Histograms

[Tim_Lookingbill]

All I know is when I edit shadow detail using PS's curvesin a gamma encoded DSLR jpeg

that depicts a scene with equal amounts of shadows as mids and highlites, I have about

two square sections off of 000RGB or the equivalent of about 1/10 the curve edit area to

edit all that detail.

 

When I edit scanned negatives depicting similar scenes I have a lot more area along the

curve (almost a third) to pull detail out of the shadows.

 

Explain that.

[bernardwest]

<i>An exposure scale has been substituted for a tonal scale</i><br><br>

 

Tom, they are directly related, as we are dealing with a LINEAR sensor. Read all the posts, I think it has been explained pretty well. Like I said for Van, do the test yourself and you will see that the rest of us are right.<br><br>

 

As for ETTR, no one is saying you have to do it, just explaining that if it is an option it is a way of capturing more data from a scene. If you don't do heavy editing or print very large, then it's probably not worth the trouble.

[tom_t4]

Bernie, we all know in here what a linear sensor is, but there is more that you do not understand. You sure have taken Reichmans argument hook/line/and sinker with no supporting documentation. As for the rest of us being right, who are you referring to? There is only you and Van, and occasionally Tim. So you took it until yourself to claim yourself a winner?

 

Bernine, I read a little more today before responding. You admit a lack of understanding of the zone system, which is the foundation of the exposure system (color,b/w, or digital). A dictionary is not a substitute for a textbook. You would benefit more learning it, and placing your highlights with a spot meter in your preferred zone then wasting your efforts with ETTR and staring at the LCD and wondering if you got the shot. Pros are getting their shot with 1 exposure, on 8x10 film, and faster, using a spot meter. Now I understand why both of you completely ignored my previous comments,because you could not refute it. It is obvious your level of understanding is too basic for a continued discussion.

 

Enjoy guys, I got better things to do then get into this cycle. We have two camps here, and I am not one accepting the ETTR theory.

[bernardwest]

<i>Bernie, we all know in here what a linear sensor is, but there is more that you do not understand. You sure have taken Reichmans argument hook/line/and sinker with no supporting documentation.</i><br><Br>

 

I provided supporting files and histograms above. Van has convinced himself that there must have been a mistake, and you haven't even acknowledged them. Where's your evidence other than a jumbled paragraph which makes little sense?<br><br>

 

<i>As for the rest of us being right, who are you referring to? There is only you and Van, and occasionally Tim. So you took it until yourself to claim yourself a winner?</i><br><br>

 

I provided a link to a previous discussion where the majority of people explained very clearly why ettr works. Check it out. Also do a search on any popular technical forum such as dpreview, luminous-landscape and here at PN and you will see that Van, yourself and a couple of other flat-earthers are by far in the minority. For heavens sake, Thomas Knoll, Bruce Fraser! Do these names not mean anything to you?<br><br>

 

<i>Now I understand why both of you completely ignored my previous comments,because you could not refute it. It is obvious your level of understanding is too basic for a continued discussion.</i><br><br>

 

Your comments had no focus no theme and made little sense. Try and describe your ideas more fully with some evidence and then we might be able to discuss what it is you are trying to say. As for zones, we've seen in this discussion reference to an 8, 9, 10 and 11 zone system? Which is it? This definitively proves my point that zones are an arbitrary system which can have no relevance in a mathmatically precise argument.

[van_camper]

"Your comments had no focus no theme and made little sense. Try and describe your ideas more fully with some evidence and then we might be able to discuss what it is you are trying to say. As for zones, we've seen in this discussion reference to an 8, 9, 10 and 11 zone system? Which is it? This definitively proves my point that zones are an arbitrary system which can have no relevance in a mathmatically precise argument. Where's your evidence other than a jumbled paragraph which makes little sense?" Bernie

 

Bernie, Toms main point was easy to comprehend. What's so hard to understand about expecting a pattern in the histograms if the same pattern is in the mathematical model that Reichman proposed. If there is any loss of data, or a trend towards declining values, it must show itself in the histogram.Your searching for excuses to discredit, rather then refute it (what you do best).

