[biz-engineer]

Dear all,
trying an experiment to achieve better image quality with the K-3 by increasing exposure time, lets say +1 Ev and then decreasing by -1 Ev at RAW to JEG conversion. I expected that the same images would be the same image with less noise for the overexposed image, but I end-up with two images having a different contrast. The overexposed image has less contrast then expected. How do you explain this?

[Adam]

Global exposure adjustments by definition decrease contrast. You'll want to adjust shadows independently of highlights in order to achieve the desired effect.

[ROBEFFY]

Global exposure adjustments by definition decrease contrast. You'll want to adjust shadows independently of highlights in order to achieve the desired effect.

Would you mind elaborating on that? I'll help narrow the request.

If I want to increase contrast, which is something I find I desire at times, I should shoot at the optimum or slightly under exposed. ( I believe that under exposed is better than over exposed from a blown out, no data in over exposed shot )

Afterwards, in Post Processing I can adjust the contrast on shadows to make them pop?

Does that seem like a good approach?

Thanks
Robeffy

[MSL]

Let me offer a simple example that has no bearing in the real world. Your full contrast range goes from 0 (black) to 255 (white). Assume you shoot a properly exposed image so it gives a histogram where brightness from 50-200 show a simple, single peak like an inverted V or U (or something in between those two letters). Now if you underexpose your shot, you may shift the peak so the brightness range now covers 25-175, and if you overexpose the range covers 75-225. In all cases, if I define the contrast as the difference between the lightest and darkest pixels, will be the same.

Now go back to your original image. Lets say that every pixel that was in the 50-75 range you darken, and with more darkening applied to those that were already dark to begin with, so that just those pixels now cover 25-75. You brighten the whites the same way so your 150-200 ranged pixels now span 150-225. Now your total image spans pixels with brightness from 25-225 and your picture will have more overall contrast.

If this doesn't make sense, I'll go find some real histograms to show you.

[UncleVanya]

Let me offer a simple example that has no bearing in the real world. Your full contrast range goes from 0 (black) to 255 (white). Assume you shoot a properly exposed image so it gives a histogram where brightness from 50-200 show a simple, single peak like an inverted V or U (or something in between those two letters). Now if you underexpose your shot, you may shift the peak so the brightness range now covers 25-175, and if you overexpose the range covers 75-225. In all cases, if I define the contrast as the difference between the lightest and darkest pixels, will be the same.

Now go back to your original image. Lets say that every pixel that was in the 50-75 range you darken, and with more darkening applied to those that were already dark to begin with, so that just those pixels now cover 25-75. You brighten the whites the same way so your 150-200 ranged pixels now span 150-225. Now your total image spans pixels with brightness from 25-225 and your picture will have more overall contrast.

If this doesn't make sense, I'll go find some real histograms to show you.

I'm following but curious how do you do that without shooting multiple pics and using HDR type software to create that - I'm getting the feeling that isn't what you mean.

[stevebrot]

I expected that the same images would be the same image with less noise for the overexposed image, but I end-up with two images having a different contrast.

Blame the default import settings for your RAW converter. The extra data per pixel you were aiming for is still there, though you will have to work the curves to take advantage of it.


Steve

[MSL]

I'm following but curious how do you do that without shooting multiple pics and using HDR type software to create that - I'm getting the feeling that isn't what you mean.

No you don't need multiple pictures to this. Rather, you do what @adam suggested. Instead of just brightening or darkening the entire image, which is to a first approximation what overexposing or underexposing a shot will do, you selectively brighten light areas or selective darken dark ones. Using the free program Faststone, you do this by adjusting the highlights or the shadows. Taking my example above, and just focusing on the highlights, you use an algorithm that says that if the Pixel luminosity (PL) > 150, add a value which depends on the current PL. You would want it be zero at PL = 150 so that you don't have a sudden jump in brightness at an arbitrary cutoff and increase a bit as PL approached 255. A real algorithm would be a little more sophisticated than this. Or you could do this by adjusting the curves, which would allow all sorts of selective tweaking.

The bottom line is that to enhance contrast you need to expand the overall contrast (difference between darkest and brightest regions). So going back to my previous example, if I did nothing more than multiply the luminosity by a constant - so every pixel were made 10% brighter than its current value - I'd go from a range of 50-200 to 55-220. So not only would my image get a bit brighter overall, it would also increase in contrast - which is not the same thing as just adding 20 to every pixel which would take the range from 70-220.

[UncleVanya]

No you don't need multiple pictures to this. Rather, you do what @adam suggested. Instead of just brightening or darkening the entire image, which is to a first approximation what overexposing or underexposing a shot will do, you selectively brighten light areas or selective darken dark ones. Using the free program Faststone, you do this by adjusting the highlights or the shadows. Taking my example above, and just focusing on the highlights, you use an algorithm that says that if the Pixel luminosity (PL) > 150, add a value which depends on the current PL. You would want it be zero at PL = 150 so that you don't have a sudden jump in brightness at an arbitrary cutoff and increase a bit as PL approached 255. A real algorithm would be a little more sophisticated than this. Or you could do this by adjusting the curves, which would allow all sorts of selective tweaking.

