October 23, 2014, 02:59:36 PM

Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.


Messages - raptor3x

Pages: 1 [2] 3 4 ... 14
16
@ Sporgon

it was already mentioned that DPP is better at develop cr2 raw expecially in the shadows, from this point your examples of the blue boat is very interesting even more with no noise reduction ... In the end if not better sensor Canon could realise a better/powerful software or why not help Adobe improve the CR2 raw converter ??? Their DPP is free, they do not need to protect sales in this regard, maybe it is the opposite, there are many adobe customers that could get more out of canon so no need to search elsewhere.

That is debatable. DPP shows less banding and noise in shadows now, but it also gets a very mushy, sometimes almost posterized, very fake, digital, waxy, no detail look IMO. Personally I'd call that putting lipstick on a pig.

Yeah, especially with the new version of DPP even the lowest non-zero level of noise reduction is very heavy handed.

17
EOS Bodies / Re: Just for Jrista: 2014 Market Data
« on: September 26, 2014, 07:45:38 PM »
So it seems the reason I've been seeing posterization issues and Sarangiman hasn't when doing the emulated high ISO trick is due to Sony's compression algorithm.  Good to know I'm not the only person seeing this.

18
EOS Bodies / Re: More Talk of an October Announcement of a DSLR [CR1]
« on: September 26, 2014, 07:15:00 PM »
Then a funny thing then happened.  All of a sudden I could not stand looking at the 60D files with anything over ISO 800.  The truth is even ISO 400 is quite bad compered to a NEX 6 at ISO 400.  I was never quite satisfied by high ISO performance but the difference was eye opening.

Huh?  Even at ISO 6400 there's almost no difference between the two.  If anything, the 60D is a bit better once it's downsampled to the same resolution.


19
EOS Bodies / Re: Just for Jrista: 2014 Market Data
« on: September 25, 2014, 02:10:53 AM »
Er, I'm going to have to beg to differ. Here's the D810 at ISO 6400 and at ISO 100 pushed 6 EV:


Can you tell which is which?

Btw, in the ISO 100 file, that road has a signal of average 7 for the green channel, where the SNR is 3.5. Red channels is like 4. These are all on a 16-bit scale. So pixels with signals of literally 4-7 in the Raw file (1 being minimum, ~16,000 being maximum).

So I'm not even talking about pushing a midtone or even a shadow +6 EV. I'm talking about pushing some of the deepest of the deep shadows 6 EV. With brighter tones - you can't tell *any* difference between ISO 6400 and ISO 100 pushed 6 stops. But here even with tones all the way down at the floor of the sensor/Raw file, it's hard to see the difference. If that doesn't wow you...

Point being: yes you can push 6 stops for certain Exmor sensors. Though it does get tricky with ACR, since it's not really built to do that. I had to use the 'Blacks' slider which ends up reducing contrast, and I adjusted until I got the same brightness in the road as the ISO 6400 file.

That's extremely impressive.  I'd guess the one on the left was pushed with some shadow tint correction since it looks a bit green on this monitor but that's very good.  The A7 is much noisier under similar conditions.

20
EOS Bodies / Re: Just for Jrista: 2014 Market Data
« on: September 25, 2014, 01:47:23 AM »
Key word here: implied. You assumed, despite me explicitly saying 'almost' and also pointing out that there are limitations due to quantization error. Which is exactly why I said some posts ago that ISO 400 is the 'magic' ISO above which there's not much benefit to hardware ISO amplification for full-frame cameras, b/c this is where one electron is counted by one digital increment in your RAW file ('unity gain ISO').

That's why I explicitly said you can choose ISO 400 rather than 6400 or 12.8k some posts above, remember?

The way you phrased it implied that you could almost keep the entire (ISO 100) dynamic range at high ISO when the reality is that you can almost keep then entire ISO 100 DR up until ISO 1600 at which point you lose ~1 stop DR per stop increase in apparent ISO (not quite sure how this will actually scale).   But fair enough, it was your comment that prompted me to test using ISO 400 instead of ISO 100 so yes, I do remember.

