Ad
Ad
Ad
Pages: [1]   Bottom of Page
Print
Author Topic: HDR - ISO / Noise / moving objects  (Read 6302 times)
Adam L
Full Member
***
Offline Offline

Posts: 188


WWW
« on: June 03, 2010, 02:18:12 PM »
ReplyReply

I understand that the bracketed images generate low/no noise when merged to HDR.    Its got me to thinking about raising the ISO so that the shutter speed is maximized.  This should minimize the amount of movement that may take place in the image.  My question is can this be pushed to the maximum ISO setting?  Are there IQ trade-offs associated with increasing ISO even on HDR?
Logged

"That's a lot of money to move a few pixels around"
Monito
Jr. Member
**
Offline Offline

Posts: 96



WWW
« Reply #1 on: June 03, 2010, 08:32:50 PM »
ReplyReply

Quote from: Adam L
I understand that the bracketed images generate low/no noise when merged to HDR.    Its got me to thinking about raising the ISO so that the shutter speed is maximized.  This should minimize the amount of movement that may take place in the image.  My question is can this be pushed to the maximum ISO setting?  Are there IQ trade-offs associated with increasing ISO even on HDR?

The way HDR (and the associated "4 stop zero noise" technique) achieves low noise is by shooting with low ISO and bracketing.  If you make an HDR with noisy high ISO, you will get a noisy HDR.
Logged

MonitoPhoto (Landscape, Architecture, Portraits: Halifax, Nova Scotia)
Adam L
Full Member
***
Offline Offline

Posts: 188


WWW
« Reply #2 on: June 04, 2010, 07:34:05 AM »
ReplyReply

Can someone confirm Monito's post?  I've been under the impression that noise is virtually zero due to multiple exposures layered on top of each other and how the application blends these together.
Logged

"That's a lot of money to move a few pixels around"
walter.sk
Sr. Member
****
Offline Offline

Posts: 1332


« Reply #3 on: June 04, 2010, 02:23:32 PM »
ReplyReply

Quote from: Adam L
Can someone confirm Monito's post?  I've been under the impression that noise is virtually zero due to multiple exposures layered on top of each other and how the application blends these together.
I shoot HDR with a Canon 1DMkii, which was a "low noise" camera when it first came out, but has been left in the dust (no pun intended) by the newer low-noise high ISO cameras.

I usually shoot a bracket of 5 images, and tried high ISO (800, 1600 and 3200) in various sequences, specifically to avoid ghosting with moving subjects.  Not thrilled with the results, although I do some large moves, tonality-wise.  The noise can make skies look miserable, so I try to minimize my ISO settings.  When I do use higher ISO, I process the images post-HDR with DFine, and paint the results in where the NR is needed.
Logged
EduPerez
Sr. Member
****
Offline Offline

Posts: 695


WWW
« Reply #4 on: June 08, 2010, 06:29:25 AM »
ReplyReply

In my experience, unless you are doing very long exposures, the timespan between photographs is far more important than the exposure time.
Logged

BartvanderWolf
Sr. Member
****
Offline Offline

Posts: 3869


« Reply #5 on: June 08, 2010, 06:44:56 PM »
ReplyReply

Quote from: Adam L
Can someone confirm Monito's post?

Yes.

Quote
I've been under the impression that noise is virtually zero due to multiple exposures layered on top of each other and how the application blends these together.

No, that would be achieved by averaging images and that would require equal exposure times for all shots, thus not expanding the captured luminosity range. The averaging of noise would increase the dynamic range some though.