The bottom line is that to enhance contrast you need to expand the overall contrast (difference between darkest and brightest regions). So going back to my previous example, if I did nothing more than multiply the luminosity by a constant - so every pixel were made 10% brighter than its current value - I'd go from a range of 50-200 to 55-220. So not only would my image get a bit brighter overall, it would also increase in contrast - which is not the same thing as just adding 20 to every pixel which would take the range from 70-220.

Thanks that helps explain a lot.

[ROBEFFY]

No you don't need multiple pictures to this. Rather, you do what @adam suggested. Instead of just brightening or darkening the entire image, which is to a first approximation what overexposing or underexposing a shot will do, you selectively brighten light areas or selective darken dark ones. Using the free program Faststone, you do this by adjusting the highlights or the shadows. Taking my example above, and just focusing on the highlights, you use an algorithm that says that if the Pixel luminosity (PL) > 150, add a value which depends on the current PL. You would want it be zero at PL = 150 so that you don't have a sudden jump in brightness at an arbitrary cutoff and increase a bit as PL approached 255. A real algorithm would be a little more sophisticated than this. Or you could do this by adjusting the curves, which would allow all sorts of selective tweaking.

The bottom line is that to enhance contrast you need to expand the overall contrast (difference between darkest and brightest regions). So going back to my previous example, if I did nothing more than multiply the luminosity by a constant - so every pixel were made 10% brighter than its current value - I'd go from a range of 50-200 to 55-220. So not only would my image get a bit brighter overall, it would also increase in contrast - which is not the same thing as just adding 20 to every pixel which would take the range from 70-220.

Thanks for that clarification!
As a LR user, I just played with the Highlights and Shadows, as well as the Whites and Blacks while viewing the histogram. Both offer increased contrast, the Whites and Blacks provide a more dramatic effect, the Highlights and Shadows offer an increase however in a different way. Not sure I can explain the difference at this time.

More for me to learn and understand. I will have to look into Faststone.

Thanks!

[MSL]

More for me to learn and understand. I will have to look into Faststone.

If you have lightroom you don't need faststone. I'm sure the LR experts around here can tell you the difference between highlights and shadows vs blacks and whites - I've never used the program.
If you can - take an image and then take screenshots of the histogram both the baseline starting point and after each tweak. I think if you then compare two histograms you'll have a much better sense of what each handle is doing.

[biz-engineer]

Hello Guys,
I've read you comments and greatly appreciate to have your suggestions and point of view. I think Steve understood my question correctly, but I'm not sure. Let me give you a bit more info.
I'm currently reading a book from Bruce Barnhaus, chapter #9 about film development and contrast control. He says that:
- changing the development time allow shifting of the high lights but has little effect on the shadows
- changing exposure better allow to shift the shadows.

In the past, with the K-3 , I tried to boost the exposure then decrease levels with the same amount in order to decrease noise, then compared the pictures. Result: yes noise was reduced, but contrast was reduced as well.
Now, while reading the book, I compared again the effect of changing exposure during the shot versus changing exposure during development using the K-3 raw converter. But I expected that adjusting the exposure at raw to jpeg conversion would be producing the same result as adjusting the exposure of the shot. But it's not. So either sensor response non linear (similar to film, sensitivity decrease as light increases) , or the raw converter curve is non linear. If the sensor is not linear in the high light zone (like film) , this is very important , this means the contrast of your base raw image will change with exposure, and of course you can correct later in post processing (as some contributors pointed out here above), but better start with a good raw file. I should share two pictures to illustrate.

[derekkite]

Hello Guys,
I've read you comments and greatly appreciate to have your suggestions and point of view. I think Steve understood my question correctly, but I'm not sure. Let me give you a bit more info.
I'm currently reading a book from Bruce Barnhaus, chapter #9 about film development and contrast control. He says that:
- changing the development time allow shifting of the high lights but has little effect on the shadows
- changing exposure better allow to shift the shadows.

In the past, with the K-3 , I tried to boost the exposure then decrease levels with the same amount in order to decrease noise, then compared the pictures. Result: yes noise was reduced, but contrast was reduced as well.
Now, while reading the book, I compared again the effect of changing exposure during the shot versus changing exposure during development using the K-3 raw converter. But I expected that adjusting the exposure at raw to jpeg conversion would be producing the same result as adjusting the exposure of the shot. But it's not. So either sensor response non linear (similar to film, sensitivity decrease as light increases) , or the raw converter curve is non linear. If the sensor is not linear in the high light zone (like film) , this is very important , this means the contrast of your base raw image will change with exposure, and of course you can correct later in post processing (as some contributors pointed out here above), but better start with a good raw file. I should share two pictures to illustrate.

Google dodge and burn. That is what the film developers called the technique. I think some software provides functions with the same name and same effect.

I have experienced what you describe. But as long as there isn't any over exposed or blown highlights, it is a matter of finding a base curve that brings back the pop. A straight raw to jpeg conversion will be flat and lifeless to our eye, so the tones and luminosity are adjusted based on a curve. Every camera manufacturer has their own, and we can recognize a Canon shot for example. Your software has predefined curves that you can apply, or you can define your own. I'm not familiar with lightroom, but more than likely it provides multiple ways to do the same thing. Contrast, brightness, saturation sliders, or tone curve editors, or selecting from a list of presets. I have found settings that work in most cases and adjust for the specific image.