No, ISO 100 pushed is not always going to be the same as ISO 400, for tones below a certain threshold where quantization error is an issue (or where downstream read noise is somewhat more significant b/c you're only counting every 3-4 photoelectrons per digital increment).

I alluded to all of these from the very beginning, but I can't always write a novel every time I'm talking about a concept.

The magnitude of this difference surprised me a bit, I didn't expect such a large difference.  Pleasant surprise.

Also, careful about resampled views in Lightroom. I actually do see a tiny bit of banding in your image - perhaps the A7 sensor is more outdated compared to the A7R? The magenta blotching can be reduced typically with the Shadows Tint slider under Camera Calibration. Also, magenta noise does seem to be more of an issue with the A7 than A7R, if I remember correctly.

You're correct about the resampling, it exaggerates the effect slightly but it's definitely there.  This is actually similar to the pattern of purple blotches I see in the A7 files at high ISO that don't respond well to noise reduction that I complained about in the other thread.

The D810 surpasses even the A7R.

I've been very tempted to rent one of these to try out both the new AF system and the supposedly improved grip.

Also, that's a +5 EV push. Your initial claim that ISO 100 + 4EV falls apart to posterization and banding and noise compared to ISO 1600 is just not true - at least not with an A7R or D810. But, yes, quantization error and the effects of non-zero downstream read noise *will* have some effect at some extreme point. But even when it does, it's very easy to remove. It's usually as subtle as the noise that comes with higher ISOs. Not the magenta blotchiness and banding your example shows. Only time I see magenta blotchiness is in the JPEG preview, b/c of the limited quality/bit-depth of some of the preview JPEGs LR uses. There can be some significantly smaller faint magenta blotchiness for signals down in the RGB = 1,1,1 area, and those can easily be removed with color NR which thankfully rarely kills much actual image detail.

The initial claim was poorly phrased; +4 stops is where I start seeing blotchiness and extra color noise and is generally the limit of how far I push Exmor files.  I generally have to push 6 stops to see the banding and posterization issues.

Also, realize LR's histogram is not the best judge b/c of the way it actually works. I'm not sure what exactly you're referring to when you're talking about 'posterization', but as long as your sampling is such that shot/statistical noise is sampled by at least one digital increment in your Raw file, the dithering effect of the noise will take care of posterization. If we do a little math and use some approximations, a signal of 16 photoelectrons varies by +/- 4 b/c of shot noise, and that signal is represented by, say, roughly 4 at ISO 100. The noise is represented by 1 digital increment, so you're fine. Yes, below this, you're not sampling the noise properly, so you may run into issues. That's where you'll benefit from using ISO 400. But if you're seriously trying to use image data from 16 photons and below... well, now, *there's* an extreme case.

This is exactly why I had issue with your initial claim that you could maintain high ISO dynamic range by just underexposing ISO 100 files.  I suppose we have different definitions of high and extremely high ISO.

And like I said, above ISO 400, most of these issues are obviated, as you yourself have seen. From that point onward, you can do huge pushes and literally see no noise cost compared to shooting at the higher ISO.

Actually, you can perform quantitative tests to figure out exactly where this 'magical ISO' is.

Overall, I'm not sure what your point is though. You can generally use the technique I mentioned, save for ridiculously low signals of I'm guessing like 30 photoelectrons or less. I thought I was OK leaving that edge case out, at the risk of writing the novel I just wrote above.

And all that said, it does seem your A7 is not performing up to the level of the A7R or D800/810. Which is not too surprising - again, it's an older sensor.

Yes, switching to ISO 400 seems to give much nicer pushes up to ISO 6400.