 

Your inexperience with sensitometry is getting tiring. This is most evident with you refusing to accept the zone system (your knowledge is based on the Kiwi dictionary) and its applications, or the belief that exposure and tonality are equivalent (review my comments on INPUT/OUPUT). The zone system, exposure scale, are inseparable. Whether you like it or not, you're using it. Your trying to argue like a champ, but you?re a newbie at sensitometry. That is apparent.

 

"Van has convinced himself that there must have been a mistake, and you haven't even acknowledged them"

 

Bernie, and you've been convinced without any documentation/proof by Reichman. Why are mfrs not commenting on this, designing it into the new D300, etc? They build them, and know it far better then Knoll or Reichman ,who have about the same talent as the rest of us for building a shoe box camera.

[van_camper]

"All I know is when I edit shadow detail using PS's curvesin a gamma encoded DSLR jpeg that depicts a scene with equal amounts of shadows as mids and highlites, I have about two square sections off of 000RGB or the equivalent of about 1/10 the curve edit area to edit all that detail.

 

When I edit scanned negatives depicting similar scenes I have a lot more area along the curve (almost a third) to pull detail out of the shadows. Explain that."

 

Tim, there is a big differnce in dynamic range between film and a dslr.

[bernardwest]

Van, since Tim is AWOL and you seem to understand what he is saying, perhaps you could explain these sentences of his:<br><br>

 

<i>One refers to quantity of light required for a 1 stop increase (doubles), and the other refers to the number of tones remaining after a 1 stop change which doesn't even make sense.</i><br><br>

 

It's this sentence which doesn't make sense. Explain why it is that tones and exposure can't be related. Examples would help your case considerably.<br><br>

 

<i>Also exposure for each pixel is random, not sequential as Reichman describes in his declining table. Each pixel may receive the least or the most exposure depending on the scene.</i><br><br>

 

What on earth does this mean? How is a pixel's exposure a random thing. It 100% deterministic. It's based 100% on the scene, apeture, Tv, and the sensor. What's a "declining table"? What does the "scene" have to do with ettr theory.<br><br>

 

<i>The right side of the histogram should show the highest peaks if you believe in ETTR, yet they do not.</i><br><br>

 

Where does this kooky idea come from? Why? Examples, picture, something!<br><br>

 

<i>Since his theory employs a pattern (a descending order of data)</i><br><br>

 

What's a "descending order of data"?<br><br>

 

<i>the zone system, which is the foundation of the exposure system (color,b/w, or digital).</i><br><br>

 

The zone system may be the foundation of exposure <b>measurement</b> but it's not of exposure itself. Photonics is the foundation of exposure.<br><br>

 

What strikes me the most about this argument, is that you and Tom won't do the test yourself! I've done it, and fine, you don't believe my results. Well let's see some of you test results! Somehow, I don't think they will be forthcoming....

[Tim_Lookingbill]

Bernie,

 

I haven't left this thread, yet. I keep checking back here for any further enlightenment and

also a better way to explain the significance of ETTR for RAW.

 

As Van already admits DSLR sensors capture scenes with less dynamic range than

negatives. But I think where this discussion gets confusing at least for me is how Van

equates a zone system of exposure with the way linear sensors record and count photons.

 

Zone system was a scene rendered film concoction for allowing exposure to predict where

the detail would be mapped and seen along the entire tonal scale of an already scene

correct gamma encoded capture.

 

You and I are talking about how the electronics within a photon counting sensor device

captures a scene which happens to be linear-NOT GAMMA ENCODED.

 

I'm still not sure what if this is really the issue because everyone is using obscure technical

terms that don't explain physically what's happening.

 

I know that ETTR works and the brighter I can capture a scene so that my shadow detail

doesn't end up being rendered with as few RGB levels as possible (as descibed in my jpegs

above) without clipping highlites the more dynamic range I have to work with.

