Home · Register · Join Upload & Sell

Moderated by: Fred Miranda
Username  

  New fredmiranda.com Mobile Site
  New Feature: SMS Notification alert
  New Feature: Buy & Sell Watchlist
  

FM Forums | Post-processing & Printing | Join Upload & Sell

  

Archive 2012 · Exposure at Capture (RAW)

  
 
cgardner
Offline
• • • • •
Upload & Sell: Off
p.1 #1 · p.1 #1 · Exposure at Capture (RAW)


Working for years in reproduction I'm very aware of how image tonal values change during the workflow. For example when I did QA for the production of maps at National Geographic in the mid-1970sto get the desired 30%, 40%, 50% and 60% halftone values in the ocean we needed to start with 27%, 36%, 45%, and 56% screen tints. This was determined by starting with 5% - 95% target and plotting how the values changed at every step of the multi-stage workflow.

My approach to digital exposure is the same. As SOP intentionally underexpose my RAW files about 1/3 stop under clipping because I've found that if I don't by the time I make a small 8-bit JPG in sRGB the file would have blown highlights, particularly in the red channel of the skin. When using flash and able to control the tonal range at will with fill, I intentionally over fill the RAW so the shadows are lighter than I want the them in the final images which keeps all the shadow detail further away from the point where noise starts to exceed signal and is seen.

The the process of doing flash testing for tonal range and expose I noticed that by simply exposing per the clipping warning so no solid whites where clipping I rarely got any clipping at the end of my workflow in the 8-bit JPG in sRGB. That makes sense because the clipping warning in camera is based on a JPG the the RAW. In the shadows I judge the exposure based on seeing detail in the playback, and when using flash knowing what ratio of key and fill will fit a full .05 - 2.3 reflected density range to my sensor. In the studio and elsewhere when convenient I use white and black towels as 3D targets and proxies for the full range.

My workflow winds up being the reverse of working with B&W negative where you'd err in exposing for shadow detail and often need to "burn-in" the highlights a bit if over development of the negative caused the highlights on the print to "clip" and wash out. With digital err on the side of underexposing the highlights and overfill the shadows to avoid clipping in the final JPG (not the RAW).

In my PP workflow I create a "Master Edit" copy of the file in CS5 and in it keep the highlights and shadows about 1/3 stop under and over exposed relative to the final eye dropper readings I want. I make the final "trim cuts" after resizing and converting it JPG and sRGB.

I check the JPG in Levels holding down the alt/opt key and clicking on highlight and shadow sliders to see what if anything is clipping. Usually the 1/3 stop allowance in the 16-bit ProPhoto RAW is enough that no tweeks are needed, by I will move the end point sliders in Levels to adjust tonal range as needed for screen files and move the middle slider back and forth off 1.0 to see how that affects the appearance.

Adjustments for prints are based on on screen appearance but how the shadows and highlights turn out on the prints, which varies depending on whether I print on my ink jet or at Costco. The same is true for mid-tone contrast. I find that a file that looks perfect on screen will tend to print darker. The amount that is noticed varies with the content of the photo. I compensate for the gain by moving the middle slider in levels left to lighten the midtones. If printing a file in several sizes I make copies scaled to size at printer native (printhead spacing) resolution applying USM and other tweeks as need for that size.

The poll is because I'm curious how many others take these workflow changes into account. I often see clipped red channels in people shots I C&C and scenics that lack a full range of detail in the JPGs posted.



Feb 01, 2012 at 07:50 AM
Alan321
Offline
• • • • • •
Upload & Sell: Off
p.1 #2 · p.1 #2 · Exposure at Capture (RAW)


I too generally try to avoid burning out the highlights at capture (whether whites or individual colours) and otherwise make it as bright as possible. The reason is that I like to keep colour in the sky and on other bright parts of landscape images, and it is easier with my D3s to recover details from dark shadows than from burnt out highlights.

Let me mention something that very few people do - what exactly is "clipping" ? The histograms you see on the camera and the exposure warnings you see on the camera are all affected by the in-camera settings of contrast and saturation and sharpness even if you shoot raw. The warnings you get at a low contrast setting may be too little too late to save the image. At high contrast the preview will look bad but the warnings are very much more evident. Also, there are far fewer pixels on the LCD preview image than in the real image and so it can take a lot of blown image pixels to make even one pixel on the LCD flash an exposure warning.

Also, the extreme left and right of the histograms are highly compressed and isolated bright or dark areas may appear as a line just a single pixel wide at the edge of the histogram box. It can be hard to see and it can be hard to appreciate how much of the image is affected. All those little yellow flowers on the dark green grass might be burnt out and you would not know because individually they are too small on the LCD and make up too little of the histogram. The situation is even worse if they were red or blue.

Another thing that ought to be obvious but is rarely mentioned is that the "expose to the right" rule needs to be accompanied by a "process back to the left" counter-rule. Otherwise you just have an overexposed shot that may have lots of "shadow" detail but is not realistic because the shadows are not dark enough. Even experts get this wrong - especially in tutorials to people who don't know any better.


A camera with greater dynamic range at middle and high ISOs makes life easier because it lets us recover from what would otherwise be lost highlights or lost shadows on a lesser camera. I love my D3s

- Alan



Feb 01, 2012 at 09:13 AM
Eyeball
Offline
• • • •
Upload & Sell: Off
p.1 #3 · p.1 #3 · Exposure at Capture (RAW)


Good points. I particularly take to heart the recommendation to protect the highlight and shadow end-points a little extra before making the final image. I do find myself running close to the edge and then clipping when I do some final sharpening and converting to sRGB. I think Lightroom contributes to this for a lot of people, too, since up until LR4 it hasn't had any soft-proofing capability. LR4 should be a step in the right direction.

I have found myself pulling back a bit from "expose to the right" lately. Even if the highlights are not critical to the image, I think they tend to give the image a very "digital" look when blown. I have also found I don't need to be quite so scared of noisy shadows, even with the "tapestry" noise on my 5D2. It's hard to generalize though - it really depends on the particular image.

Gabor Schreiner developed an excellent program he called "Rawnalyze" that was great for checking true clipping on raw files. I still use it from time to time to check my Canon files. Unfortunately, he passed away sometime back so the program doesn't support newer cameras.

Another thing I played around with a few months ago was "uni-white-balance". It is a technique where you adjust the in-camera white balance so the LCD reflects actual raw clipping (or at least as close as you can get) rather than clipping after color space conversion. On the 5D2 and several other cameras it is as simple as taking an over-exposed shot of the sky and using that blown shot to set the in-camera white balance. Some cameras don't work with this method though and they require some additional effort to balance the RGB channels.

The exercise really brought home to me how difficult it is to blow the blue or red channels in a raw image without blowing the green first. I assume the higher green sensitivity of the sensor (twice the number of green pixel elements compared to R and B) is at least partially responsible for this.

I wouldn't shoot uni-white-balance all the time (it turns the embedded thumbnails green) but I think it can be a useful learning exercise and may be helpful when you want to be particularly careful about blown channels.



Feb 01, 2012 at 10:18 AM
cgardner
Offline
• • • • •
Upload & Sell: Off
p.1 #4 · p.1 #4 · Exposure at Capture (RAW)


Good points Allen.

My strategy it based on knowing I can't trust the camera feedback to tell me what is happening with 100% accuracy which is why it is prudent to err on the side of underexposure by 1/3 or even 2/3 stop in the highlights and overfill by the same amount with reflector or flash in the foreground when possible. I don't ever change my camera style, contrast, etc. so I will have a consistent as possible evaluation via the camera feedback.

In terms of exposure if you take a white card and bracket exposures around the point it clips in the JPG playback in camera then compare the shots frame by frame with RAW files on computer and playback on camera side by side you can train your brain to recognize that clipping in camera warning = 1/3 headroom below clipping in the RAW.

But the workflow doesn't end with RAW. What you should do is take the bracketed series, make 900 x 600 JPGs in sRGB and compare clipping in those files with what you see in the camera and let that guide your interpretation of what the clipping warning of the camera is trying to tell you.

Light angle affects how much of the scene the camera DR can record. If unable to use flash I keep the sun over my shoulder so it highlights 90% of the scene and 90% winds up with "normal" looking detail requiring only minor global tonal range and contrast manipulation in RAW conversion stage. When I can use flash I know via ratio control I can record 100% or even 150% of the detail I see by eye in the foreground, but not in the background beyond the range of the flash which winds up normal in the highlights I've pegged the overall exposure but progressively underexposed by 0 - 3 stops on progressively darker scene content.

Here's a backlit ambient only exposure pegged via highlight clipping...
http://super.nova.org/TP/HSS/_MG_5034_Zones.jpg
I can push the blacks, fill and brightness controls to pull this much detail out of it in ACR..
http://super.nova.org/TP/HSS/_MG_5034_ACRedit.jpg
But in the shadows where the camera didn't record signal all I'm amplifying is the noise. So for something like a portrait or a scenic with foreground detail I will use a three prong strategy: key and fill flash to light the foreground with the same modeling as the natural light before flash is added...
http://super.nova.org/TP/HSS/_MG_5037.jpg
then using layers and midtone manipulation as in the shot above I lighten the ambient background beyond the flash to "normalize" it in midtones and shadows on a duplicate layer and blend the two with masks. Or alternately I will just crop out enough of the background so the underexposure isn't noticed...

http://super.nova.org/TP/HSS/_MG_5035_Cropped.jpg

I do not shoot for specific numbers on the targets, just and overall perceptual impression of a "normal" detail everywhere tonal range the brain perceives and remembers from probing the highlighted detail and adjusting to the details hidden in the shadows it explores in person. The camera can't adapt like eye pupil and filter impression like the brain, but that's what I try to mimic to the extent possible in manipulating the tonal range with flash and PP. The only way to do that successfully is to start with detail everywhere at capture, plus workflow "spoiiage".

In magazine printing we always printed about 2-3% over the required run on every signature in a magazine to allow for makeready and screw-ups during binding. Better to recycle some paper than go back to press to make up a shortage,




Feb 01, 2012 at 10:36 AM
plubbry
Offline
• • •
Upload & Sell: Off
p.1 #5 · p.1 #5 · Exposure at Capture (RAW)


For casual shooting I try to get exposure "right" at capture.

For more serious/careful landscape shooting I will typically ETTR. I will ETTR as far as I can without clipping any of the RGB channels. (Specular Highlights excepted of course).

Since the histogram is generated from the lowly 8-bit jpg file I have neutralized all in camera settings to get the flattest/dullest looking jpg. Furthermore, I have played some with the UniWB in attemps to get even more accuracy out of the RGB histograms.



Feb 01, 2012 at 10:49 AM
RustyBug
Offline
• • • • • • •
Upload & Sell: On
p.1 #6 · p.1 #6 · Exposure at Capture (RAW)


When I tend to shoot a lot of stuff that exceeds the sensor DR, I have to make a choice as to which end I want to protect the most and deviate from "right" accordingly.

If the scene does fit in the DR, then I typically go for "right" because that will give me lattitude on both ends. But there again, if I am favoring one end vs. the other to accentuate in PP, "right" may not be the way I go. If I were shooting more of a production style ... I'd be mostly about "right".

"Old school" says get it right as much as you can in camera, so that you have less problem/effort during processing constraints/issues ... so it is a joint effort depending on your processing approach/plan.

cgardner wrote:
My strategy it based on knowing I can't trust the camera feedback to tell me what is happening with 100% accuracy


This is why I really like my Kodak SLR/C. Although this isn't applicable for most people, it will provide RGB values feedback wherever I place the cursor, so I can readily know exactly how close I am to a blown channel in any portion of the scene. Too bad most other cameras don't do the same.



Feb 01, 2012 at 01:16 PM
Alan321
Offline
• • • • • •
Upload & Sell: Off
p.1 #7 · p.1 #7 · Exposure at Capture (RAW)


I'm past the end of my day and will have to study this thread in more detail later on.

I have done very little printing and therefore have not had too much problem making prints look like the screen image (within reason - the prints never seem to light up a dark room no matter how bright the scene was ). For this I've made things easier by having the screen brightness closely match the reflected brightness of blank printer paper in the room where I hang my prints - about 90 Cd/m2.

I mentioned previously that contrast and other settings affect the exposure warnings. That of course affects highlight headroom in the captured raw files. Some time back when I using a 1D2 (I think) I found that the settings could vary the headroom from as little as 1/3 stop to as much as 4/3 stops, which is quite a range considering the overall dynamic range that can be captured by a sensor. i.e. I could be losing or gaining a stop of DR with no apparent indication at time of capture, just by tweaking in-camera settings.

Uni WB is probably a good thing in some ways but I couldn't stand the green previews when I was trying to see how well the image looked in-camera. I do wish the manufacturers would give us user-selectable raw per-channel exposure clipping warnings, in which we could nominate the percentage of pixels that clipped. It would only take a few little LEDs on the camera, or something in the main LCD screen display. For that matter, even a raw-data histogram on-camera with clear indications of stop boundaries should be a no-brainer. There's no reason why it could not show mid tone and +/- 4 or 5 stops plus sensor limits on a histogram horizontal scale.

I did find the raw luminance histogram in Canon's DPP software to be very useful in the old days when misinformation abounded about what the RGB histograms actually represented. I had no other software that could do that. All that stuff about digital exposures being basically linear (so half the possible tonal values are in the brightest stop and each darker stop got half the values of the previous stop) just did not gel with a jpg histogram.


Something else about capture exposure settings just came to mind. I've read an article that pointed out that the matrix metering on Canon and Nikon cameras varies greatly between the two brands, perhaps by several stops in some situations. That makes it very confusing for those of us using both systems. To get something close to consistent you need to use average or centre-weighted average metering (or maybe spot) and make the necessary adjustments as we did in the old days. This is especially true if you cannot trust the histograms where it matters most - at the highlights and shadows.

- Alan



Feb 01, 2012 at 02:57 PM
matthewbmedia
Offline
• • •
Upload & Sell: Off
p.1 #8 · p.1 #8 · Exposure at Capture (RAW)


Most of the time I will aim for a balanced exposure in camera - and If the situation calls for it, I will actually Pull exposure in post rather than push it to avoid adding noise - so I will overexpose maybe 1/3rd of a stop. I still think the basic zone system principles apply to shooting and processing raw files, but I prefer to pull instead of push as I did with film.

Yes you run the danger of clipping, but the highlight recovery in aperture is so good that I will take it over an underexposed file anyday -



Feb 01, 2012 at 09:09 PM
cgardner
Offline
• • • • •
Upload & Sell: Off
p.1 #9 · p.1 #9 · Exposure at Capture (RAW)


RustyBug wrote:
"Old school" says get it right as much as you can in camera, so that you have less problem/effort during processing constraints/issues ... so it is a joint effort depending on your processing approach/plan.


The question is where should the aim point for "right" be? Right in RAW or right in the worst case format at the end of the workflow the small JPGs. If you shoot for the requirements of the latter you will find the JPGs based playback/histogram/clipping warning while not perfect do a good job of predicting results.

For example if you know you can recover that extra stop of highlight detail in RAW beyond the point clipping shows in playback that will with manipulation produce an OK RAW file. But if you then take that manipulated RAW file and make an 8-bit 900 x 600 pixel JPG for your web page you will often find the highlights blowing as indicated on the camera at capture in one or more channels.

Digital is more analogous to the negative / print process than transparencies because you can dodge and burn with both.

Many think there is no opportunity do corrections with transparencies, but working on the reproduction side I was a aware of several:

I had a 4 x 5 internegative made whenever I made a print from a transparency which allows for corrections of exposure (burning in shadows for greater DR) when making the negative and resulted in less grain from the transparency being visible on the print vs. direct Cibachrome prints.

In the 70s through the mid 2000s when digital files replaced transparencies completely we rarely got an original transparency to drum scan for the magazines we printed. Most were 4x5 duplicate transparencies made from 35mm. As with the internegs for prints you could correct for color bias from WB error or processing, dodge/burn and wound up with less grain in the final scanned image.

In the days before Photoshop catalogs with several images on a page would be duped to final reproduction size and color/exposure corrected so all matched. Kodak made a color duplicating transparency film that allowed the emulsion to be floated off the base of the dupe film and into place on a 1 or 2 - page spread sheet of mylar, which would then be drum scanned. I cut the number of scans (and their correction) by about a factor of 10 and resulted in more consistent images. But even then the scans were to look lighter on proof than they would appear on the final web printed catalog, allowing for dot gain, wet trapping of inks and other press variables.



Feb 02, 2012 at 12:53 AM
RustyBug
Offline
• • • • • • •
Upload & Sell: On
p.1 #10 · p.1 #10 · Exposure at Capture (RAW)


Yes, internegatives are an option for transparencies, my workflow preference was to avoid them as best possible.

+1 @ D&B like neg/print, while +1 @ blown like chrome, +1 @ pull vs. push given option.

Lately, I've been starting with my workflow with a linear gamma @ 32 bit and selectively applying my gamma at both ends and in the middle via layers, opacity and blending ... still experimenting to develop a workflow for "wider than sensor" DR scenes.



Feb 03, 2012 at 12:37 PM
Peter Figen
Offline
• • • • •
Upload & Sell: On
p.1 #11 · p.1 #11 · Exposure at Capture (RAW)


Drum scanning is still the best, fastest and easiest way to deal with color transparencies, as long as you have a great scanner, software and, most importantly, operator. Same for color negs but the software and operator are even more crucial then.


Feb 03, 2012 at 01:07 PM
Alan321
Offline
• • • • • •
Upload & Sell: Off
p.1 #12 · p.1 #12 · Exposure at Capture (RAW)


cgardner wrote:
The question is where should the aim point for "right" be? Right in RAW or right in the worst case format at the end of the workflow the small JPGs. If you shoot for the requirements of the latter you will find the JPGs based playback/histogram/clipping warning while not perfect do a good job of predicting results.

For example if you know you can recover that extra stop of highlight detail in RAW beyond the point clipping shows in playback that will with manipulation produce an OK RAW file. But if you then take that manipulated RAW file and make
...Show more

Doesn't this come down to how you make that small jpg for your web page ? i.e. the extent to which you are automating the file conversion versus carefully tweaking it manually all the way. I haven't specifically tried the example you suggested but I don't see how you can avoid the problem without manual tweaking. e.g. lets say you take your raw capture ignoring a certain amount of exposure warnings because you "know" you have headroom in the raw data vs the jpg preview - but you can only recover that headroom if you manipulate the raw to jpg conversion differently from whatever settings were used in camera to produce those warnings. If you use the same settings then of course you'll get the same overexposures. If you do change the conversion parameters and do produce a raw to jpg conversion that looks ok, but then do a downsized web version that gives clipped highlights then it can only be because the for-web conversion has been different - or because you inadvertently applied some jpg parameters other than simple downsizing when you converted from the ok full-size jpg to a smaller jpg (extra contrast or saturation or sharpening, for example).


- Alan



Feb 05, 2012 at 04:53 PM
cgardner
Offline
• • • • •
Upload & Sell: Off
p.1 #13 · p.1 #13 · Exposure at Capture (RAW)


The point I'm trying to make is that one can't avoid the changes from 16-bit RAW > 16-bit ProPhoto edit > 600 x 900 pixel 8-bit sRGB JPG in the workflow even with careful manual control of the conversion.

But once you know how much the tonal range changes and highlights clip do to the PP workflow you can COMPENSATE for them by underexposing the RAW by 1/3 to 2/3 stops below what seems optimal.

In other works the goal shouldn't be an optimally correct RAW capture with the solid whites at 254 but rather to record the solid whites at around 245 in the RAW, anticipating they will wind up at in the JPG 254 without doing anything to change them.



Feb 05, 2012 at 05:26 PM





FM Forums | Post-processing & Printing | Join Upload & Sell

    
 

You are not logged in. Login or Register

Username       Or Reset password



This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.