Ad
Ad
Ad
Pages: [1] 2 »   Bottom of Page
Print
Author Topic: Post processing images exposed to the right  (Read 9650 times)
willie45
Newbie
*
Offline Offline

Posts: 30


« on: April 12, 2008, 05:34:07 PM »
ReplyReply

Hi

I have been reading about exposing to the right on this site and also on the "Camera to Print videos. I have read quite a few of the debates here about its various merits and demerits. I am keen to use this technique and find out the resuls for myself. I do have one very basic question though.

The question is probably borne out of lack of experience with ACR as much as anything. I am able to busk my way around the sliders to get a reasonable result but I know I really need to get more knowledge of its workings and workflow to progress.

Anyway, once I have the image on my computer in and my histogram is well to the right without blowing highlights, do I just use all my controls as normal including the exposure slider. This will mean I have to bring the exposure down to  minus something or other and much more so than witha conventional exposure. I am assuming that is right because I can't see how else you could make it work. I know this is a bit of a daft question but just wanted to make sure I hadn't missed out on something,

Willie
« Last Edit: April 12, 2008, 05:35:53 PM by willie45 » Logged
michael
Administrator
Sr. Member
*****
Offline Offline

Posts: 4871



« Reply #1 on: April 13, 2008, 08:02:57 AM »
ReplyReply

Yes, that's all there is to it.

Just "normalize" the exposure so that it looks correct using the Exposure slider. What you'll have done though is to move the data-rich portion of what has been recorded down into the data poor mid-tone and quarter tone area, thus giving yourself a more robust image to work with.

Michael
Logged
CynthiaM
Full Member
***
Offline Offline

Posts: 105



WWW
« Reply #2 on: April 13, 2008, 08:48:04 AM »
ReplyReply

Michael:

With regard to "exposing to the right," my understanding is that the histogram on the LCD is displaying information based on a jpeg, even if camera is set to shoot raw.  Should one allow any blinking highlights to remain if shooting raw and trying to expose to the right?  If you keep dialing in negative exposure compensation until no blinkies remain, do you essentially dial away usable data that a raw capture could utilize/recover?
Logged

Cynthia Merzer
My Home page
CatOne
Sr. Member
****
Offline Offline

Posts: 381


WWW
« Reply #3 on: April 13, 2008, 10:18:23 AM »
ReplyReply

Quote
Michael:

With regard to "exposing to the right," my understanding is that the histogram on the LCD is displaying information based on a jpeg, even if camera is set to shoot raw.  Should one allow any blinking highlights to remain if shooting raw and trying to expose to the right?  If you keep dialing in negative exposure compensation until no blinkies remain, do you essentially dial away usable data that a raw capture could utilize/recover?
[a href=\"index.php?act=findpost&pid=189189\"][{POST_SNAPBACK}][/a]

Correct, the histogram is based on a JPEG.  You can typically pull back some detail in the RAW file which the "blinkies" indicate is "blown out."  But, the ability to recover isn't infinite... personally I'd not try to recover more than about 2/3 stop, and if I had huge areas of a photo (say, a white shirt in sunlight) that were all blinking... I'd re-shoot.

ETTR is nice, and it can help you capture more detail in a RAW photo because of the math of how things work, but practically today's cameras (if you have one made in the last couple years) are so good with high ISO performance in shadow noise that I don't think it's worth worrying TOO much about it.  Take the shot as you normally would, maybe dial in 1/3 to 2/3 stops of exposure compensation, and turn on the highlight warnings to make sure you don't overbake it *too* much.

One of the nice things about digital *is* the histogram on the back of the camera.  I was still spending a bunch of time spot metering and worrying about exposure when Marc Muench came to me and said "why not just run on aperture priority, take the shot, see what you get, dial in the right exposure and shoot again... takes a couple seconds and eliminates the need to meter."  Which actually is very helpful for landscape (for some things, knowing exposure correctly is still helpful ;-)  It's saved me a lot of time and it's not too hard to get a correct exposure... like it was with film where the feedback loop was a day at least.
Logged

willie45
Newbie
*
Offline Offline

Posts: 30


« Reply #4 on: April 13, 2008, 11:30:48 AM »
ReplyReply

Thanks for your reply Michael. I supposed it must be so but then started to wonder and wanted to make sure. I will go out and do it.

