LiveND for Noisereduction/Nightshot (vs. HR and HHHR)

Hi,

I tried to analyze how LiveND on the EM1.3 is working and based on the data/limitations I came to the following conclusion:

For example the LiveND32:
The fastest possible (combined) shutterspeed is 1/2sec. When deviding this by 32 this results in 1/60sec (=sensor readout time of the EM1.3). My interpretation is that the ND mode translates in the Number of pictures taken. E.g. when we set ND32 and 1/2sec it will take 32 pictures at 1/60sec and averages them into one final image with a virtual shutterspeed of 1/2sec (32 x 1/60sec). Since it will shoot 60fps @ 1/60sec this will generate smooth transition of motion.

When checking the other LiveND option this looks the same (fastest possible shutterspeed / # of shots = 1/60sec)


LiveND Settings and shortest total (virtual) exposure time

When using LiveND32 (combining / averaging 32 single shots @ 1/60sec) this should give us 5 EV stop improvment

Doing a short test in my dim lit basement. All shot at ISO 800 - F4.0 - 1/5sec (LiveND = 32x 1/5sec = 6sec). Developed in Olympus Workspace: just adding +2EV, noise reduction etc. turned off. Highres-shots Downscaled to 20MP in Affinity Photo and then cropped:


20MP Single shot: +2EV - ISO 800 - F4.0 - 1/5sec


20MP LiveND32: +2EV - ISO 800 - F4.0 - 32x 1/5sec = 6sec


80MP HR shot downscaled to 20MP: +2EV - ISO 800 - F4.0 - 1/5sec


50MP HR shot downscaled to 20MP: +2EV - ISO 800 - F4.0 - 1/5sec

As you can see the LiveND32 shot has the lowest noise. The both HR shots have more details and also lower noise than the 20MP singel shot (but not as good as the LiveND32)

My conclusion with still a lot of open questions (on using LiveND for Noise reduction):
1. It is working pretty good, so I see some potential there
2. Why not use a single low ISO long exposure vs LiveND32:
a) with ISO800 and 5EV improvment we end up at ISO25 which is not available
b) LiveND32 could also be used with ISO64 in good light resulting in ISO 2
3. In good light there might be some limitations, since shortes single frame exposure time is 1/60sec (can be too long in bright light)
4. HHHR might still be the better choice: also reducing the noise plus adding detail at the same time. HHHR combines up to 16 shots (=max. 4EV improvement).
5. Motion should be handled better with LiveND vs HHHR (since this is the purpose of LiveND). But: maybe using HHHR at 1/60sec creates a similar effect (on a quick check of a tree in the wind this looked OK, but not perfect).

Thanks for this!

I cannot wait till EM5IV (or Pen-F II) implements these features!

ND16 (4 stops) is my ND of choice for wide-angle daylight live-ND (in S mode), combined with hhhr in A mode.. plus C1 set for normal shooting.

--
Emil Pozar, http://www.alpfabet.com

Thanks for posting this.

abgestumpft said:

LiveND Settings and shortest total (virtual) exposure time

Click to expand...

When you use LiveND, say ND32, the result is an average of 32 individual shots. Since noise is random, averaging can very effectively reduce noise.

LiveND is actually available to every digital camera if you are willing to do the averaging on a computer. I do that regularly. This is my approach to raising exposure of dark area within an image.

Space the final frontier said:

LiveND is actually available to every digital camera if you are willing to do the averaging on a computer. I do that regularly. This is my approach to raising exposure of dark area within an image.

Click to expand...

Not quite: liveND starts the next exposure right after reading a line (so the line is reset right after reading), so this rules out any sequence taken with mechanical shutter. With electronic shutter... it depends how it works, some may leave a gap between exposures some may not.

So there may be big differences when there is continuous movement in the frame. But otherwise the effect is the same, for example in a landscape shot where everything is more or less static and what movement there is would smooth out with enough shots anyway e.g. waves or trees.

prsc said:

Space the final frontier said:

LiveND is actually available to every digital camera if you are willing to do the averaging on a computer. I do that regularly. This is my approach to raising exposure of dark area within an image.

Click to expand...

Not quite: liveND starts the next exposure right after reading a line (so the line is reset right after reading), so this rules out any sequence taken with mechanical shutter. With electronic shutter... it depends how it works, some may leave a gap between exposures some may not.

