Faking an ND Filter for Long Exposure Photography

Sep 24, 2013

Guest Author

We love it when our readers get in touch with us to share their stories. This article was contributed to DIYP by a member of our community. If you would like to contribute an article, please contact us here.

Faking an ND Filter for Long Exposure Photography

Sep 24, 2013

Guest Author

We love it when our readers get in touch with us to share their stories. This article was contributed to DIYP by a member of our community. If you would like to contribute an article, please contact us here.

Join the Discussion

Share on:

So, I’ve had this idea bouncing around in my head for a bit, and figured it might help me to get off my ass and actually try it out if I described my thought process.

Long Exposure - Shanklin
Long Exposure – Shanklin by Richard ‘Tenspeed’ Heaven, (cc-by), 5 sec  f/32 ISO 100

I’ve been wanting to get some ND filters to experiment with daytime long exposures for a while now. The problem is that I’m lazy. So when I say “for a while now”, I really mean that it’s been like 3 years.

I had previously written about using median stacks to remove noise from an image, as an easy way to remove non-static objects from a scene, and to create interesting artwork. It’s those last two things that got me thinking…

All of the Playboy Centerfolds from 1970
All of the Playboy Centerfolds from 1970

This is one of the first amalgamation images I did after seeing Jason Salavon’s incredible work. This is all twelve centerfolds averaged together. While I think it’s neat, it’s more interesting when you begin to view it in the context of a typical long exposure image.

Long Exposure Groin
Long Exposure Groin by Paul Chaloner, (cc-by-nd), 30 sec  f/22 ISO 200

Do you see some visual similarities between the water in these long exposure images, and the centerfold image?

This similarity led me to realize the main point of this entire post:

Theoretically, there should be no difference between a single long exposure image, and multiple short(er) exposed images that have been averaged together.

It helps to visualize the fact that when taking a long exposure image, the camera is averaging over the duration of the exposure. I know this is probably “Duh!” for most people out there, but for me it was a “Eureka!” moment (it takes me a bit longer to get things).

If you want to skip the numbers and get straight to the results, click here.

Looking at the Exposure

So let’s have a look at the numbers for the image, Long Exposure Groin , above. According to the photographer, Paul Chaloner, the image was exposed for 30 seconds, at f/22 ISO 200. For this discussion, I am going to assume that the ISO is already set to its lowest acceptable value.

With that in mind, we can see the relationship of shutter speed and f-stop in this table:

0.7 1 1.4 2 2.8 4 5.6 8 11 16 22 32
130 115 18 14 12 1 2 4 8 15 30  

This table shows the relationship of Aperture ( f ) to shutter speed (seconds) to maintain the same exposure (in full stop increments). The parameters for the base exposure are shown in green.

What this means is that if I wanted to have the same exposure, but wanted to change one of my parameters (either f-stop or shutter speed), I need only move left or right on the table to get the required values.

To illustrate, perhaps I wouldn’t want to shoot stopped all the way down to f/22. Instead I might prefer to shoot at f/11. No problem looking at the table for f/11 yields:

0.7 1 1.4 2 2.8 4 5.6 8 11 16 22 32
130 115 18 14 12 1 2 4 8 15 30  

I can see that at f/11, my new shutter speed would be 8 seconds. Pretty elementary so far, and really just Exposure 101.

Things get a bit more interesting when we consider how some of these are shot. There is some more information about the photo which isn’t present in the EXIF data.

Neutral Density (ND) Filters

The EXIF data on the photo shows that the image was shot at ~3:30PM local time. Just going off of the Sunny 16 rule, I would be shooting at f/16 and 1200ths of a second (at ISO 200).

So how did the photographer get his shutter speed all the way up to 30 seconds? Simple. A Neutral Density (ND) filter.

ND filters are pretty simple things, in theory. Just reduce all wavelengths of light by a certain amount.

The image above was, according to the photographer, shot with a B&W ND110 (10-stop) filter.

Fiddling with the Exposure

Now things get a little more interesting. Let’s assume for a moment that the image uses a 5-stop ND filter to achieve its exposure. Let’s also assume that we want to keep the aperture at the same setting ( f/22).