Also thanks Cat One for your helpful comments.

Willie
Logged
CynthiaM
Full Member
***
Offline Offline

Posts: 105



WWW
« Reply #5 on: April 14, 2008, 10:54:26 AM »
ReplyReply

Thanks for your reply.  I just never know how much in the way of blinking highlights to allow to remain.  Being that it's RAW, I'll dial in negative exposure compensation until the point where there are barely any "blinkies;" minimal, almost to the point where you have to look twice to see them.
Logged

Cynthia Merzer
My Home page
Panopeeper
Sr. Member
****
Offline Offline

Posts: 1805


« Reply #6 on: April 14, 2008, 11:18:39 AM »
ReplyReply

Quote
I just never know how much in the way of blinking highlights to allow to remain. 
There is no way to know that, *except* when shooting with the total neutral setting (no sharpening, contrast 0, saturation 0, tone adjustment 0) and with a "neutral WB"; see this thread, particularly the last post: http://luminous-landscape.com/forum/index....22&#entry189022, though you should read the beginning as well.

That way you can achieve perfect ETTR; however, the preview and the thumbnail images will look very strange this way.

As to the raw processing: you need to know your raw data and how it will be treated by the particular raw processor. The clipping indication has nothing to do with pixel saturation. ACR can make adjustments you have no idea of; see the thread http://luminous-landscape.com/forum/index....showtopic=24354 on this topic.
Logged

Gabor
CynthiaM
Full Member
***
Offline Offline

Posts: 105



WWW
« Reply #7 on: April 16, 2008, 10:39:09 AM »
ReplyReply

Quote
As to the raw processing: you need to know your raw data and how it will be treated by the particular raw processor. The clipping indication has nothing to do with pixel saturation. ACR can make adjustments you have no idea of; see the thread http://luminous-landscape.com/forum/index....showtopic=24354 on this topic.
[a href=\"index.php?act=findpost&pid=189431\"][{POST_SNAPBACK}][/a]

So I guess what I need to do before reformatting a card is to put it back in the camera (not all the time, but just to get some sense of the difference between my camera's histogram and the raw converter), look at the histogram on the camera, and compare it to whether or not the raw processor is showing that I'm blowing out highlights?
Logged

Cynthia Merzer
My Home page
Panopeeper
Sr. Member
****
Offline Offline

Posts: 1805


« Reply #8 on: April 16, 2008, 10:59:32 AM »
ReplyReply

Quote
So I guess what I need to do before reformatting a card
Off topic, but why would you reformat the card instead of deleting the files or the folder you don't need any more?

Quote
is to put it back in the camera (not all the time, but just to get some sense of the difference between my camera's histogram and the raw converter), look at the histogram on the camera, and compare it to whether or not the raw processor is showing that I'm blowing out highlights?
If the image was shot with such a setting and WB as described above, then you can see a more or less raw-like histogram, but nothing more. That is suitable for shooting, particularly because this is the best you can have (except with some MFDBs, which show true raw histograms).

However, for the detailed raw data analysis you need a program dedicated to that purpose. There are several such, but I can refer only to one of them, for I am using only that (as that is mine): Rawnalyze

It's user's guide can be used as a raw primer as well.
Logged

Gabor
Mike V
Newbie
*
Offline Offline

Posts: 6


« Reply #9 on: April 17, 2008, 11:45:00 PM »
ReplyReply

Read These:


Adobe Whitepaper: Raw Capture, Linear Gamma and Exposure

Adobe Technical Paper: Highlight Recovery in Camera Raw


.
Logged
bjanes
Sr. Member
****
Offline Offline

Posts: 2794



« Reply #10 on: April 18, 2008, 06:31:14 AM »
ReplyReply

Quote
Yes, that's all there is to it.

Just "normalize" the exposure so that it looks correct using the Exposure slider. What you'll have done though is to move the data-rich portion of what has been recorded down into the data poor mid-tone and quarter tone area, thus giving yourself a more robust image to work with.

Michael
[a href=\"index.php?act=findpost&pid=189180\"][{POST_SNAPBACK}][/a]

This method makes sense since increased exposure in the camera and the decrease in exposure with the ACR exposure slider are both linear scaling operations. For example, if you expose to the right by 1 stop, all pixels in the image are multiplied by 2 and an exposure correction of -1 in ACR reverses the process.