Cheers,
Bart
Logged
Plekto
Sr. Member
****
Offline Offline

Posts: 551


« Reply #6 on: June 28, 2010, 12:08:19 PM »
ReplyReply

And that's the real problem.  There is no free lunch.  If all of your exposures have a lot of noise, the end result will also suffer from it as well, though it will result in a blurry or grainy appearance more than individual defects.  Me?  I shoot at the camera's highest quality almost all of the time(100/200) and bracket as required.  Often 1 stops is far in excess of what I need.   A typical setting would be 0, -1 , +1 all tightly bracketed together.   As a rule, computer software does a better job of cleaning up noise than the in-camera processing does.

http://luminous-landscape.com/forum/index....showtopic=17775 - Zero Noise software/technique
BTW, if an admin is reading this, can we get that post bookmarked/etc so that it's easy to find?   Newer people here likely don't even know it exists, yet it's amazing in what it does(free as well - can't beat that)
Logged
Ray
Sr. Member
****
Offline Offline

Posts: 8939


« Reply #7 on: June 29, 2010, 09:17:18 AM »
ReplyReply

Quote from: Adam L
I understand that the bracketed images generate low/no noise when merged to HDR.    Its got me to thinking about raising the ISO so that the shutter speed is maximized.  This should minimize the amount of movement that may take place in the image.  My question is can this be pushed to the maximum ISO setting?  Are there IQ trade-offs associated with increasing ISO even on HDR?


Adam,
You've got the right idea. Merging bracketed exposures at high ISO will not result in as low a noise as bracketed images taken at base ISO, but still some improvement. If you have a tripod and the subject is static, it's always preferable to bracket exposures at base ISO for maximum dynamic range and minimum noise.

However, if circumstances dictate that a higher shutter speed than what is practical at base ISO is required, for whatever reason, slight subject movement, lack of IS in the lens, no tripod and poor lighting etc, then bracketing at high ISO will produce cleaner results with a higher dynamic range than a single shot at base ISO.

For example, let's say the exposure for a full ETTR at f2.8 and ISO 100 is 1/30th. There are some dark shadows containing wanted detail which will be noisy, even at base ISO. You decide to bracket exposure, but 1/30th is the slowest speed which might produce an acceptably sharp result in the circumstances, but it is still a bit risky. Plus 2 stops requires a shutter speed of 1/8th which is useless in the circumstances. Even +1EV at 1/15th is not on.

In my opinion, the best option would be to bracket exposure at ISO 800. Plus 2EV, the slowest shutter speed, then becomes 1/60th.

1/60th at ISO 800 , using a Canon or Nikon DSLR, will produce cleaner shadows than 1/60th at base ISO under the same lighting conditions. 1/60th at ISO 800 will even produce cleaner shadows than 1/30th at base ISO, although I stand to be corrected if anyone wishes to carry out a practical test to prove me wrong.  
Logged
BobFisher
Guest
« Reply #8 on: June 29, 2010, 10:05:39 AM »
ReplyReply

You may even see an increase in noise in the HDR images vs. the individual input shots due to the nature of HDR in enhancing everything - colour, detail, even noise.  HDR is, essentially, meant for non-moving subjects.  Something like water being an exception.  If you've got a moving subject then you may need to fall back on the deghosting features of some HDR applications.  Some work better than others and won't work for subject movement during an exposure, only movement between exposures.  If you need to trade off a higher ISO to freeze action then you may also need to accept a higher level of noise.

Ray, ETTR really has nothing to do with HDR shooting.  It's a technique preferred by some to get the most out of a single shot.  And it should matter little what the shutter speed/ISO combination is as long as you're pushing the histogram to the right edge.  If, under your scenario, 1/60 @ 800 pushes the histo to the 'right' spot but 1/30 at base ISO doesn't then it's not a valid comparison of the ETTR methodology.
Logged
Ray
Sr. Member
****
Offline Offline

Posts: 8939


« Reply #9 on: June 29, 2010, 07:17:55 PM »
ReplyReply

Quote from: BobFisher
Ray, ETTR really has nothing to do with HDR shooting.  It's a technique preferred by some to get the most out of a single shot.  And it should matter little what the shutter speed/ISO combination is as long as you're pushing the histogram to the right edge.  If, under your scenario, 1/60 @ 800 pushes the histo to the 'right' spot but 1/30 at base ISO doesn't then it's not a valid comparison of the ETTR methodology.