Stops 10 9 8 7 6 5 4 3 2 1 0  
0.7 1 1.4 2 2.8 4 5.6 8 11 16 22 32
130 115 18 14 12 1 2 4 8 15 30  

What we’re essentially doing is removing the 5-stop filter from the lens. To maintain the same exposure, we have to shift those 5 stops of light somewhere. Since we want to keep the aperture at f/22, this means that our shutter speed has to become faster to accommodate.

As we can see in our table, moving over 5 stops yields a new shutter speed of 1 second.

So, if keep the aperture (and ISO) the same, and remove the ND filter, then we will now have to reduce the shutter speed to one second to maintain the same exposure.

The new exposure settings are: 1 second @ f/22 (still ISO 200).

Shutter Speed is too Fast!

So we’ve easily managed to compute a new shutter speed to take the same exposed image. The problem is that now we have a (relatively) short shutter speed compared to the original. With such a short shutter speed we will no longer have nearly the same level of blur for moving objects in the scene.

We’re getting to the meat of this post now.

I contend that with the new exposure value of 1 second, that we can achieve the same level of blur by averaging the same number of frames to equal the original exposure value.

That is, the original shutter speed was 30 seconds. Without an ND filter, our new shutter speed to maintain exposure is 1 second.

I’m saying that with 301 = 30 frames @ 1 second, averaged together, we’ll get the same result as the original image.

Testing the Theory

Things are more fun when we actually start testing out this idea. So I headed out this past weekend and found a nice fountain near my home to see if I could illustrate what I’m thinking.

Here is a single frame from one of the shots:

1⁄15 sec at f/11, ISO 200
115 sec at f/11, ISO 200

I had my camera setup on a tripod, and just let the burst mode run with this (9 fps for the win!). So I ended up with 55 images of the scene.

As in my Enfuse tutorial, I first made sure all my images were correctly aligned using Hugin’s align_image_stack:

C:\Program Files\Hugin\bin\align_image_stack -a OUT FILE1 FILE2 FILE3

My tripod is a little wobbly, so I wanted to make sure everything was well aligned. Obviously this could be skipped if you’re sporting a fancy, heavy tripod.

After all the images are aligned, I just use Imagemagick the same way I did for averaging my other images:

convert *.tif -evaluate-sequence mean -alpha off OUT.tif

So 55 images × 115 = 3.6 second equivalent exposure. Here’s the result:

3.6 sec f/11 ISO 200
3.6 sec f/11 ISO 200 (mouseover to compare to single frame)

This is the same as if I had used a 5-stop ND filter, and exposed for the same result.

Here is a 100% crop to show some detail better:

crop - detailscrop - details


I’d say that the results so far are in line with what I expected. That is, multiple shorter exposures average blended together will yield the same visual results as a much longer exposure.

Cool! So now I have a way to replicate the results usually obtained with an ND filter, without having to get one.

Now, this way may seem a bit fussy, but really it was pretty simple. Most modern cameras can fire off quite a few shots per second. The setup is no different than shooting with an ND filter (you’ll still have to have a steady tripod setup, and the shot framed up and ready to go).

The only difference is that you are now firing off multiple shots instead of one long shot.

Processing is pretty straightforward once you have all the images, too. Those two commands are all that is needed (and really if my tripod was steady enough, I could have skipped the align_image_stack step).

So the next time you’re out shooting and forget (or don’t have) your ND filter, try setting up the tripod anyway, and firing off a bunch of images instead!

[Bonus] – An added bonus of using this technique is that you’ll be actively reducing the noise in your image through averaging, which is normally the opposite of running your sensor for 30+ seconds in a single go.

Something Neat

I also learned something really neat while writing this. Apparently Imagemagick can also read in video files! So I can also just setup a camera, and take a video of the scene in question. I’ll get about 30 frames per second from my camera, so a 10 second video gets me 300 frames (albeit at HD resolution). In my case, I had 435 frames to quickly average from my video:

435 frames averaged directly from HD video
435 frames averaged directly from HD video

Of course, doing it this way means your view has to be really, really still. Otherwise you’ll have to extract all those frames and align them first to keep still objects sharp.