So there may be big differences when there is continuous movement in the frame. But otherwise the effect is the same, for example in a landscape shot where everything is more or less static and what movement there is would smooth out with enough shots anyway e.g. waves or trees.

Click to expand...

It would be interesting to compare:

1. LiveND32
with
2. manual approach: shooting 32 frames (with frame count limiter) @ 60fps (Electronic shutter) and 1/60sec exposure time and average them afterwards (either in camera or via computer)

Maybe Olympus just automated what astrotripper wrote a while back in this thread
HOW-TO: LiveND, 14 stops DR, ISO 25 RAW on any Olympus camera

To me it looks like the EM1.3 is not doing any kind of allignment on LiveND (e.g. there is also no post-prossing time spend like on HHHR were the camera calculates several seconds after the HHHR shot was taken)

Interesting comparison

For static subjects looks to be a quite convenient way to minimize the noise without getting huge files to resample on the PC.

What I also noticed is that the moiré, on the black wallet and on the charger at the bottom of the pictures, disappeared in the resized HR shots but not in the LiveND.

is the moiré present in the original files or it also the HR files do not show it?

Stizzu68 said:

Interesting comparison

For static subjects looks to be a quite convenient way to minimize the noise without getting huge files to resample on the PC.

What I also noticed is that the moiré, on the black wallet and on the charger at the bottom of the pictures, disappeared in the resized HR shots but not in the LiveND.

is the moiré present in the original files or it also the HR files do not show it?

Click to expand...

The moire is also not present in the original HR shots:


HR Tripod shot - 100% crop - ISO 800 - +2EV correction in Oly Workspace - no noise reduction applied


HHHR shot - 100% crop - ISO 800 - +2EV correction in Oly Workspace - no noise reduction applied

ND0 without an ISO restriction would be a blast

--
Emil Pozar, http://www.alpfabet.com

Hi,

I did a quick test between:

1. LiveND8


2. 8 Shots at 1/60sec, 60fps stacked in Camera using this procedure: https://www.dpreview.com/forums/post/63214336
I don't know why, but this resulted in a WB shift (orange/brown) which I tried to correct


3. the same shots from 2. stacked in Affinity Photo
I loaded the same 8 shots I stacked in camera in Affinity Photo:
File -> New Stack
This automatically creates the stack with blending mode "Median" -> I exported to JPG without any other manual action



Here are the results


1. LiveND8




2. 8 fotos stacked in camera + WB corrected




3. Same 8 single shots autostacked in Affinity Photo (Median)



To me the LiveND8 and the in-camera stacking of 8 single shots look very similar (the character of the blur). It looks like the LiveND will take a predefined number of images (depending on selected ND filter) at 1/60sec and 60fps and then merges them (automated procedure of this: https://www.dpreview.com/forums/post/63214336 )

Based on this my conclusion of pros&cons of LiveND vs Manual approach is:

1. Advantages: LiveND
a) Automated Process (one click solution) - no post processing needed
b) only one final image (no need to handle 8, 16, 32,... images)
c) Live preview of effect

2. Advantages: Manual shooting
a) no limit of fastest shutterspeed (1/60) and ISO (max ISO800 with LiveND). For motion blur effect you should stick to 1/60sec, but for noise reduction of static subjects this is not needed
b) should work better with many images and handheld: can make use of auto-allignment features in post processing. E.g. with LiveND32 I have to make sure I can handheld 1/2sec, with manual shooting you just have to make sure that single frame is not blurred
c) you can make use of different effects: like median or mean blending methods...

abgestumpft said:

prsc said:

Space the final frontier said:

LiveND is actually available to every digital camera if you are willing to do the averaging on a computer. I do that regularly. This is my approach to raising exposure of dark area within an image.

Click to expand...

Not quite: liveND starts the next exposure right after reading a line (so the line is reset right after reading), so this rules out any sequence taken with mechanical shutter. With electronic shutter... it depends how it works, some may leave a gap between exposures some may not.

So there may be big differences when there is continuous movement in the frame. But otherwise the effect is the same, for example in a landscape shot where everything is more or less static and what movement there is would smooth out with enough shots anyway e.g. waves or trees.

Click to expand...