Bob,
I think you might have missed the point. It's the shutter speed required for a full ETTR shot which may be the critical factor when making a decision to increase ISO, in the absence of a tripod, or when either the subject or the cameraman is moving.

As you probably know, the auto-alignment of images in Photoshop is now so good, it really isn't necessary to use a tripod when bracketing provided the longest exposure in the bracketed series is fast enough for a sharp hand-held shot, and provided your camera has a reasonably fast frame rate.

The longest exposure in a bracketed series will inevitably be longer than the exposure required for a full ETTR (otherwise what's the point?). In such circumstances, if that shutter speed is too slow, bracketing at a higher ISO will not only potentially produce cleaner results (than a single ETTR shot at base ISO), but also sharper results than bracketing at base ISO when that longest exposure is too slow.

In circumstances where there is significant movement in parts of the scene, such as fast moving clouds for example, the automated process of merging to HDR may not produce satisfactory results. In such a situation it may be possible to manually combine feathered selections of sky and foreground (for example), taking the sky from one of the underexposed images and the darker foreground from one of the overexposed images.

By way of further explanation, perhaps I have assumed too much that you are all familiar with the high-ISO advantages of the CMOS design in Canon and Nikon cameras whereby the signal is boosted in analogue form at high ISO before digitisation. This has the effect of resulting in a greater signal-to-noise in the darker parts of an image.

You can test this for yourself. Take any scene with a high SBR (subject brightness range). Make an ETTR shot at base ISO, then shoot the same scene at ISO 800 using the same shutter speed and f stop.  The second shot will of course be overexposed, but you should find that the shadows and lower midtones will be significantly cleaner than the ETTR shot at base ISO.

Of course, certain camera designs lack this facility for analogue boost, such as many MFDBs and P&S cameras.
« Last Edit: June 29, 2010, 07:40:45 PM by Ray » Logged
BobFisher
Guest
« Reply #10 on: June 29, 2010, 07:44:19 PM »
ReplyReply

I come back to my earlier point.  ETTR has nothing whatever to do with HDR or shooting for HDR.  Comparing HDR to ETTR is completely and totally useless.  See my other point on the effect of noise in HDR as well.  

And I'll take a tripod (and cable release) over 9fps and fast shutter speeds any day, thanks.
Logged
Ray
Sr. Member
****
Offline Offline

Posts: 8939


« Reply #11 on: June 29, 2010, 08:05:48 PM »
ReplyReply

Quote from: BobFisher
I come back to my earlier point.  ETTR has nothing whatever to do with HDR or shooting for HDR.  Comparing HDR to ETTR is completely and totally useless.  See my other point on the effect of noise in HDR as well.  

And I'll take a tripod (and cable release) over 9fps and fast shutter speeds any day, thanks.


Tell me Bob, would you use HDR when an ETTR is sufficient to provide an image with clean shadows without blowing highlights in other parts? The purpose of HDR is to overcome the limited dynamic range of the camera. It is used when an ETTR is insufficient in respect of DR. It's used when the SBR of the scene exceeds the DR of the camera.

Using high ISO is sometimes a necessity when shutter speeds are too slow for a full ETTR, whether bracketing or not.

Bracketed shots at high ISO will result in greater DR and cleaner shadows than a single ETTR shot at high ISO. I can't understand the difficulty here.

Another issue is determing the ideal exposures for bracketing.  If your camera provides a maximum auto-bracketing range of +/- 2EV (as Canon DSLRs do), there would be little point if the +2EV exosure turned out to be no more overexposed than a correct ETTR. In order to ensure that this does not happen, you need to be aware of the sort of exposure that would produce an ETTR, which might be set at the -1EV level for safety, rather than -2EV.

