example of 6 stop push in post

[3kramd5]

Nope. I found plenty of photos where I locally increased exposure to +4 (which in ACR I believe is equal to 4 stops) in order to blow out an inadequately-lit white background, but can't find anything close to 6 where I was trying to retain detail.

 

[EdB]

My attempt at a six stop push. Not sure I'm doing it right.

 

[quod]

Does anyone have an example of an image where they needed a full 6 stop push?

The point behind the multi-stop pushes is that with some sensors (Exmors), there is substantially less decline in image quality when pushing an image when compared to Canon sensors. So, the theory is that you can take a picture at ISO 3200 with 9 stops of dynamic range, or you can take it at ISO 100 with 14 stops of dynamic range and push that image up. The Exmors push shadows very well. I don't think anyone actually does this because if the required push is that substantial, it makes more sense to HDR or composite the shot.

 

[privatebydesign]

Does anyone have an example of an image where they needed a full 6 stop push?

The point behind the multi-stop pushes is that with some sensors (Exmors), there is substantially less decline in image quality when pushing an image when compared to Canon sensors. So, the theory is that you can take a picture at ISO 3200 with 9 stops of dynamic range, or you can take it at ISO 100 with 14 stops of dynamic range and push that image up. The Exmors push shadows very well. I don't think anyone actually does this because if the required push is that substantial, it makes more sense to HDR or composite the shot.

Well that is the theory, in practice heavily lifted shadows lose tonality and look like crap at anything over web size.

It is just maths, if you have x number of tones per stop, as we do in digital imaging be it 8bit, 14 bit or 16bit, if you then spread that one stop of tones out over six stops you have one sixth the tonal range. Those bottom six stops might not have any/much noise, but they look like mud because they don't have the tonality we are used to.

 

[quod]

Well that is the theory, in practice heavily lifted shadows lose tonality and look like crap at anything over web size.

Test an Exmor before making this statement. I'm not saying they are perfect, but they are in a different league than Canon.

Here's a 2-stop push with the 645Z. No degraded IQ to my eye. I get similar results from my A7RII:

http://starvingphotographer.com/pentax-645z-real-world-dynamic-range/

 

[privatebydesign]

Well that is the theory, in practice heavily lifted shadows lose tonality and look like crap at anything over web size.

Test an Exmor before making this statement. I'm not saying they are perfect, but they are in a different league than Canon.

Here's a 2-stop push with the 645Z. No degraded IQ to my eye. I get similar results from my A7RII:

http://starvingphotographer.com/pentax-645z-real-world-dynamic-range/

I print for many photographers and have done everything from iPhones to scanned large format film, not saying this to try to sound like a big shot, I'm not, just saying it for background relevance. I have printed many Exmor images which is why I felt qualified to make the comment.

We were talking about 6 stop lifts not 2 stop lifts. There is a three fold difference (on a linear capture scale) between 2 stops and 6 stops which equates to three times less tones, with a gamma curve applied (which there is) there is a difference in luminence of 64 times. I can lift an iPhone image two stops and print to 18", I know that because I did it this week.

The 'answer' to higher IQ is not less noise where there is practically no signal, the answer is more signal, or more efficient photon to electron conversion and deeper FWC.

So do this test, take a picture of a ball or smoothly curved object with side light such that you get very wide falloff and gradation, expose it optimally. Now under expose by six stops. Now put both images in two windows in PS next to each other after lifting the underexposed image six stops and tell me you can't see a difference.

Now we are in a period of 'good enough' I appreciate that, VHS was technically less strong than Betamax but it was 'good enough', MP3's are 'good enough', heck for the majority of people phones make 'good enough' cameras. But, for me, six stop lifting is far from good enough.

 

[quod]

So do this test, take a picture of a ball or smoothly curved object with side light such that you get very wide falloff and gradation, expose it optimally. Now under expose by six stops. Now put both images in two windows in PS next to each other after lifting the underexposed image six stops and tell me you can't see a difference.

No, I think you should waste your time doing this. In fact, you should do it with your phone because we are in an era where that is good enough, right?

 

[bwud]

Here's an example of a pretty extreme DR squish.

As Private By Design mentioned, with the full push (+4 exposure, +100 shadows), while there is little noise in the shadows, there's also little detail. It just looks flat - there's no depth.

That's because it's a shadowed area - it never had the chance to record significant data. While I can make it brighter without noise dominating, I can't create detail from nothing.

Finally attached is a half-assed attempt to balancing it better: +100 shadows, -.2 exposure, -14 black clipping, which is nowhere near 6 stops.

 

[jrista]

I just can't see myself ever really needing more than this.

That's great. But, if you are saying that no one else should ever need more than this, that's not true.

Not sure about six stops, but I can see cases where others might need more than your example shows.

I agree. It doesn't look like a six-stop push. A few stops, and it looks good, but I wouldn't call it a six stop shadow push.

I recall an example from a couple of years ago provided by JRista, in which he was trying to duplicate a scene of backlit sunflowers near sunset and realized that his Canon (I believe he was using a 5DIII) was simply not up to the task. The difference between his example and yours was the need to maintain maximum detail and decent brightness in the flowers (your scene does not require as much exposure in the grass) and, I believe, the sun was a bit higher in the sky and brighter.

