got the right idea when it comes to using ACR to create 8-bit PS files, but as Rutt pointed out there is a difference between 8-bit and 16-bit files created by ACR. The 16-bit file can hold all of the intensities that were in the RAW image, so there really isn't a need to use ACR to expand out the intensites before making a PS file. In fact Rutt's set Shadows to 0 is pretty good advice if you're making 16 bit files. It doesn't hurt anything since all of the original raw file intensities will be in your 16-bit PS, and you can make a curves or levels adjustment inside PS and get the same results as if you used the Shadow and Exposure sliders in ACR.

rutt · March 29, 2005

Thanks, Eric. I was about to respond, but you saved me the trouble.

Let me try to say it in slightly different words. So long as different pixel values in the RAW image map into different pixel values in the internal PS representation after conversion, no information is lost. You can use curves or whatever to spread them out more or condense them.

pathfinder · March 29, 2005

cletus wrote:

Yes you do. You've got the right idea when it comes to using ACR to create 8-bit PS files, but as Rutt pointed out there is a difference between 8-bit and 16-bit files created by ACR. The 16-bit file can hold all of the intensities that were in the RAW image, so there really isn't a need to use ACR to expand out the intensites before making a PS file. In fact Rutt's set Shadows to 0 is pretty good advice if you're making 16 bit files. It doesn't hurt anything since all of the original raw file intensities will be in your 16-bit PS, and you can make a curves or levels adjustment inside PS and get the same results as if you used the Shadow and Exposure sliders in ACR.

Ok I see what your'e saying. So then it is just whether you are more comfortable doing the change in ARC or using curves in PS. Hmmmm.... Despite the fact that ARC data is a linear array and 16 bit files are corrected for a gamma adjusted array?

If s,o then there is no reason to make 16 bit files if the corrections are made first in ARC? Just go directly to 8 bit files? I know that is the opinion held by some workers, but perhaps, not by all. This seems to get murkier rather than clearer. Good discussion Cletus and Rutt.

rutt · March 29, 2005

pathfinder wrote:

Ok I see what your'e saying. So then it is just whether you are more comfortable doing the change in ARC or using curves in PS. Hmmmm.... Despite the fact that ARC data is a linear array and 16 bit files are corrected for a gamma adjusted array?

This is the only part of the picture I don't completely understand. I'm working on it, doing my homework, but I'm not there. But let me use logic instead of true understanding for a minute. Just so long as ARC doesn't end up mapping different pixel values to the same pixel value (and so long as the mapping is monotonic) then no information is lost in the conversion. Curves allow any function to be expressed, so we can always use them to acheive whatever we want. Now it may be that the "gamma adjusted array" in PS has some properties that require ACR to lose information in some cases, and perhaps it loses more with a more compressed histogram. That's the part where my knowledge is incomplete. But, I really guess we have some angels dancing on the head of our pin at this pointt.

What is undoubtedy true is that ACR will lose information if the shadow or exposure values are too high.

pathfinder · March 29, 2005

rutt wrote:

This is the only part of the picture I don't completely understand. I'm working on it, doing my homework, but I'm not there. But let me use logic instead of true understanding for a minute. Just so long as ARC doesn't end up mapping different pixel values to the same pixel value (and so long as the mapping is monotonic) then no information is lost in the conversion. Curves allow any function to be expressed, so we can always use them to acheive whatever we want. Now it may be that the "gamma adjusted array" in PS has some properties that require ACR to lose information in some cases, and perhaps it loses more with a more compressed histogram. That's the part where my knowledge is incomplete. But, I really guess we have some angels dancing on the head of our pin at this pointt.

What is undoubtedy true is that ACR will lose information if the shadow or exposure values are too high.