It would be interesting to compare:

1. LiveND32
with
2. manual approach: shooting 32 frames (with frame count limiter) @ 60fps (Electronic shutter) and 1/60sec exposure time and average them afterwards (either in camera or via computer)

Maybe Olympus just automated what astrotripper wrote a while back in this thread
HOW-TO: LiveND, 14 stops DR, ISO 25 RAW on any Olympus camera

To me it looks like the EM1.3 is not doing any kind of allignment on LiveND (e.g. there is also no post-prossing time spend like on HHHR were the camera calculates several seconds after the HHHR shot was taken)

Click to expand...

Yeah it is not doing any alignment. However I think that is just a logical next step for the LiveND functionality, although it will complicate its use somewhat. I mean they already have quite good alignment algorithms for the HHHR anyway, so may as well use those for LiveND (and the processing time isn't as bad as with HHHR since it's just align and average).

abgestumpft said:

Hi,

I did a quick test between:

1. LiveND8

2. 8 Shots at 1/60sec, 60fps stacked in Camera using this procedure: https://www.dpreview.com/forums/post/63214336
I don't know why, but this resulted in a WB shift (orange/brown) which I tried to correct

3. the same shots from 2. stacked in Affinity Photo
I loaded the same 8 shots I stacked in camera in Affinity Photo:
File -> New Stack
This automatically creates the stack with blending mode "Median" -> I exported to JPG without any other manual action

Here are the results


1. LiveND8


2. 8 fotos stacked in camera + WB corrected


3. Same 8 single shots autostacked in Affinity Photo (Median)

To me the LiveND8 and the in-camera stacking of 8 single shots look very similar (the character of the blur). It looks like the LiveND will take a predefined number of images (depending on selected ND filter) at 1/60sec and 60fps and then merges them (automated procedure of this: https://www.dpreview.com/forums/post/63214336 )

Based on this my conclusion of pros&cons of LiveND vs Manual approach is:

1. Advantages: LiveND
a) Automated Process (one click solution) - no post processing needed
b) only one final image (no need to handle 8, 16, 32,... images)
c) Live preview of effect

2. Advantages: Manual shooting
a) no limit of fastest shutterspeed (1/60) and ISO (max ISO800 with LiveND). For motion blur effect you should stick to 1/60sec, but for noise reduction of static subjects this is not needed
b) should work better with many images and handheld: can make use of auto-allignment features in post processing. E.g. with LiveND32 I have to make sure I can handheld 1/2sec, with manual shooting you just have to make sure that single frame is not blurred
c) you can make use of different effects: like median or mean blending methods...

Click to expand...

Does Live ND take any time to process the images? I know that HHHR takes 10 seconds or more for each shot. Also, how easy is it to turn on Live ND? Could you realistically shoot every landscape photo in good light as way of emulating lower base ISO, without feeling tedious?



For example, I know that automatic two image overlay/merge in EM5III is easy to trigger from the main menu, and it does help reduce the noise. But after each shot, I have to dive into the menu again to re-enable it which makes it a hassle to use successively. I haven't found a way to assign this option to a button either. So while I appreciate the ease of automated merging of two images, realistically it's not worth it for me to use every time I take a photo.

So how do you think Live ND would work if someone wanted to use it as the default shooting method to essentially leave-on by default for noise reduction purposes?

There is no extra processing time. You can assign LiveND to a button that then acts as a toggle, and if you hold the button down you can turn the wheel to change the ND level.

AlmostDoctor said:

Does Live ND take any time to process the images? I know that HHHR takes 10 seconds or more for each shot. Also, how easy is it to turn on Live ND?

Click to expand...

There is no post-processing time at all (at least there is no processing-bar shown like on HHHR).

AlmostDoctor said:

Could you realistically shoot every landscape photo in good light as way of emulating lower base ISO, without feeling tedious?

Click to expand...

Yes, except that you have some limitations:
1. Shortest single frame shutterspeed: 1/60sec -> can be too long in good light
2. Max ISO 800 (not a problem in good light)

Also LiveND32 (1/2sec total virtual exposure time) can be a challenge to handhold (not an issue with tripod of course). So other modes like LiveND16 or LiveND8 might work better.

Right now I would prefer HHHR (and normal HR on Tripod, when available) over LiveND in good light (exept there is lot of motion in the picture): it also reduces the noise, but also adds details, and can be used at faster shutterspeeds, does some auto-allignment of the shots (HHHR),...