You would then get a 3 stop increase in DR. If that's not sufficient for your purposes, you would need to use a tripod and set the exposure manually. One of the advantages of the Nikon DSLRs is they provide a wider range of exposures for auto-bracketing as well as the facility to auto-bracket ISO at a fixed shutter speed and f stop.
« Last Edit: June 29, 2010, 09:00:26 PM by Ray » Logged
BobFisher
Guest
« Reply #12 on: June 29, 2010, 09:27:26 PM »
ReplyReply

I don't use ETTR.  Period.
Logged
Ray
Sr. Member
****
Offline Offline

Posts: 8939


« Reply #13 on: June 30, 2010, 12:14:41 AM »
ReplyReply

Quote from: BobFisher
I don't use ETTR.  Period.

Why not? Please enlighten us as to your technique that makes the use of ETTR redundant.

My understanding is, those who bracket exposure for the purpose of merging to HDR do so in order to increase the dynamic range of the capture and reduce shadow noise. Why go to the trouble of taking multiple exposures to merge to HDR when a single ETTR exposure may suffice.

The ETTR method is a handy reference point.
Logged
BobFisher
Guest
« Reply #14 on: June 30, 2010, 06:21:53 AM »
ReplyReply

The original post was dealing with HDR.  I've indicated that bringing in a discussion of ETTR is a red herring.  I regret getting drawn into it and won't discuss it further.  This thread is about HDR.  Period.
Logged
Ray
Sr. Member
****
Offline Offline

Posts: 8939


« Reply #15 on: June 30, 2010, 10:27:05 AM »
ReplyReply

Quote from: BobFisher
The original post was dealing with HDR.  I've indicated that bringing in a discussion of ETTR is a red herring.  I regret getting drawn into it and won't discuss it further.  This thread is about HDR.  Period.

Well that is a pity. I thought for a moment I might learn some new technique for assessing exposure settings for HDR purposes. But I see you are not in the communicative mood.

I personally always try to get an assessment of what would be the exposure for an ETTR, at a particular f stop and ISO,  before I select the exposures and range for auto-bracketing. I find from experience that when I fail to do this, either through laziness or because I'm in a hurry, using the camera in aperture priority mode, I tend to make mistakes, such as the longest exposure being unusable because it's too slow, or realising after the fact that the longest exposure is not long enough and provides little or no benefit beyond a single ETTR shot.

As I've mentioned before, if I'm autobracketing +/- 2EV, I would try to get a shutter speed appropriate for ETTR purposes at the -1EV point in the range so I can get an increase in DR of 3 stops. If the +2 EV point then proves to be too slow for a hand-held shot, I would raise ISO, camera in manual mode, until the shutter speed at +2 EV were fast enough for a sharp image.

This gets to the heart of the question that Adam, the OP, asked; essentially is there any point in taking shots for HDR purposes at high ISO, and are there image quality trade-offs.

The answer is, there are trade-offs in IQ when using high ISO, but in certain circumstances, particularly hand-held shots with lenses that have no IS, bracketed shots at high ISO can provide a worthwhile reduction in noise and an increase in DR, although I would emphasise that such technique is more successful at overcoming camera shake than movement in the subject. If movement within the subject takes place during the relatively long intervals between each exposure, no matter how fast the shutter speeds used, the misalignment amongst the frames may be too much for the automated HDR process to handle successfully. In such circumstances, selective manual merging to HDR might be the best option.
Logged
BenGreaves
Newbie
*
Offline Offline

Posts: 7


« Reply #16 on: July 04, 2010, 05:36:33 AM »
ReplyReply

If you like, what you can do is actually mask some of one of the original images to get the movement in the image displaying the right look that you are after.  Or you can always create an HDR image from a single raw.  This does raise the noise level a little bit and does rely on the original raw being a good exposure but its nothing the imagenomic or noise ninja cant solve in a jiff
Logged
Pages: [1]   Top of Page
Print
Jump to:  

Ad
Ad
Ad