The above situation would take place with a short scale subject where the dynamic range of the subject is less than that of the camera and a normal exposure would leave empty space on the right of the histogram. If the dynamic range of the subject is equal to that of the camera, normal exposure and ETTR are the same.

If the dynamic range of the subject is greater than that of the camera, you must make a compromise and have the option of preserving the highlights or the shadows.  To preserve the highlights, you would place them just short of clipping or perhaps allow some clipping and use highlight recovery. If you favor the shadows, then highlight clipping will be present.

In the excellent LL ACR tutorial, Thomas Knoll discusses how he corrects images. He uses the exposure control to set the white point, holding down the ALT key (windows) to show highlight clipping. He then sets the black point with the black slider, also holding down the ALT key. He then uses the brightness control to adjust the distribution of tones between the white and black point. Brightness is nonlinear.

He doesn't explicitly state if he still uses this method for highlight recovery, but does note that an alternate approach is to use the exposure control to adjust the overall brightness of the picture and then use the recovery slider to control the highlights. The recovery slider is nonlinear and affects the highlights more than the remaining tones.

For highlight recovery, one may use either the exposure or the recovery slider and he gives no advice in this situation. In the case of a high dynamic range subject where the highlights are blown but the midtones are correct, use of the recovery slider is suggested. In the tutorial Jeff shows how to use curves to tease out highlight detail in such a case.

Bill
Logged
ejmartin
Sr. Member
****
Offline Offline

Posts: 575


« Reply #11 on: April 18, 2008, 10:11:40 AM »
ReplyReply

Quote
What you'll have done though is to move the data-rich portion of what has been recorded down into the data poor mid-tone and quarter tone area, thus giving yourself a more robust image to work with.

Michael
[a href=\"index.php?act=findpost&pid=189180\"][{POST_SNAPBACK}][/a]

Actually, the issue is not the "data richness" of any portion of the tonal range, if by that you mean the number of raw levels in any given segment of the raw data (as measured e.g. in stops down from raw saturation).  I believe that is the explanation given in the essay on ETTR on this site, but it is incorrect.  Rather, the point is that by exposing to the right, one achieves a higher signal to noise ratio in the raw data.  The number of available raw levels has nothing to do with it.

For example, here is the S/N ratio of the 5D (vertical axis, in stops; for dB, multiply by six), as a function of raw level (horizontal axis, in stops -- that is, horizontal coord x is raw level 2^x+1):



By exposing to the right, one takes advantage of the highest possible S/N one can record at a given ISO setting.  Note that, in the midtones and highlights (the top three stops), one achieves the same S/N ratio at lower ISO at lower raw level, roughly one stop lower raw level for one stop lower ISO (e.g. ISO 800 at raw level 2^11 has about the same S/N as ISO 400 at exposure 2^10; both would be achieved for a given subject with the same shutter speed and aperture).  So for midtones and highlights, there is little benefit to ETTR, and the danger of clipping highlights.  For shadows, there is a definite benefit to ETTR since raising the raw level by a stop gives one more S/N improvement than lowering the ISO at the same raw level.  The benefit is actually NOT in the tonal range "having the most raw levels".
« Last Edit: April 18, 2008, 10:42:34 AM by ejmartin » Logged

emil
Jonathan Wienke
Sr. Member
****
Offline Offline

Posts: 5759



WWW
« Reply #12 on: April 18, 2008, 10:17:20 AM »
ReplyReply

Quote
I believe that is the explanation given in the essay on ETTR on this site, but it is incorrect.  Rather, the point is that by exposing to the right, one achieves a higher signal to noise ratio in the raw data.  The number of available raw levels has nothing to do with it.