At the time, many were critical of his criticism of Canon sensors (myself included). I now realize that was unfair and I was wrong. (belated apology JRista)

.......thanks...? (Sorry...I'm....honestly not sure if that is real?)

 

[jrista]

Most dynamic range test I have read seems to set the standard of a 6 stop push in relation to good dynamic range. I have really never needed more than a couple of stops, however now that I have the 1DX2 I thought I would give it a try just to see if I could do it. I took this shot of a sunrise and exposed for the sky which made the trees and the remaining landscape black. When I opened the image in ACR I only had to push the exposure .2 stops. I did boost the shadows to 100% and I am sure that is equivalent to a couple of extra stops Which would mean that I had an overall push of what, 2.5 stops? I just can't see myself ever really needing more than this. Does anyone have an example of an image where they needed a full 6 stop push?

Personally I like landscape shots into the sunset. As unfocused said, I tried to do that some years back with my 5D III, and in a single frame it wasn't up to the task:

I did try HDR, however that was very, very problematic around the sun, where the HDR blending simply didn't hold up well and produced funky posterization and other artifacts that I had to spend a good deal of time manually processing out:

I have meant to get back out to those fields and try again, but...never really had the will to bother with the 5D III. Been waiting for Canon to produce a lower noise full frame DSLR to try again...gave up waiting a while ago when I got into astrophotography. Maybe the 5D IV will finally deliver the kind of quality I am looking for...I truly hope so. I've been waiting for it since 2009...

 

[Mikehit]

So do this test, take a picture of a ball or smoothly curved object with side light such that you get very wide falloff and gradation, expose it optimally. Now under expose by six stops. Now put both images in two windows in PS next to each other after lifting the underexposed image six stops and tell me you can't see a difference.

No, I think you should waste your time doing this. In fact, you should do it with your phone because we are in an era where that is good enough, right?

WOW! talk about a misquote.

 

[Aglet]

Most dynamic range test I have read seems to set the standard of a 6 stop push in relation to good dynamic range. I have really never needed more than a couple of stops, however now that I have the 1DX2 I thought I would give it a try just to see if I could do it. I took this shot of a sunrise and exposed for the sky which made the trees and the remaining landscape black. When I opened the image in ACR I only had to push the exposure .2 stops. I did boost the shadows to 100% and I am sure that is equivalent to a couple of extra stops Which would mean that I had an overall push of what, 2.5 stops? I just can't see myself ever really needing more than this. Does anyone have an example of an image where they needed a full 6 stop push?

Personally I like landscape shots into the sunset. As unfocused said, I tried to do that some years back with my 5D III, and in a single frame it wasn't up to the task:

I did try HDR, however that was very, very problematic around the sun, where the HDR blending simply didn't hold up well and produced funky posterization and other artifacts that I had to spend a good deal of time manually processing out:

I haven't seen a "need" for a 6 stop push in post (not including the shadow compression that occurs in the print driver) but I've pushed Exmor files considerably with perfectly usable results so, similar to the one-shot sunset JRista shows was a failure with Canon, was a success with Nikon.

camera jpg

pushed raw

I don't remember the exact amount of push in the shadow areas of this shot but when I was playing with it I could bring out the footprints and sand texture in the lower right without any serious loss of tonality... It just didn't look how I wanted it to for the overall image.


above images were posted in the sunset gallery in 2012, the year I was done waiting for Canon to improve sensor performance :
http://www.canonrumors.com/forum/index.php?topic=8105.msg161888#msg161888

 

[Sporgon]

You guys have gotta be kidding. The originals of those sunset shots are so under exposed it would be unbelievable if I hadn't seen it here.

Keep under exposing the Canon like that and you'll sure get crap shadows. You'll also get crap totality on an exmor type sensor but that seems to be quite acceptable to some.

 

[privatebydesign]

You guys have gotta be kidding. The originals of those sunset shots are so under exposed it would be unbelievable if I hadn't seen it here.

Keep under exposing the Canon like that and you'll sure get crap shadows. You'll also get crap totality on an exmor type sensor but that seems to be quite acceptable to some.

I was hoping it wouldn't have to be me to take on those two bullshit 'examples', thanks Sporgon

 

[Aglet]

Keep under exposing the Canon like that and you'll sure get crap shadows.

that's for sure, and obvious, altho latest generation (7d2 & 80d) have, at least, seemed to do away with the picket-fence FPN that made "lack of tonality" the least objectionable downside to shooting like this with a Canon.

You'll also get crap totality on an exmor type sensor but that seems to be quite acceptable to some.

Are you speaking from experience? i don't recall you giving an exmor body more than a perfunctory test and possibly without altering your workflow.

What's being pushed isn't necessarily coming from the very bottom 1 or 2 bits of exmor info.
There's enough tonal data, that when combined with natural textures in some shots, makes for acceptable output even at large print sizes. Certainly a heck of a lot better than a single-shot Canon image if you try to wring out too much shadow info!

