Pages: [1]   Go Down

Author Topic: long exposure simulation  (Read 1655 times)

Eric Brody

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 487
    • http://www.ericbrodyphoto.com
long exposure simulation
« on: August 10, 2021, 08:22:33 pm »

I enjoy long exposure photography. I have a set of ND filters which I use often. I was recently reading about using multiple individual exposures to accomplish the same effect without filters. This involves using Photoshop, scripts, stacking and then combining the images.

I have avoided superwide lenses with bulbous front elements because using filters with them is complex and/or expensive.

I’d greatly appreciate hearing from those who have used this technique with such lenses.

Thanks.

Eric
Logged

mcbroomf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1534
    • Mike Broomfield
Re: long exposure simulation
« Reply #1 on: August 11, 2021, 06:05:28 am »

I've used the technique (median stacked) but not often and only once with a kind of comparison.  Unfortunately my ND filter slipped off the lens part way though (and into the river) so I only have pre-dawn images taken with it (for comparison against median stacked) and I've not processed them, but will later today.

Rather than resize for this forum here's a link to 3 images so far;
1st was a single capture earlier at dawn when a long exposure was needed (15 secs).  No ND.
2nd was a 16 file median stack @1/25th (after sunrise, an hour later)
3rd was one of the files, so single exposure of 1/25th
All were taken at ISO 100 so I was doing this to smooth the water rather than reduce noise

What's missing and I will add later today are an ND long exposure pre-dawn (single image @40 secs or so) and a 16 image stack pre-dawn with no ND filter to compare.  It looks like I may have a set at 640 ISO as well as 100 ISO from the pre-dawn series.  I also have more than 16 images taken post sunrise so I can add a 64 image stack later as well to compare to the 16 image stack.
All taken with a 21mm Loxia on Sony A7R3. 

Median Stack album
(If you click on the (i) icon top right, an info palette opens with exposure info)

Mike

Logged

jrsforums

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1288
Re: long exposure simulation
« Reply #2 on: August 11, 2021, 10:34:15 pm »

Doesn’t ‘median’ eliminate uncommon items, image to image?  Useful to take multiple images and remove moving objects, such as car & people

‘mean’ combines the images and can give the smoothing of water effect which you can get from a long exposure.
Logged
John

mcbroomf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1534
    • Mike Broomfield
Re: long exposure simulation
« Reply #3 on: August 12, 2021, 05:49:33 am »

I will check on both when I get the chance and add the comparison (did not get any time yesterday).

You are right that median eliminates moving objects and I've used that in museums.  Will be interesting to see if mean gives better water blur.
Logged

David Eckels

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3526
  • It's just a camera.
    • Website
Re: long exposure simulation
« Reply #4 on: August 12, 2021, 08:49:37 am »

Mean is the way to go. The shorter the shutter speed, the more exposures you need to get a smooth effect. In my experience 1/30 sec exposures require about 10 consecutive frames, but I've done it at 1/125 sec with 30-40. With my D850, I can combine multiple exposures in one frame (up to 10). Using a few of these, I can get quite a nice effect though it can be a bit tedious.

mcbroomf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1534
    • Mike Broomfield
Re: long exposure simulation
« Reply #5 on: August 12, 2021, 02:26:30 pm »

I added some more images to the album.  The last 4 are now sunrise scenes; 16 shot median, 16 mean, 32 median, 32 mean.

The water flow in the Buffalo river at this point is very slow and there was almost no wind that day so there at 1st glance there's not much to choose from between mean and median.  Looking closer though I see that the reflection of the top of the buildings is "tighter" in the median image than the mean.  I like both.  This would be very different if the scene were a fast flowing stream, waterfall or breaking waves.
Logged

sbay

  • Full Member
  • ***
  • Offline Offline
  • Posts: 225
    • http://stephenbayphotography.com/
Re: long exposure simulation
« Reply #6 on: August 17, 2021, 12:53:30 pm »

I do image stacking all the time. Two main reasons, to get the equivalent of a longer exposure time (for smoother water/clouds) or to reduce noise (generally astro landscapes).

You can do this in PS with smart objects but it is slow. However there are other programs that can do this much faster such as Long Exposure Stacker, image magick, etc. Sony used to be able to do this in camera with the app Smooth Reflections.

Mean results in slightly softer images but can leave ghosts if objects move. Median tends to be crisper but eliminates objects that are only in one frame like a moving airplane, car, person, etc. Can sometimes result in weird artifacts in smooth areas though.