At lower ISO with some cameras, quantization errors (AKA lack of RAW levels) can be a significant factor in the noise level. But in many cases, especially at higher ISO, other forms of noise are much more significant. So the correct answer is "it depends".
Logged

ejmartin
Sr. Member
****
Offline Offline

Posts: 575


« Reply #13 on: April 18, 2008, 10:40:52 AM »
ReplyReply

Quote
At lower ISO with some cameras, quantization errors (AKA lack of RAW levels) can be a significant factor in the noise level. But in many cases, especially at higher ISO, other forms of noise are much more significant. So the correct answer is "it depends".
[a href=\"index.php?act=findpost&pid=190407\"][{POST_SNAPBACK}][/a]

No.  Any 12-bit camera with 11.7 stops or less of DR (all but perhaps one), and any 14-bit camera with 13.7 or less stops of DR (which is all of them) do not exhibit quantization errors to a degree that affects image quality in any way whatsoever.  Quantization error averages about .3 of the quantization step (one raw level), and sensor read noise is well over one raw level on all current DSLR's (except perhaps the Pentax K10D, though I haven't examined that myself).  All Canon and Nikon bodies have no possibility of exhibiting issues of quantization error, as long as one is shooting raw.
« Last Edit: April 18, 2008, 10:41:25 AM by ejmartin » Logged

emil
bjanes
Sr. Member
****
Offline Offline

Posts: 2794



« Reply #14 on: April 18, 2008, 10:58:13 AM »
ReplyReply

Quote
Actually, the issue is not the "data richness" of any portion of the tonal range, if by that you mean the number of raw levels in any given segment of the raw data (as measured e.g. in stops down from raw saturation).  I believe that is the explanation given in the essay on ETTR on this site, but it is incorrect.  Rather, the point is that by exposing to the right, one achieves a higher signal to noise ratio in the raw data.  The number of available raw levels has nothing to do with it.
[{POST_SNAPBACK}][/a]

I think that the signal to noise explanation of the benefits of ETTR is the most compelling and the number of tone levels is secondary. The brightest f/stop of a 12 bit file contains half the tonal levels, 2048 out of the total of 4096. The tonal response of the human eye is approximately logarithmic and the eye can distinguish only about 70 levels per f/stop as explained by [a href=\"http://www.normankoren.com/digital_tonality.html]Norman Koren.[/url].

Even if you underexpose by 2 f/stops, you would still have 512 levels in the brightest f/stop, which is plenty to prevent posterization even with rather extreme tone curves. However, the number of levels in darker f/stops is also decreased and at some point posterization might appear in the shadows. In most cases, shadow quality is limited by noise rather than posterization.

Bill
Logged
ejmartin
Sr. Member
****
Offline Offline

Posts: 575


« Reply #15 on: April 18, 2008, 11:11:04 AM »
ReplyReply

Quote
I think that the signal to noise explanation of the benefits of ETTR is the most compelling and the number of tone levels is secondary. The brightest f/stop of a 12 bit file contains half the tonal levels, 2048 out of the total of 4096. The tonal response of the human eye is approximately logarithmic and the eye can distinguish only about 70 levels per f/stop as explained by Norman Koren..

Even if you underexpose by 2 f/stops, you would still have 512 levels in the brightest f/stop, which is plenty to prevent posterization even with rather extreme tone curves. However, the number of levels in darker f/stops is also decreased and at some point posterization might appear in the shadows. In most cases, shadow quality is limited by noise rather than posterization.

Bill
[a href=\"index.php?act=findpost&pid=190429\"][{POST_SNAPBACK}][/a]

As far as posterization is concerned, the issue is always whether the number of bits exceeds the S/N ratio in stops.  At the top end, the S/N ratio maxes out around 8; you need little more than 256 levels in the highest stop of exposure to prevent posterization (512 to be absolutely safe).  At the bottom end (e.g. of the 5D), the lowest stop has S/N ratio about one, and one bit is sufficient since it's half noise anyway.  12 bits is never needed for posterization suppression, it's needed because the dynamic range between saturation and S/N ratio one is close to twelve stops at the lowest ISO's.
« Last Edit: April 18, 2008, 11:12:30 AM by ejmartin » Logged

emil
Panopeeper
Sr. Member
****
Offline Offline

Posts: 1805


« Reply #16 on: April 18, 2008, 11:20:17 AM »
ReplyReply

Quote
Actually, the issue is not the "data richness" of any portion of the tonal range, if by that you mean the number of raw levels in any given segment of the raw data (as measured e.g. in stops down from raw saturation). I believe that is the explanation given in the essay on ETTR on this site, but it is incorrect. Rather, the point is that by exposing to the right, one achieves a higher signal to noise ratio in the raw data. The number of available raw levels has nothing to do with it

The noise is one aspect, one reason for ETTR; the number of levels is another one. Particularly with cameras, which create only a few levels in the dark areas, the number of levels is an issue. By exposing as high as possible, the raw processor can work with more levels even in the dark areas.