 

[Aglet]

You guys have gotta be kidding. The originals of those sunset shots are so under exposed..

actually, they're perfectly exposed to retain the highlite data adjacent to the sun.

You wanna stack a bracketed shot for waves?.. go ahead.
I don't have that kind of time or patience which is why I chose better tools to acquire cleaner raw data.

Exmor = post-processing bliss vs Canon's constipated raw files for the kind of shooting I do on a regular basis.

BUT - I'm not trying to make you switch systems. You keep using what you're happy with.
I'm just saying I'm MUCH happier since I switched.

 

[bwud]

What's being pushed isn't necessarily coming from the very bottom 1 or 2 bits of exmor info.

Same is true with other sensors.

Regardless, shadows are shadows, and there is a significant lack of detail there as compared to mid tones, for obvious reasons. If you want to record more detail in the shadows, you have to expose for them, which may result in clipped highlights. For a given sensor, you can maximize shadow detail without clipping the other end by ETTR. With Exmor sensors, decreasing noise has probably reached diminishing returns (I have no practical experience with the latest Canon sensors and thus won't make a similar statement).

 

[Larsskv]

You guys have gotta be kidding. The originals of those sunset shots are so under exposed it would be unbelievable if I hadn't seen it here.

Keep under exposing the Canon like that and you'll sure get crap shadows. You'll also get crap totality on an exmor type sensor but that seems to be quite acceptable to some.

I was hoping it wouldn't have to be me to take on those two bullS___ 'examples', thanks Sporgon

+1

 

[Aglet]

What's being pushed isn't necessarily coming from the very bottom 1 or 2 bits of exmor info.

Same is true with other sensors.

Regardless, shadows are shadows, and there is a significant lack of detail there as compared to mid tones, for obvious reasons. If you want to record more detail in the shadows, you have to expose for them, which may result in clipped highlights. For a given sensor, you can maximize shadow detail without clipping the other end by ETTR. With Exmor sensors, decreasing noise has probably reached diminishing returns (I have no practical experience with the latest Canon sensors and thus won't make a similar statement).

that's pretty much correct, there's not much detail to be had in dark shadow areas so if you're pushing them so hard all you get is a posterized mess lifted into lower midtones then it's grossly overdone and other methods should have been considered.
If you get a stripey noisy mess with a small push it likely only means you're shooting some version of Canon that does not do well for this kind of post work.
... which is why I think the "lack of tonality" argument is moot.
If the file had enough shadow detail to do a good push without making a mess then there's no problem. This is easily accomplished with ABC cameras and even some Canon raw files that have enough SNR in the raw file.
"Properly exposed" is more applicable to a Canon raw file when you want to push it. The exmor sensored cameras just have more latitude and less noise.

There's little point re-igniting the flames of this argument. The simple fact is you can shoot the same scene with the same settings with Canon and some Exmor-equipped camera and the latter will provide more shadow editing latitude at low ISO (typically 800 or less). It saves time and can produce acceptable results with much less effort.
And, in the example I provided, there's no "lack of tonality" because there's still plenty of shadow data remaining.

EDIT: The only real merit to a 6-stop push is to show the relative SNR (signal to noise ratio) quality of the file. It's unlikely to be used to that extent for practical or even artistic reasons.

I've occasionally re-hot a scene with a bit more exposure to gain "better tonality" data when I underexposed too far trying to maintain highlite tonality. So even us Exmor users will occasionally bracket a little or sacrifice some hilite data to optimize what we want. It's a just a different tool with different abilities and some of us are happier or more comfortable with that.

 

[9VIII]

Most dynamic range test I have read seems to set the standard of a 6 stop push in relation to good dynamic range. I have really never needed more than a couple of stops, however now that I have the 1DX2 I thought I would give it a try just to see if I could do it. I took this shot of a sunrise and exposed for the sky which made the trees and the remaining landscape black. When I opened the image in ACR I only had to push the exposure .2 stops. I did boost the shadows to 100% and I am sure that is equivalent to a couple of extra stops Which would mean that I had an overall push of what, 2.5 stops? I just can't see myself ever really needing more than this. Does anyone have an example of an image where they needed a full 6 stop push?

Personally I like landscape shots into the sunset. As unfocused said, I tried to do that some years back with my 5D III, and in a single frame it wasn't up to the task:

I did try HDR, however that was very, very problematic around the sun, where the HDR blending simply didn't hold up well and produced funky posterization and other artifacts that I had to spend a good deal of time manually processing out:

I have meant to get back out to those fields and try again, but...never really had the will to bother with the 5D III. Been waiting for Canon to produce a lower noise full frame DSLR to try again...gave up waiting a while ago when I got into astrophotography. Maybe the 5D IV will finally deliver the kind of quality I am looking for...I truly hope so. I've been waiting for it since 2009...

To a certain extent I have to agree with the peanut gallery.
The final HDR shot looks great, but also has the sun mostly overexposed again, if that had been the case in the first shot it might have produced some usable detail in the flowers.
What I really have to wonder is how it would have turned out coming off a 6D. Canon did fix most of the banding, the biggest problem with the 5D series wasn't a total lack of DR, it was how it responded to adjustments within the range it did have.