21
EOS Bodies / Re: Just for Jrista: 2014 Market Data
« on: September 25, 2014, 01:27:12 AM »

It actually works well (usually without trouble) up to 5 1/3 to 5 2/3 stops. You may start seeing some issues starting at around 5 stops...depends on the exposure and how deep your really going (you can still ETTR with an Exmor camera). Push an Exmor raw that far, and with a good exposure you still won't see color noise, and those sensors are entirely devoid of any kind of banding unless your utterly cracked (there are guys on DPR who have used the brush to boost exposures from the D800 up to 10 stops, where they eventually finally to notice extremely minor banding amongst all the random color noise.) The Exmor advantage is a little over five stops, so that's generally what you can get with a strait exposure-slider and highlight/shadow-slider push in LR. Doing so does not result in anything remotely resembling the kind of noise you get when you push a Canon image a mere couple of stops, let alone three or four or more on top of some explicit shadow pushing with the shadow slider.

I've never been quite comfortable with the results from the A7 I have now pushed any more than 4 stops.  Then again, I never push my 5D3 shadows more than a stop without using MagicLantern.

You can work the data more...you can tweak it more carefully and preserve and enhance color. It's more work...and it may or may not be important to the ultimate goals. Simple fact of the matter is...you have the option with a camera built around an Exmor sensor. You don't have the option with a Canon sensor.

Just to clarify, this was never a Canon vs Sony issue.  Just a difference in my experience with Exmor vs what was being claimed.

I see banding and color noise, but no posterization. Posterization is basically "cartoonization", which results in the effect you see in this image:

The spikey behavior in the histogram represents posterization.  This isn't even really a function of the sensor at all, just that we don't have enough information to provide continuous color contours.  It's not going to appear as clearly as your exampe though due to the random noise.  In any case, Sarangiman clarified that he meant that higher ISOs were emulated with higher base ISO rather than always sticking with ISO 100 which, surprisingly to me, actually makes a huge difference.

The pinkish-red and blue blotchiness of the shadows in the image you have shared is a form of color noise. That could very likely be due to the lossy compression Sony employs in their raw files...which is one of the things I truly DO NOT like about their cameras. I cannot fathom why Sony would gimp their own cameras by using a non-raw "raw" image format...it aint RAW if your throwing away data. Bleh.  :-\

The severity of it is actually exaggerated slightly by LR's downsampling method but the blotchiness shown is very real.  This is also the same kind of behavior I was mentioning in the other thread where I was complaining about the A7 not responding well to noise reduction at high ISO.

22
EOS Bodies / Re: Canon Says Higher Resolution Sensors Are Coming Soon
« on: September 25, 2014, 01:08:47 AM »

Sure, I don't disagree with that. However, to outright state that their camera sensors are the best on the market is taking it too far. There are many other words and phrases that could be used to describe Canon sensors and still give Canon customers the sense that Canon is working to improve things. But to outright state that they don't know what measurements could possibly be indicating their sensors are lacking in the dynamic range department, or to outright state that their sensors are the best in the world...that's insane.

Especially when so many customers in the market KNOW it's a bold faced lie. Because that's exactly what it was...a bold faced lie. :P

This article is motivating me to rent a D810 for this weekend,  :-\.

23
EOS Bodies / Re: Just for Jrista: 2014 Market Data
« on: September 24, 2014, 11:46:03 PM »
And even if you do push 7, 8, or 9 stops, it's still not going to result in the ugly read noise or banding you'll see with gentle pushes of Canon DSLR files. So I still have no idea about the banding, color noise, posterization, and whatnot raptor3x was talking about from a 4 EV push. Which he's now, apparently, changed to a 6+ stop push - but even then, I don't see anything near ugly posterization, banding, or even significant noise. A tiny bit more noise than native ISO when you're pixel-peeping, yes, but hardly worth raising a fuss about.

This is what I mean when I'm talking about posterization and banding.  It's obviously an extreme case but it is there if you dig deep enough.  Edit:  I'm surprised you wouldn't think posterization would be an issue with underexposing using ISO 100, that's not even a property of the sensor.