That is interesting, because in my workflow if I have an image without a lot of very dark tones in the origianl image, I WILL raise the shaoow slider until I see the very first pixels blow, or until the image looks better - whichever comes first, and then I convert to a 16bit ProPhoto space for most of my final image editing, and then convert to 8bits. Maybe this is just making work for oursleves. When I get some free time, I'll try a few both ways - yours and mine - yours as 16bit image in PS, mine in ARC and see what falls out.

rutt · March 29, 2005

pathfinder wrote:

Maybe this is just making work for oursleves. When I get some free time, I'll try a few both ways - yours and mine - yours as 16bit image in PS, mine in ARC and see what falls out.

It's not that far from what I do. I make sure I have no blow bits on either end and then fix up in PS. But I'm more comfortable with the power of curves than with the simple sliders in ACR, so I almost always at least see if I can get something from a curves step.

One thing that I have found is that increasing/decreasing ACR exposure changes how NeatImage processes the image. Increasing exposure adds noise. So does shadow/hightlight, but not necessarily curves. So I try to be conservative with increasing exposure (but will happily decrease in order to get blown areas in range.)

pathfinder · March 29, 2005

rutt wrote:

It's not that far from what I do. I make sure I have no blow bits on either end and then fix up in PS. But I'm more comfortable with the power of curves than with the simple sliders in ACR, so I almost always at least see if I can get something from a curves step.

One thing that I have found is that increasing/decreasing ACR exposure changes how NeatImage processes the image. Increasing exposure adds noise. So does shadow/hightlight, but not necessarily curves. So I try to be conservative with increasing exposure (but will happily decrease in order to get blown areas in range.)

The reason you begin to get noise with overexposed bits is that ARC ( unlike most RAW converters ) will try to fill in data as long as at least one channel of the RGB space remains unblown. Once all three slots, R G AND B, are blown it can no longer reconstruct pixels. But if only one channel is at 255, it will try and reconstruct the other two- that is why you can sometimes gain almost 1 f stop of exposure correction. This of course leads to more noise, but the alternative is no exposure correction at all. At least that is my undrestanding from B Fraser's book about ARC.

cletus · March 29, 2005

pathfinder wrote:

Ok I see what your'e saying. So then it is just whether you are more comfortable doing the change in ARC or using curves in PS. Hmmmm.... Despite the fact that ARC data is a linear array and 16 bit files are corrected for a gamma adjusted array?

I had forgotten about ARC making gamma adjustments before creating the PS file. I guess it is possible (though I really doubt it) that the gamma correction could alter things enough so that you would wind up with a difference between a tweaked in ARC 16 bit file and and PS only adjusted image.

rutt · March 29, 2005

pathfinder wrote:

The reason you begin to get noise with overexposed bits is that ARC ( unlike most RAW converters ) will try to fill in data as long as at least one channel of the RGB space remains unblown. Once all three slots, R G AND B, are blown it can no longer reconstruct pixels. But if only one channel is at 255, it will try and reconstruct the other two- that is why you can sometimes gain almost 1 f stop of exposure correction. This of course leads to more noise, but the alternative is no exposure correction at all. At least that is my undrestanding from B Fraser's book about ARC.

When you need exposure correction, then of course you need it and fine, you'll end up with noise. But I more often turn down exposure in order to recover blown information than turn it up in order to spread out the data acros the the histogram. Turning it up (the equivalent of "pushing" film) seems to make a lot of noise fast, so that's what I try to avoid (and usually can avoid.) Turning down the shadow value doesn't seem to increase noise. All of this runs a counter to your reasoning, and so falls into the "homework required" category.

rutt · March 29, 2005

cletus wrote:

I had forgotten about ARC making gamma adjustments before creating the PS file. I guess it is possible (though I really doubt it) that the gamma correction could alter things enough so that you would wind up with a difference between a tweaked in ARC 16 bit file and and PS only adjusted image.

Yeah, and if you did, would you be able to perceive it? Those were the angels on the pin I was mentioning.

cletus · March 29, 2005

pathfinder wrote:

If s,o then there is no reason to make 16 bit files if the corrections are made first in ARC? Just go directly to 8 bit files?