[bernardwest]

<i>Zone system was a scene rendered film concoction for allowing exposure to predict where the detail would be mapped and seen along the entire tonal scale of an already scene correct gamma encoded capture. You and I are talking about how the electronics within a photon counting sensor device captures a scene which happens to be linear-NOT GAMMA ENCODED.</i><br><br>

 

Tim, you're absolutely spot on with this observation. Last night I did a bit of reading on sensitometry and on sleeping on it, what you've just said is exactly what I was thinking when I woke up this morning. Everything I read about sensitometry involves the measurement of density on film, which itself is determined by exposure AND DEVELOPMENT(~gamma encoding). This, I'm almost certain now, is where VAn and Tom are going wrong. As you say, linear raw data undergoes no development and therefore no gamma encoding (or you could say it applies a gamma of 1). Therefore, sensitometry and zones really have no relevence to digital, let alone ETTR theory.<br><br>

 

So let's hear Van and Tom address this point specifically. Does sensitometry involve exposure AND development (gamma encoding)? And if that is the case, then how can sensitometry relate to digital capture? Let's have it then!

[Tim_Lookingbill]

I'll provide a visual I hope will explain what I just said.

 

The image below is an attempt to capture a wide dynamic range scene with a gamma

encoded incamera jpeg taken with my Pentax K100D. Since the oak tree was the main

subject in the scene, I exposed to the LEFT to preserve shadow detail according to the

incamera histogram so I could post process later.

 

The insets below are 200% previews with the first having a brightening curve applied to

the shadows to show pixel detail. The second inset shows how much more midrange pixel

detail=(more photons counted) I had to work with over the shadow detail.

 

If I shot in RAW I'ld have to expose to the right working with the same histogram. And in

doing this I'm pretty sure because of the extreme dynamic range I'ld probably still get

blown out clouds in the RAW converter but I'ld have far more levels to work with to make

up the shadow detail in the tree.

 

Not sure about that last bit though. Haven't shot a similar scene in RAW, yet.<div></div>

[bernardwest]

Tim, i think ETTR is best applied when the dynamic range of a scene you want to capture is less than that which the camera can record. That way you can overexpose a scene without suffering clipping in any important highlights. In the example you posted, you would have to make a choice what was most important to your image - shadows or highlights. Obviously the tree was the focus, but are you willing to wear a totally clipped sky? If the answer was no, then you have no choice but to expose the shot as you did and bring the shadow detail up in post. This is obviously where raw is an advantage to jpg, as you have 16x more levels to play with (in a 12 bit raw capture). The best approach would be something like Guillermo's approach over at LL, and take two shots 4 stops appart and blend in post.

[van_camper]

"Since Tim is AWOL and you seem to understand what he is saying, perhaps you could explain these sentences of his."???Bernie, Toms business is his, but I can see you're having problems understanding a few basic issues. Anyways, now all of a sudden everything is Latin, and you have your big list of things you do not understand. I think what Tom said was very clear (and Tim had no complaints). Looks like now your looking for something else to argue about?grammar in this case.

 

"Since his theory employs a pattern (a descending order of data). What's a "descending order of data?" ???.I assume you have studied the table Reichman proposed, I think the pattern (reduction of tonal values by ? with each stop) is pretty clear? Toms concept that a pattern should exist between Reichmans table and a histogram is a legitimate question. Do you think you can answer it now that I explained it?It is at the grass routes level of our ETTR issue? EXPLAIN why it should or should not be visible in the histogram!

 

"What strikes me the most about this argument, is that you and Tom won't do the test yourself! I've done it, and fine, you don't believe my results." ........ Why waste the time and effort, you never looked at my chart from 2 weeks ago. Besides, Reichmans website has some pretty histograms to look at.

 

"Explain why it is that tones and exposure can't be related. Examples would help your case considerably."?.. Shees, I have explained it at least 10 times (look under INPUT/OUTPUT I had discussed), and how it relates to contrast. Try this, go to Curves in Photoshop, then adjust curves so it is steeper. Now the gamma has changed. For the same exposure you now get a different result (eg- input 100 and output 150 turns to input 1oo and output 200). I can't explain it any better then that (unless you got some beer handy, with pen and pencil and you live close by). Not only must you deal with linearity, but also slope change (gamma).

 

Well, nuff said. Reichmans ETTR is only a proposal. He as of yet not proven it. No data, documentation, just a table he made up because he says "that's the way it is". Looks to me like he left it to us to disprove/prove. Funny guy. If this were a dissertation, it would have been thrown out.

[bernardwest]

Sorry Tim, just realised I claimed you were AWOL. That should have been Tom.