AlmostDoctor said:

For example, I know that automatic two image overlay/merge in EM5III is easy to trigger from the main menu, and it does help reduce the noise. But after each shot, I have to dive into the menu again to re-enable it which makes it a hassle to use successively. I haven't found a way to assign this option to a button either. So while I appreciate the ease of automated merging of two images, realistically it's not worth it for me to use every time I take a photo.

Click to expand...

On the EM1.3 you can assign LiveND to a button: this will enable (last used setting) and disable it with a simple button press. When pressing&holding this button, you can also change the ND Settings with front- or backdial: LiveND2, ND4, ND8, ND16, ND32

AlmostDoctor said:

So how do you think Live ND would work if someone wanted to use it as the default shooting method to essentially leave-on by default for noise reduction purposes?

Click to expand...

If you plan to use it regularly I would assign it to a custom mode and / or button.

abgestumpft said:

AlmostDoctor said:

Does Live ND take any time to process the images? I know that HHHR takes 10 seconds or more for each shot. Also, how easy is it to turn on Live ND?

Click to expand...

There is no post-processing time at all (at least there is no processing-bar shown like on HHHR).

abgestumpft said:

Could you realistically shoot every landscape photo in good light as way of emulating lower base ISO, without feeling tedious?

Click to expand...

Yes, except that you have some limitations:
1. Shortest single frame shutterspeed: 1/60sec -> can be too long in good light
2. Max ISO 800 (not a problem in good light)

Also LiveND32 (1/2sec total virtual exposure time) can be a challenge to handhold (not an issue with tripod of course). So other modes like LiveND16 or LiveND8 might work better.

Right now I would prefer HHHR (and normal HR on Tripod, when available) over LiveND in good light (exept there is lot of motion in the picture): it also reduces the noise, but also adds details, and can be used at faster shutterspeeds, does some auto-allignment of the shots (HHHR),...

abgestumpft said:

For example, I know that automatic two image overlay/merge in EM5III is easy to trigger from the main menu, and it does help reduce the noise. But after each shot, I have to dive into the menu again to re-enable it which makes it a hassle to use successively. I haven't found a way to assign this option to a button either. So while I appreciate the ease of automated merging of two images, realistically it's not worth it for me to use every time I take a photo.

Click to expand...

On the EM1.3 you can assign LiveND to a button: this will enable (last used setting) and disable it with a simple button press. When pressing&holding this button, you can also change the ND Settings with front- or backdial: LiveND2, ND4, ND8, ND16, ND32

abgestumpft said:

So how do you think Live ND would work if someone wanted to use it as the default shooting method to essentially leave-on by default for noise reduction purposes?

Click to expand...

If you plan to use it regularly I would assign it to a custom mode and / or button.

Click to expand...

yes but I really do not think that it is something that most people will use often. At least for me in general LiveND and HR have several limitations related to the presence of moving subjects that cannot become the "normal" way to take pictures.

If I would be most of the time not happy with the quality or noise of the m43 files I would look for a larger sensor camera.

But it is good to have it and in some special occasions can be very useful

I have in mind to give it a try to the LiveND to replace a real-ND filters for waterfall shots or pano with clouds but still did not have time to try it.

Another advantage of manual stacking is that you can "eliminate" moving objects. You can effectively eliminate moving people in a busy intersection by taking a number of shots and then stack them in a photo editor.

Thanks, this is very interesting. I think you could apply a bit of noise reduction to the HR images and improve the fine grain noise while still preserving more detail than is in the LND image. Also, look at the moire noise on the black folder at the bottom - it's absent in the HR images as well as much more detail being there. A very useful and helpful test. Thanks.

Oly

Had some time this morning to play with LiveND albeit a controlled setting. I underexposed this shot -5EV then pushed exposure in post.

Here is the scene as exposed (left) then +5EV (right)



Here are the results side-by-side by method. All size adjusted to match the standard 20MP frame.



The LiveND32 looks the cleanest to my eyes but it looks quite similar to HHHR. If there is a difference there, I'm not sure it affects the deliverable in the end. Even the noise in the HR shot isn't intrusive - perhaps a stop behind HHHR and LiveND. All are better than the standard 20MP image - no surprise there. :-D

What you gain in shadow detail and latitude, you also lose in highlight roll off and detail.