I don't think I would want to say that ARC adjustments do away with the need for 16 bit files. Even with ARC adjustments, you're still going to get more intensity levels in a 16 bit file than you could in an 8 bit file. More intensity levels mean more detail and more information for the PS tools to work with.

Where things get really crazy is when you start talking about the subjective side of things... like "will I be able to see the difference in a print made from an 8 bit file that was properly adjusted in ACR and a print made from a 16 bit file that was created with little or no adjustments made in ACR and then curve adjusted in PS?" In general my answer would be :nah

However, there are probalby times where you want use the tools in PS to make extreme changes to an image (big color corections, radical curves adjustments, etc...) where I would bet that the difference between a 16 bit file and a 8 bit file (even if the 8 bit file was created from a properly ACR'd RAW file) would be night and day.

pathfinder · March 29, 2005

rutt wrote:

When you need exposure correction, then of course you need it and fine, you'll end up with noise. But I more often turn down exposure in order to recover blown information than turn it up in order to spread out the data acros the the histogram.

That's right - if you have overexposed your highlights, you want to turn down the exposure slider. I agree. But where do you think the information for the blown highlights comes from? That was my point and I guess I did not explain it well either. The highlights that retain data in one of the original color channels are reconstructed by ARC. Yes, there is not a lot of noise dialing back exposure, but there must be some since you are reconstructing partially blown highlight pixels.

Turning it up (the equivalent of "pushing" film) seems to make a lot of noise fast, so that's what I try to avoid (and usually can avoid.) Turning down the shadow value doesn't seem to increase noise. All of this runs a counter to your reasoning, and so falls into the "homework required" category.

Agreed -- Significant underexposure is really harder to correct than slight overexposure, and causes much more noise in the image - again because there is more noise in the darker part of the image to start with, and there is much, much less data in the darker levels of a normal exposure, let alone an under exposed image. With a 16 bit image with a potential 64,000 bits, the first stop of a level has 32,000 bits, the second level, 16,000 bits, the third level 8,000 bits, the 4th level 4,000 bits, the 5th level 2,000 bits, the 6th level 1,000 bits, the 7th level 500 bits, and the 8th level 256 bits. Try interpolating with only 256 bits versus interpolating with 32,000 bits. And that would be for a normal exposure, rather than an underexposed frame.

rutt · March 29, 2005

Makes sense, Jim. It seems we have come to a point of mutual understanding and perhaps even agreement. Whew.

Andy · March 29, 2005

ok, my head hurtz now

good discussion.

pathfinder · March 29, 2005

andy wrote:

ok, my head hurtz now

good discussion.

Clear as mud, right Andy??

Andy · March 29, 2005

pathfinder wrote:

Clear as mud, right Andy??

well, i do alot of this, but i don't think about it too much. i process more by feel than by textbook. i could do more, i s'pose... i'll stock up on motrin

rutt · March 29, 2005

andy wrote:

well, i do alot of this, but i don't think about it too much. i process more by feel than by textbook. i could do more, i s'pose... i'll stock up on motrin

I suppose if your feel is good enough you don't need knowledge. My alma mater's motto was "Non Satis Scire". The literal translation is, "To know is not enough." But a college buddy of mine really hit the nail on the head when he made a more postmodern translation, "To know is not necessary."

Andy · March 29, 2005

rutt wrote:

I suppose if your feel is good enough you don't need knowledge. My alma mater's motto was "Non Satis Scire". The literal translation is, "To know is not enough." But a college buddy of mine really hit the nail on the head when he made a more postmodern translation, "To know is not necessary."

well, i like that motto - i only have so many brain cells, y'know? but all kidding aside, i'm counting on guys like you rutt, to point out when i could put in some extra "in post" and get more out of an image

appreciate your contributions here on dgrin, very much, and in this thread in particular

ACR default shadow value

Comments