24
EOS Bodies / Re: Just for Jrista: 2014 Market Data
« on: September 24, 2014, 11:43:13 PM »

I think the simple point is, when you need to push 6 stops...you can. And you can work the data to recover lost color fidelity. With a Canon camera, you can't. Well, you can push 6 stops, but as PBD's (and many other's examples, and hopefully soon my own, as I just rented an A7r and Metabones EF adapter) images demonstrate, the shadows are so riddled with noise that is so bad, no matter how much you work it, it isn't going to get any better.

With Sarangiman's window example, you could HDR that...but your almost guaranteed to get integration artifacts, and removing ghosts isn't always the solution...sometimes you have to manually work it, sometimes you have to excessively bracket (one of the original videos demonstrating ACR's 32-bit 20-stop TIFF HDR toning used 15 shots in bracket to handle an HDR of the interior of a darkened plane with a very bright window...the 15 shots weren't for DR, they were to avoid artifacting around and in that window.)

It isn't generally a common situation to need to push 6+ stops. There are cases where it could be very useful, and for certain kinds of photography, it could be more useful than in other kinds of photography. PBD's example images are far from extreme. Before the shadow lift, the shadows under the awning were not pitch black and buried deep. They were light shadows, you could see into them...and even lifting that, in a realistic edit, produced very clean, usable results on the Exmor camera, and some heavy banding and not much detail (recoverable or otherwise) on the Canon camera. Canon's banding and read noise reaches up to the lower midtones. The 7D II may prove to have solved the banding issue, if so, WONDERFUL! That's a good sign. There is still the random read noise, though, and that's still going to eat away at detail...possibly right up into the lower midtones. NOT having that problem would just be...wonderful.

The whole point was that even Exmor RAW files can't be pushed 6 stops without trouble.  Sarangiman's original claim was that you could shoot at high ISO and retain almost all of the dynamic range by shooting at ISO 100 and underexposing, which implied you're better off shooting underexposed at ISO 100 and then pushing however many stops to get to your desired emulated ISO.  This works quite well up to ISO 1600 where you start seeing some minor color splotches and a slight magenta cast, but nothing major.  As you try to push further though it starts falling apart pretty quickly.  What did surprise me though is that a 4 stop push from ISO 400 looks much better than the six stop push from ISO 100.  I was under the assumption than an ISO 100 file underexposed by 2 stops was essentially identical to an ISO 400 file, but that's clearly not the case as shown below.

25
EOS Bodies / Re: Just for Jrista: 2014 Market Data
« on: September 24, 2014, 09:17:40 PM »
"I'm sorry; not seeing any color noise, posterization, and banding. Not really seeing much of a noise cost at all to shooting ISO 100."

I think we should all be able to agree that there is also no color noise, posterization, or banding in my "real world" 100% crop as well, though that does only have an approximately three+ stop lift.

To see the banding you have to lift shadows ridiculous amounts, 6+ stops.  The posterization though can be seen in histograms relatively early on though but only really becomes a problem when, again, making silly shadow lifts.  The color noise surprises me though, as this is something I see when lifting ~4 stops or so.   Maybe an A7 vs A7R thing.

Ok, keep in mind that at 6+ stops you need to expect noise levels of ISO 6400 and above. C'mon. Let's be reasonable.

And yes, more than 5 stops of pushing and you might see some effects of downstream read noise or quantization error, compared to the ISO 6400 or higher (not pushed) shot.

And, yes, different sensors will perform slightly differently here. I believe the A7 is a slightly older sensor, but it shouldn't be accounting for the drastic differences in your experience vs. what I just showed above.

A 6+ stop push is where I'm most interested because most of my shooting is above ISO 3200.  Once you get to this point the midtones look noticeably worse for an ISO 100 shot pushed up than a native high ISO shot; this is the issue I have with the claim that you keep all the dynamic range.  You can keep the highlights but the shadows and midtones turn to garbage.  The reason I mentioned the histogram is because while you can't really see the posterization at ISO 1600 in the image, you can see it in the histogram already and it becomes visible in the pictures once you start trying this same technique at even higher ISO values.