I often use this technique in conjunction with ND filters. For example, a 10-stop ND filter might give you a 10s exposure in daylight. So instead of getting a 16-stop filter, just average 10, 20, 40, etc shots with the 10-stop to get to minutes in daylight.

For astro-landscapes, I might take 5 2min exposures and average them to get lower noise on the foreground instead of one 10-minute shot. This is very helpful as some types of noise build up with longer exposures and occasionally shots get ruined because of headlights from passing cars. With the averaging case, you just throw out the one with the car and use the rest.

The main drawback is that if an object is moving fast, it can leave gaps between shots. E.g. A car trail split amongst two pictures will have a gap due to the delay from one shot to the next.

Guillermo Luijk

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2005
    • http://www.guillermoluijk.com
Re: long exposure simulation
« Reply #7 on: September 12, 2021, 07:38:07 pm »

I find more accurate and also enriching to be able to choose between mean and median just by looking at their mathematical defnitions, instead of crazy testing both on a camera. If you look it this way, you'd know mean is the appropiate operation to mimic long exposure simulation since it basically implements exactly what a ND filter does: add photon counts, which median doesn't.

BTW to be perfectly accurate, the averaging should be done in the linear domain, i.e. using gamma=1 images. Averaging gamma encoded images (like those from sRGB or Adobe RGB RAW processings) will provide great results but slightly different to those from a genuine long exposure.

Median is another story, and although useful for some purposes (eliminate moving objects, removing shot noise,...), it is definitively not the appropiate tool to achieve long exposure.

Mean: https://www.overfitting.net/2021/05/apilado-por-media-simulando-iso-ultra.html
Median (vs mean): https://www.overfitting.net/2021/05/apilado-por-mediana-para-eliminar.html

Regards


jrsforums

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1288
Re: long exposure simulation
« Reply #8 on: September 13, 2021, 08:49:49 am »

I find more accurate and also enriching to be able to choose between mean and median just by looking at their mathematical defnitions, instead of crazy testing both on a camera. If you look it this way, you'd know mean is the appropiate operation to mimic long exposure simulation since it basically implements exactly what a ND filter does: add photon counts, which median doesn't.

BTW to be perfectly accurate, the averaging should be done in the linear domain, i.e. using gamma=1 images. Averaging gamma encoded images (like those from sRGB or Adobe RGB RAW processings) will provide great results but slightly different to those from a genuine long exposure.

Median is another story, and although useful for some purposes (eliminate moving objects, removing shot noise,...), it is definitively not the appropiate tool to achieve long exposure.

Mean: https://www.overfitting.net/2021/05/apilado-por-media-simulando-iso-ultra.html
Median (vs mean): https://www.overfitting.net/2021/05/apilado-por-mediana-para-eliminar.html

Regards

How would you do the ‘mean’ processing in the linear domain?  Would using a ‘linear profile’ (vs. Adobe or camera profiles) in LR/ACR before sending to Photoshop do that?
Logged
John

Guillermo Luijk

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2005
    • http://www.guillermoluijk.com
Re: long exposure simulation
« Reply #9 on: September 13, 2021, 01:13:02 pm »

How would you do the ‘mean’ processing in the linear domain?  Would using a ‘linear profile’ (vs. Adobe or camera profiles) in LR/ACR before sending to Photoshop do that?
Photoshop allows to do that even on gamma encoded images (I'm translating into English, some words could be different):

'Edition' -> 'Colour adjustments...' -> Blend RGB colours using gamma: 1.00

However as I pointed, there will be no big difference, and you may even prefer the non-linear stacking result. Just check both.

Regards

fdisilvestro

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1852
    • Frank Disilvestro
Re: long exposure simulation
« Reply #10 on: September 14, 2021, 05:22:59 am »

How would you do the ‘mean’ processing in the linear domain?  Would using a ‘linear profile’ (vs. Adobe or camera profiles) in LR/ACR before sending to Photoshop do that?

The "Linear profiles" in LR/ACR and linear Gamma ICC profiles are two different things. Linear DCP profiles (ACR/LR) refer to a tone curve applied to the image, usually to simulate film response.

The Tone Response curve used for encoding is separate and varies depending the ICC profile for the rendered image. E.g. Adobe RGB uses Gamma 2.2, Prophoto RGB uses Gamma 1.8 and sRGB has a specific tone curve that sometimes is approximated to a Gamma 2.2

Guillermo is correct that you should use Linear Gamma to have "Scientifically correct" mean, but non-linear stacking could give you a pleasant result anyway.

jrsforums

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1288
Re: long exposure simulation
« Reply #11 on: September 14, 2021, 10:04:02 am »

Thanks to both of you
Logged
John
Pages: [1]   Go Up