It may also mean that on the higher exposed shoot you are clipping one of the color channels. Even if that isn't the case the applied effects may need to be stronger.

[dracluff]

Hello Guys,
I've read you comments and greatly appreciate to have your suggestions and point of view. I think Steve understood my question correctly, but I'm not sure. Let me give you a bit more info.
I'm currently reading a book from Bruce Barnhaus, chapter #9 about film development and contrast control. He says that:
- changing the development time allow shifting of the high lights but has little effect on the shadows
- changing exposure better allow to shift the shadows.

In the past, with the K-3 , I tried to boost the exposure then decrease levels with the same amount in order to decrease noise, then compared the pictures. Result: yes noise was reduced, but contrast was reduced as well.
Now, while reading the book, I compared again the effect of changing exposure during the shot versus changing exposure during development using the K-3 raw converter. But I expected that adjusting the exposure at raw to jpeg conversion would be producing the same result as adjusting the exposure of the shot. But it's not. So either sensor response non linear (similar to film, sensitivity decrease as light increases) , or the raw converter curve is non linear. If the sensor is not linear in the high light zone (like film) , this is very important , this means the contrast of your base raw image will change with exposure, and of course you can correct later in post processing (as some contributors pointed out here above), but better start with a good raw file. I should share two pictures to illustrate.

Dodge and burn refes more to making a print than developing a negative which is what the book is refering to from what I can tell. I am not sure you can take the development process and exposure of film and compare it so literally with digital. What he is describing with film involves the action of the developper. When you take an image on film the light interacts with the silver halide particles in the emulsion on the film base, reducing some of the halides on the particle surface to silver. The more light the more particles are turned to silver. Now during the development of a negative the amount of halide that is reduced to silver by the developing process is related to the amount of silver reduced by the exposing light. Highlights are much more dense than shadows (remember this is a negative) therefore it takes more time to develop the highlights fully than develop the shadows fully. For this reason you can vary the contrast of the final negative by adjusting the developing time. If you reduce the developing time the shadows will be fully developed but the highlights will not, thereby reducing the contrast that was initally captured in the scene. Essentially what it sounds like he is describing is to expose for the shadows (over expose the highlights) then develop for a shorter time to underdevelop the highlights (or develop for the highlights). This can bring the contrast back in line with what photographic paper can handle as it typically has a dynamic range smaller than film.

In digital I do not think the same rules apply. For one film does not clip the highlights the same way digital sensors do so there is more latitude to overexpose. Secondly the development process in digital is very different from negative film development. In film the number of halides reduced to silver by the light might be analogous to counting of photons (or electrons produced by a photon) at each photosite. In film, a developper is still required to produce a real image where as with digital the image is there and the 'developing' of the raw file (to me) is more akin to developing + prinitng the negative.

Not sure if that makes any sense (things seem to be clearer in my mind then what I tend to put into words). I can not answer the change in contrast with change in exposure accurately unless your change of exposure is due to changing aperture. Wider apertures may result in a loss of contrast.
Dan

[fgaudet]

I'm currently reading a book from Bruce Barnhaus, chapter #9 about film development and contrast control. He says that:
- changing the development time allow shifting of the high lights but has little effect on the shadows
- changing exposure better allow to shift the shadows.

One important thing regarding this is that the author of the book is talking about developing a negative, not the capture of an image. You will get the best dynamic range (and therefore best contrast) with a properly exposed image.

What the author is doing is dodge and burn (as it was mentioned by derekkite), creating more contrast out of an already captured image.

What you are doing is shifting exposure... Let say you start with 14 stops of dynamic range in your original image, you push it towards the dark in post processing (-1)... you still have 14 stops of DR but the top most 1 stop is pretty much only padding (white) and your darks are one stop lighter... so in your process, not only did you introduce white, you also got rid of blacks and lighten up your shadows... losing about 2 stops of dynamic range (and contrast) in the process

[MSL]

chapter #9 about film development and contrast control.

There is a significant difference between film and digital. Film struggles with the shadows and low light exposures, but will saturate when overexposed without necessarily losing the color - i.e. things get brighter but not blown out. So the response is definitely non-linear and if you look at the color response curves from old film packs you'll see this (I remember these from my film days a couple of decades ago).
Digital is great at recovering shadow regions (when you capture RAW in particular) but overexposure tends to blow out highlights that cannot be recovered.

In my examples above I deliberately set up the histogram to be between 50-200 because I didn't want to deal with the issue of what happens when you get to 0 or 255. Once you hit either extreme there is no going back - those details are lost. The other issue that I really didn't want to get into, is that over or under exposing won't just shift the histogram - it will change its shape too. That can include widening it or compressing it - because neither sensor nor lenses are perfectly linear across all color channels - so that just by changing the exposure you can either increase or decrease contrast depending on how the camera+lens combination respond.