26
EOS Bodies / Re: Just for Jrista: 2014 Market Data
« on: September 24, 2014, 08:42:03 PM »
"I'm sorry; not seeing any color noise, posterization, and banding. Not really seeing much of a noise cost at all to shooting ISO 100."

I think we should all be able to agree that there is also no color noise, posterization, or banding in my "real world" 100% crop as well, though that does only have an approximately three+ stop lift.

To see the banding you have to lift shadows ridiculous amounts, 6+ stops.  The posterization though can be seen in histograms relatively early on though but only really becomes a problem when, again, making silly shadow lifts.  The color noise surprises me though, as this is something I see when lifting ~4 stops or so.   Maybe an A7 vs A7R thing.

27
EOS Bodies / Re: Just for Jrista: 2014 Market Data
« on: September 24, 2014, 08:14:43 PM »

And some people wonder why I try to make it a point to dispel these fallacies?

P.S. Yes I know the ISO 100 shot could've been processed better, but I only spent 30s processing just to try & waste as little time as possible with this. Window lighting is pretty challenging and requires proper use of masks in PS if you want to do it right. Doesn't change my point.

Not all of the issues will show up the moment you try to do this for ISO 1600, although even in your own example I can already see shadow areas that have a significant magenta cast compared to the ISO 1600 version.  Additionally, I can almost guarantee that if you post the histogram for the ISO 100 image you'll already see the signature of posterization in the shadows.  Once you get up to ~6400 though you end up with very nice highlights and garbage midtones and shadows.  This was an idea I was quite excited about when the A7R first came out, but in practice it's just not that useful.

28
EOS Bodies / Re: Just for Jrista: 2014 Market Data
« on: September 24, 2014, 02:44:29 PM »
*And yet shooting at ISO 1600 - which deprives the sensor of 4 stops of light compared to ISO 100 in the traditional sense - is OK... hmm perhaps zlatko didn't understand that changing the ISO setting on your camera doesn't actually change the native sensitivity of the sensor. All it does is amplify the data. Well, me raising the exposure +4 stops is 'amplifying the data' - just at a different step of signal processing. Why is one valid and the other not? And that's the point - it *is* valid for Nikon/Exmor, just not for Canon, where you have to amplify early on. So this requires a shift in thinking, and it doesn't change my point: being able to amplify later is advantageous b/c it gives you stops and stops of highlight headroom, essentially giving you the full DR of the sensor at higher ISOs (there are limitations at extremely high ISOs, but I won't go into that here).

I thought this was what you were getting at.  This was literally one of the first things I tried when I first bought an A7R and unfortunately, this is not very useful at all.  Because middle grey will be buried so deep within the shadows, color noise, posterization, and horizontal banding (yes, Exmor has banding noise as well if you dig deep enough) all become major issues.  Even when emulating ISO 1600.

29
Lenses / Re: DXOMark Reviews Zeiss Otus 85mm f/1.4
« on: September 24, 2014, 10:35:33 AM »
I find it really interesting that the lens has, apparently, 1/3rd stop less vignetting on the D800 than the 5D3.

30
EOS Bodies / Re: Am I the only one excited about the new 7D mk2?
« on: September 23, 2014, 07:27:50 PM »
They're measuring the DR at ISO 6400 to be equal to the 5D3.

No, it's less.
It is better than for the 7D though, but about half a stop (maybe +/- 1/3 stop as the masked vs main area and high ISO ratings might differ).

The dynamic range of both the 7D2 and 5D3 at ISO 6400 is 8.3EV based on the measurements from the sample RAWs for the 7D2 and DXO for the 5D3.

It remains to be seen. I measured higher for 5D3 and even 5D2, but I didn't apply actual ISO rating correction for the 5D3 as DxO did. Their 5D3 copy also might be a little bit on the weak side for high ISO read noise compared the two I've tested.

I find this really interesting.  Do you think this is due to copy variation or due to a difference in the way the tests are conducted?  Do you see differences in high ISO DR for short exposures in bright light vs longer exposures in dim light?

Pages: 1 [2] 3 4 ... 14