[bernardwest]

<i>Not only must you deal with linearity, but also slope change (gamma).</i><br><br>

 

Van, you're a master at avoiding the issue. Did you miss Tim and my last posts? Why are you still talking about gamma? Gamma has nothing to do with linear raw data and ETTR. Gamma in digital only comes into the picture with rendered raw (ie. jpgs, tiffs etc). So i put the question to you again: "Does sensitometry involve exposure AND development (gamma encoding)? And if that is the case, then how can sensitometry relate to digital capture? Let's have it then!". For the whole of this debate you've been pretty much trying to ram 'sensitometry' down my throat, and now all of a sudden you've got nothing to say on the issue?<br><br>

 

<i>"What strikes me the most about this argument, is that you and Tom won't do the test yourself! I've done it, and fine, you don't believe my results." ........ Why waste the time and effort, you never looked at my chart from 2 weeks ago.</i><br><br>

 

Of course I looked at it, but i've explained a number of times that density (and now gamma) have nothing to do with digital. Give me your definition of sensitometry and explain how gamma and development relate to digital. As for "why waste the time and effort", it's clear why you won't do it because you are uncertain whether you are right or not. I know exactly what the outcome will be. I'll even post some more data soon for you to ignore again.<br><br>

 

<i>EXPLAIN why it should or should not be visible in the histogram!</i><br><br>

 

The histogram directly represents the scene in front of the camera and how that interacts with exposure settings. Take a photo of two different scenes and you will get two different looking histograms. ETTR doesn't require the histogram to be stacked on the right side. I don't know where you two got this idea from. All ettr means is increasing exposure so that the most saturated thing (of interest) on the sensor occurs further to the right in the histogram (you don't deny that as you increase exposure you move the histogram to the right do you?). Say you've got a scene which shows as linear raw ranging from level 0 to 180. Let's say that there are 10 pixels which register a tone of 180. Increase exposure and what happens? Those 10 pixels will now occur at a higher level (lets say 255). In no way does this mean that the histogram will be stacked to the right. You and Tom seem to be suggesting that the number of pixels registering at 255 will now be some huge number. Where are you getting these pixels from? Infact, due to the effect of reversing quantization errors those 10 pixels will now be spread out over a number lower levels, and level 255 may only show a couple of pixels. Basically by overexposing you are stretching the histogram from 180 levels to 255. Clearly you've got 75 more levels of data now. Look at my two examples from earlier, and you can see the second histogram is just a stretched version of the first one. No gross stacking to the right is needed.

[bernardwest]

Here's the test results i promised. These are shots of my wall defocussed and one stop apart. These are from linear tiffs with no white balancing and no colour profile. The clipping points are indicated in the histogram data (unfortunately you can't see the cursor position to obtain these points, but i can send the original tiffs if you want to check this yourself). Each exposure is one stop apart and you can clearly see that the histogram gets compressed/stretched, and values all reduce by half for each stop of reduced exposure. I've included in the zip file three small jpgs of the tiff (nb. these are wb'ed and profiled) so you can read the exif data and satisfy yourself that there is no error (wishfull thinking on my part??). I can't see how you can keep denying the fact that if you halve exposure, you halve tonal values.<div></div>

[bernardwest]

here's the zip with the small jpgs

[Tim_Lookingbill]

Bernie,

 

Pardon my ignorance and/or failing memory, what did you adjust on your camera to

determine one stop variance? Since there are several adjustments like ISO, shutter and

aperture with the last two controlled by focal range depending on priority setting, I was

never clear or maybe I forgot what to adjust to get a one stop difference in exposure.

 

I'm testing my camera's dynamic range on various scenes and my camera depending on

focal length jumps say for shutter speed from 1/60th to 1/80th skipping 1/70th and the

same happens on aperture in manual or program. I don't want to adjust ISO because of

noise. Or is it whatever the camera picks that's one stop.

 

Great demo image BTW. Really made it more clear what this is about.

[bernardwest]

Tim, they were all shot at f4, with the first at 1/50th, second at 1/100th and the third at 1/200th. I did this in manual as using exposure compensation with aperture priority on the camera was not giving me true one stop differences (ie. it was something like 1/50 1/125 and 1/250).