Quote
No. Any 12-bit camera with 11.7 stops or less of DR (all but perhaps one), and any 14-bit camera with 13.7 or less stops of DR (which is all of them) do not exhibit quantization errors to a degree that affects image quality in any way whatsoever

Let's keep in eyes that the subject is mainly *photography*. Theoretically (with absolutely noise-free pixels) the 11th stop would consist of TWO levels.
« Last Edit: April 18, 2008, 11:20:51 AM by Panopeeper » Logged

Gabor
ejmartin
Sr. Member
****
Offline Offline

Posts: 575


« Reply #17 on: April 18, 2008, 01:30:26 PM »
ReplyReply

Quote
The noise is one aspect, one reason for ETTR; the number of levels is another one. Particularly with cameras, which create only a few levels in the dark areas, the number of levels is an issue. By exposing as high as possible, the raw processor can work with more levels even in the dark areas.

You have missed the point.  The number of levels is irrelevant, as long as the noise fluctuations exceed 1.3 levels.  S/N is the ONLY quantitative justification for ETTR.

Quote
Let's keep in eyes that the subject is mainly *photography*. Theoretically (with absolutely noise-free pixels) the 11th stop would consist of TWO levels.
[a href=\"index.php?act=findpost&pid=190436\"][{POST_SNAPBACK}][/a]

Theoretically, with noise-free pixels dynamic range is infinite and an arbitrarily large bit depth is justified for faithful recording of image data.  In *actual* cameras, there is a noise floor, and there is no problem that the 11th stop has only two levels if the noise is plus or minus well over one level -- except insofar as one doesn't like the fact that the noise is over one level.  If the eleventh stop down from saturation were made up of four or eight levels, or more, by the use of higher bit depth of recording, it would make no difference provided the noise floor remains the same; the noise totally masks the finer gradations of the extra levels.
Logged

emil
bjanes
Sr. Member
****
Offline Offline

Posts: 2794



« Reply #18 on: April 18, 2008, 03:29:34 PM »
ReplyReply

Quote
As far as posterization is concerned, the issue is always whether the number of bits exceeds the S/N ratio in stops.  At the top end, the S/N ratio maxes out around 8; you need little more than 256 levels in the highest stop of exposure to prevent posterization (512 to be absolutely safe).  At the bottom end (e.g. of the 5D), the lowest stop has S/N ratio about one, and one bit is sufficient since it's half noise anyway.  12 bits is never needed for posterization suppression, it's needed because the dynamic range between saturation and S/N ratio one is close to twelve stops at the lowest ISO's.
[{POST_SNAPBACK}][/a]

I am not aware of such a relationship between the S/N and posterization, but doubt very much that 256 levels are needed in the brightest f/stop whereas only 70 levels are needed according to the Weber-Fechner law. Look at Norman Koren's link and let me know what you think.

Of course, the number of levels needed to prevent posterization depends on the gamma and fewer levels are need at a gamma of 2.2 than one of 1.0. It is generally accepted that 8 bit gamma 2.2 images are sufficient to prevent posterization in a typical reflection print with a 100:1 luminance ratio.

[a href=\"http://www.normankoren.com/digital_tonality.html]Norman Koren[/url]
Logged
Panopeeper
Sr. Member
****
Offline Offline

Posts: 1805


« Reply #19 on: April 18, 2008, 07:36:39 PM »
ReplyReply

Quote
You have missed the point.  The number of levels is irrelevant, as long as the noise fluctuations exceed 1.3 levels.  S/N is the ONLY quantitative justification for ETTR
I think you see this only from the point of dynamic range and ignore other concerns of real life raw image procressing. When the raw values are increased, more raw levels are required to maintain continuous mapping. Such increase can happen already simply due to the WB application and adjustment specifically aimed at redistributing the levels increases the tendency to posterization.
Logged

Gabor
Pages: [1] 2 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad