Precise Digital Exposure

[AlterEgo]

painting

good one

[bernie west]

you fail to consider that it is not my description - it is how processing was explained (more then once) by developers, that is exposure correction stage (pull/push/leave as is) in ACR/LR is after demosaick, wb/color transform (dcp has several parts, matrices /CM, FM/ and HueSatMap LUTs are applied before exposure correction)
 

Well it is your description here in this thread.  Do you understand it, or are you just parroting something without understanding it? 

If acr/lr is working in 16bit then when white balance is applied to a nominally over-exposed image a whole bunch of pixels will max out at 65535.  Pulling back the exposure slider will be unable to create detail out of that blown data.  It will only be able to render those areas as some shade of grey.  So either you are wrong, or acr/lr are working in a larger bit-depth than 16bit.  If that's the case, then you've got to explain how it is that both DCRAW and LR render nominally blown highlights the same way.  DCRAW I'm pretty sure only works in 16 bits (well, it did when I last played with this stuff in 2009 or so).  Can you explain all this, or are you uninterested in the details of your beliefs?

[BobD]

...I would not advise using LR/ACR... especially with the current process version (PV2012)...
...the BaselineExposure... increase or decrease the rendered values by the amount of the baseline adjustment.
...Most light meters (including those built into our cameras) are calibrated to yield 12% saturation rather than the 18%

I understand what you are saying about LR's adaptive technology and I try to debunk the 18% myth by referencing Thom Hogan’s article “Meters Don’t See 18% Gray” in my book.  However, what I learned from the Zone System was to understand AA’s procedure then empirically apply it to my personal “zone system”… my Exposure/Development; my +/-N develop times; my cold-light enlarger; etc. 

The keyword here is “system”… an empirically defined personal system. Not in a theoretical scientific realm… but real world photography… “your” real world photography! (Your camera, Your raw software, Your workflow.)

Photographers are supposed to take pictures not run tests and be scientists.

The goal in the analytical labs at Polaroid was for scientist to refine processes THEN develop procedures (systems) to be used by technicians. Many technicians might not understand the “why” of the procedures but were good observers and craftsmen. The real good craftsmen grew to become associate scientists.  Not unlike a “picture-taker” becoming a “photographer.

My method of exposure is intended to empirically test “our personal system” then get us photographers out shooting and not running endless test.

[digitaldog]

Photographers are supposed to take pictures not run tests and be scientists.

Sorry but I don't see how the two are mutually exclusive.
When I was shooting film as a pro, I always ran tests. Film tests for each emulsion, working with my E6 lab for ideal CC filtration to further balance. Testing stobes, differing ways of printing. The statement is one useful to those who want to suggest photography is just about pointing a camera and capturing a moment. That's a large part of photography but a large part, perhaps not equally so IS running tests, and understanding the science behind the process. I don't know it's much different from a painter. Painters are supposed to paint and not mix let alone create their own pigments. I'm not a painter, but I think not.

In terms of taking pictures, a photographer I respect once told me: Great pictures are made, not taken.

[AlterEgo]

Well it is your description here in this thread.  Do you understand it, or are you just parroting something without understanding it? 

If acr/lr is working in 16bit then when white balance is applied to a nominally over-exposed image a whole bunch of pixels will max out at 65535.  Pulling back the exposure slider will be unable to create detail out of that blown data.  It will only be able to render those areas as some shade of grey.  So either you are wrong, or acr/lr are working in a larger bit-depth than 16bit.  If that's the case, then you've got to explain how it is that both DCRAW and LR render nominally blown highlights the same way.  DCRAW I'm pretty sure only works in 16 bits (well, it did when I last played with this stuff in 2009 or so).  Can you explain all this, or are you uninterested in the details of your beliefs?

again - you fail to comprehend that those are not my "beliefs" - I am merely repeating you (as you can't find out yourself) what more than once was said by Adobe developers (E.Chan) about where the exposure happens in ACR/LR pipeline in terms of operations with the data  


if you are interested in the data types that Adobe are using internally during various operations then you probably need to see not dcraw.c code, but rather Adobe's DNG SDK code (available publicly) - that might give you an idea (because they use that code in ACR/LR).


if you are interested to find out how Adobe achieves the sequence of operations they disclosed (demosaick -> ... -> wb/color transform till/including HueSatMap tables -> ... -> exposure ->... ) w/o losing any necessary data please feel free go and dig there, but that's not something I am interested in here, sorry, so don't drag me into "bits" discussion... or you can chase E.Chan and challenge his description about the sequence of operations with the data in ACR/LR (that is, if I understand you correctly, that exposure operations /relevant UI sliders/ in ACR/LR code must be done before WB as you, apparently, want to say, not after, as he says)... here I want to include again one his many quotes = ""the DNG processing model performs a linearization of the original raw image values followed by demosaicing, then white balance. All of the other image stages follow. So to answer your question, all of the image ops except for linearization (which isn't under user control anyways) happens after demosaicing. "

[BobD]

Andrew,

"but a large part, perhaps not equally so IS running tests, and understanding the science behind the process."
My comment was never intended to mean that becoming a photographer meant you no longer have to practice and keep learning your craft!

"Great pictures are made, not taken."
As is the learning process throughout our life, we tried to simplify our lives.
I feel this is the same for us photographers… To simplify our photographs

When I do my photography talks, I try to emphasize that…
..."Our goal is to not take snapshots... but, to make snapshots.

[digitaldog]

"but a large part, perhaps not equally so IS running tests, and understanding the science behind the process."
My comment was never intended to mean that becoming a photographer meant you no longer have to practice and keep learning your craft!

And IMHO that involves testing, some science and of course, 'taking' pictures.

[bernie west]

again - you fail to comprehend that those are not my "beliefs" - I am merely repeating you (as you can't find out yourself) what more than once was said by Adobe developers (E.Chan) about where the exposure happens in ACR/LR pipeline in terms of operations with the data  


if you are interested in the data types that Adobe are using internally during various operations then you probably need to see not dcraw.c code, but rather Adobe's DNG SDK code (available publicly) - that might give you an idea (because they use that code in ACR/LR).


if you are interested to find out how Adobe achieves the sequence of operations they disclosed (demosaick -> ... -> wb/color transform till/including HueSatMap tables -> ... -> exposure ->... ) w/o losing any necessary data please feel free go and dig there, but that's not something I am interested in here, sorry, so don't drag me into "bits" discussion... or you can chase E.Chan and challenge his description about the sequence of operations with the data in ACR/LR (that is, if I understand you correctly, that exposure operations /relevant UI sliders/ in ACR/LR code must be done before WB as you, apparently, want to say, not after, as he says)... here I want to include again one his many quotes = ""the DNG processing model performs a linearization of the original raw image values followed by demosaicing, then white balance. All of the other image stages follow. So to answer your question, all of the image ops except for linearization (which isn't under user control anyways) happens after demosaicing. "

I'm not questioning what Chan says, all I'm saying is that there's a chance that you might not have understood what he was saying.  In the end, it doesn't really matter, as it is clear that LR can pull back nominally clipped channels in the rendered data, just like all raw converters can.  Whether it does this before the WB or after with a larger bit space, doesn't really make much difference. 

[BobD]

Andrew,

We are on the same "wavelength" on this - trust me.
We all need to be a little more empirical by testing what we read - not just doing what we read. (see my examples below).
Bob

1. I plotted many bracketed exposures many time testing the top 2 stop occur in the top 10% of our exposure. (Yes, I read George Jardine's" article about revisiting the Zone System a while back explaining this... but, I'm an empirical kind of guy... I need to test it on "my" system)


2. When testing the dynamic range of my camera sensor using Sekonics DTS software and the L758, I determined that
> the dynamic range of my camera's sensor was different with a camera raw captured saved in AdobeRGB1988 space
> rather than shooting a camera "jpg" shot in AdobeRGB1988 space (as recommended).

[digitaldog]

Andrew,
We are on the same "wavelength" on this - trust me.

I know we are. I'm just not as comfortable as you are with the statement: Photographers are supposed to take pictures not run tests and be scientists.
It runs counter to everything you've written prior to that text, that's all.

[AlterEgo]

I'm not questioning what Chan says, all I'm saying is that there's a chance that you might not have understood what he was saying.

well, suggest an alternative explanation to a very clear statement (and not only one, but just one out of many) that WB follows right after demosaick and all other things later...  E.Chan is a technical person, not from marketing department, so I have no reason to question that direct statement... now, again, if you curious, how they manage to do this - DNG SDK code has C/C++ data types for you to study what they do as 16 bit signed/unsigned integers, what they do in 32/64 and float... or it seems that with ACR9/LR6 being released he is out here (I saw him replying today on Adobe forums) - so ask him

[Anders_HK]

1. I plotted many bracketed exposures many time testing the top 2 stop occur in the top 10% of our exposure. (Yes, I read George Jardine's" article about revisiting the Zone System a while back explaining this... but, I'm an empirical kind of guy... I need to test it on "my" system)

Bob,

I politely view your adoption of 99% per the above basis as incorrect, for the following three reasons. If I am missing something with your justification, please advise.

1) I do not see basis for why to choose 99%.

2) A picture is not mere based upon optimum white, but is important based on textures and tonal values, which with digital colour photography for the latter relates to R, G, and B.

3) Please refer to the attached to what Ansel referred to as Zone System Ranges. Contrary to there, today we typically call Dynamic Range in digital what he referred to as Full Black to Pure White. What he referred to as Dynamic Range is in fact Tonal Range. Additionally there is Texture Range. I will argue that what is correct to base the exposure upon towards the ends of your sensors exposure range, are actually the last texture or tonal values respectively before clipping of channels (before and after recovery settings). Firstly lost is the texture, later after are the tonal values in each channel R, G, B which does not necessarily occur at same time). If I recall correct, I determined my 2 1/3 upon upper point without recovery and without clipping any channel where I could still view texture, while 3 1/2 the last channel before clipping with max highlight recovery that I could still clearly view tonal value.



Anders

[ErikKaffehr]

Hi,

Regarding shadows, a proper ETTR exposure is always optimal. It is simply the maximum exposure that avoid clipping non specular highlights.

I would suggest to check the raw files with a tool showing a proper histogram, without manipulation in the raw processor. What I have found on my P45+ and also on the Sony Alpha 99 I have is that blinkies and pretty accuarate indications of clipping.

Here is an example, taken from my latest outdoor shoot: http://echophoto.dnsalias.net/ekr/Articles/ETTR2/20100721-CF046421.iiq

This exposure is slightly overexposed, which is quite obvious in RawDigger:


Lightroom silently applies some highlight recovery:


Capture One correctly indicates clipping:


Reducing exposure in C1 fixes clipping (by highlight recovery):


Now, in this case I looked at the blinkies and histogram on my P45+, and decided that correct exposure was 0.5 stops down. So the final exposure had a raw histogram like this:


Now, I could have used my Minolta Spotmeter F to adjust highlight exposure to +3 EV and gotten the same results. Using the histogram, I need to make an additional exposure, of course. Also, the spot meter can give indications how dark the darks are.

More modern cameras often have "zebras" which give adjustable limits on exposure and can be used with live view.

So to sum it up:

ETTR is always right, but it needs to be decided what is OK to clip. Blinkies are helpful in that regard.

Histograms shown in the raw converter always show manipulated data. It is therefore better to analyse raw histograms like those produced by RawDigger.

This is a decent example of an image with relatively high dynamic range:



Note that the disc of the sun is clearly visible:


But the near silhouette of the rowers can be made quite clean:


The raw histogram of this image indicates around ten stops of luminance range:




Here is a subject with larger luminance range, the Alpha 99 SLT reproduces highlights decently, but also gives good details in the darks:



In this case the P45+ had somewhat less exposure, based on camera histogram:


And the P45+ straggled in the shadows:


With the P45+, a spotmetered exposure may have resulted in a better exposure. Around one stop difference between SLT 99 and P45+.

A HDR merge using Lumariver HDR fixed he shadows issue on the P45+:


Best regards
Erik

Lets get back on subject…

Where to locate the maximum of photons collected, and how ?? ?
Ensure shadows will be adequate collected ?

EXAMPLES:

How to expose the attached scenes precisely; Ziczac Bridge, View from Resturant, Bierstadt (painting by)?

Consider exposing for allowing in processing for an extended shoulder transition into highlights of around 1/3 stops more than available highlight recovery (Velvia slide film, page 19-20 in my paper), and for maintaining as much quality pixel info as possible in the "important" parts of image.

I can spot meter for (A) or (B) above, or if I want to expose per knowing more for any of remaining in above. After, all I need is one shot. To add, although my Leaf back has an excellent RAW histogram, due 80MP it is slower to use than a DSLR and more battery hungry.

Anders

[ErikKaffehr]

Hi,

Apart from the discussion about exposure determination, the images that Anders HK has posted are both really good and quite illustrative.

Best regards
Erik

[bjanes]

Apart from the discussion about exposure determination, the images that Anders HK has posted are both really good and quite illustrative.

I agree that Anders' images are lovely, but I think that attempts to adapt zone principles to digital are misdirected since film and digital sensors have a very different response to light. Digital is linear whereas film responds proportionally to the logarithm of exposure. The linear part of the H&D is linear only when plotted on a log-log scale (log exposure vs log(1/transmission). Figure 10.6 in this link demonstrates the differences between film (in this case radiographic film) and digital.

The s-curve of the raw converter response does not correspond to the toe and shoulder of the H&D plot of film exposure.

Bill

[digitaldog]

I agree that Anders' images are lovely, but I think that attempts to adapt zone principles to digital are misdirected since film and digital sensors have a very different response to light. Digital is linear whereas film responds proportionally to the logarithm of exposure.

That's always been my understanding and belief too so it's good to hear confirmation on this.

[hjulenissen]

I just found the article that this image seems to stem from:

"An objective protocol for comparing the noise performance of silver halide film and digital sensor", Frédéric Cao, Frédéric Guichard, Hervé Hornung, Régis Tessière
https://www.dxo.com/sites/dump.dxo.com/files/dxoimages/ei/sci-publications/2012%20Film_vs_Digital_final_copyright.pdf

[bjanes]

I just found the article that this image seems to stem from:

"An objective protocol for comparing the noise performance of silver halide film and digital sensor", Frédéric Cao, Frédéric Guichard, Hervé Hornung, Régis Tessière
https://www.dxo.com/sites/dump.dxo.com/files/dxoimages/ei/sci-publications/2012%20Film_vs_Digital_final_copyright.pdf

Thanks for the link to a very good article by the scientists at DXO. The plot is for SNR, not the characteristic curve, but is nonetheless interesting.

I downloaded the graph for the characteristic curve of a hypothetical film and used Photoshop to measure the values on the axes and interpolate the Y-axis values for X-axis values and converted from density to opacity in Excel (opacity = 10^density) and plotted the values in Excel. The plot shows a logarithmic response as expected. Digital is linear as the article states.

Bill

[Anders_HK]

... a proper ETTR exposure is always optimal. It is simply the maximum exposure that avoid clipping non specular highlights.

I would suggest to check the raw files with a tool showing a proper histogram, without manipulation in the raw processor.

Hi Erik,

Apologies for a belated reply. I agree that what you state is the theory. However, in practicing photography we can be faced with situations where similar as the example with the sun you posted, there is less haze (stronger light from sun), whereby we need decide to not ETTR on basis of the strongest light within the sun. This is because we else will not get sufficient data for mid tonal range of a picture, and because the collection of optimum high values within the dynamic range within sun is not of most important.

"a tool showing a proper histogram" - My Leaf back show me a RAW histogram, including extended ends for what in principle can be recovered. I also do rely on the histogram in Capture One (including at my choice for defaults. However, the important to my image is what results I can (have capability to) achieve in a processed image (no recovery and/or towards full recovery). That will be my picture. Therefore is why in my spot metering I consider both the RAW values at extremes, and at what spot metered values are beyond these to the maximum I can recover in Capture One.

Further, when studying test images of a grey card per viewing them in Capture One, at both extremes firstly lost will be the texture, second the tone, and third clipping. For my image, I wish to control on marginally safe side to where in image that texture and tone extends at defaults in Capture One (default RAW values) and at maximum recovery. That will help me define exposure for how I wish optimum basis RAW for my later processing towards a picture. This is how I precise can visualise for assuring of captured dynamic range versus that of a scene.

So why do I not simply rely on my Leaf histogram fully? For situations where I demand full control for precise allocation of capable dynamic range of my sensor (with and without recovery in post) per pin pointing spot metered values in a scene.

Best regards,
Anders