Just to be complete, the same results can be obtained through G’MIC in GIMP. G’MIC has a command for doing average blending of all layers in the image:

Layers → Blend [average all]

I didn’t do it in GIMP because 55 full resolution images tends to choke things up (and I really didn’t need to see each layer).

About The Author

Pat David is a hobbyist photographer and a software developer, you can follow his blog here, flickr stream here and check out his GIMP tutorials here. This post was originally published here.

Filed Under:

Tagged With:

Find this interesting? Share it with your friends!

Guest Author

Guest Author

We love it when our readers get in touch with us to share their stories. This article was contributed to DIYP by a member of our community. If you would like to contribute an article, please contact us here.

Join the Discussion

DIYP Comment Policy
Be nice, be on-topic, no personal information or flames.

Leave a Reply

Your email address will not be published. Required fields are marked *

15 responses to “Faking an ND Filter for Long Exposure Photography”

  1. Tim Avatar

    Hallelujah. I’ve been doing this for a few years now, for much the same reasons.

    However, there are two differences to be wary of.
    First, you’re actually getting better image quality. As you know, ISO noise halves with every doubling of the number of frames: 55 source frames is the noise-equivalent of an ISO rating 5.8 stops lower than 200. Epic win.
    Second, you do have to be careful of the nature of motion in the image. Assume the situation and filtration are such that you’re doing 30-second exposures of passing clouds. If you have a 2s self-timer delay to let the tripod settle, then that’s going to give the effect of multiple fronts / scallopped edges to the clouds. Bad. Similarly with a waterfall: you get patterns of irregular flow gushing down in tresses which might resonate with the frequency of shooting (fps). However, if your scene is something like water flowing around rocks, then you get to choose several out of many source frames according to the precise position of spray around said rocks – errant flecks can be removed so you control the nature of the misty haze.

    Other reasons for stacking everything:
    1) maybe you don’t want to use f/22 because of lens diffraction. If your lens is sharpest at, say, f/8 then that’s 3 stops to make up therefore use 2^3 = 8 frames and you’ll equal the same exposure time.
    2) maybe the camera doesn’t go over a minute exposure
    2b) even if it does, the long-exposure sensor noise is horrendous

    1. Kenneth Younger III Avatar
      Kenneth Younger III

      Pardon my ignorance, but could you just drop down to f/8, take 3 frames and just add them vs 8 frames and averaging?

  2. joe_average Avatar

    yes, very cool stuff! glad you brought it to the community!

    very good points from tim. kudos. to belabor his point: anything with periodic motion will tend to ‘stack’ the images in a biased way (aka digital aliasing). fast motion, compared to shutter speed, causes choppiness, and will not average together.

    astronomers have been using this technique to “average out” the atmospheric turbulence twinkle for a couple decades.

    also, I have to mention that median (most frequent value) and mean (averaging) are different mathematical operations. each are very useful in image processing, but can produce very different image. so use the terms carefully.

    1. Pat David Avatar
      Pat David

      Joe brings up a good point, don’t mix up median for mean!
      I had previously used median blending to remove noise and/or moving objects from a scene quickly and easily.

      For a good comparison, my favorite piece using these methods that I have made is “Mr. Presidents” – an averaging of all of the U.S. Presidential portraits.

      Here is the purely mean average result vs. a median stack blended into the average. Same source material, two different mathematical operations that produced two very different results.

  3. David O'Sullivan Avatar
    David O’Sullivan

    I have been doing this for a while too, I do mine in photoshop, from lightroom open in layers then adjust the transparancy of each layer to be the average of the layers below it. So the bottom is 100% then 33, 25, 20, etc.

    Doing it this way means you can selectively mask parts of the image, so if you want a long exposure look but with someone standing in the image you can:


    I find around 40 images gives a smooth result, it doesn’t get smoother with more


    and about 20 images give a quite different but still beautiful look to the water



    I had always meant to write up a tutorial but never quite got around to it :-)

    1. Mark Berry Avatar
      Mark Berry

      David, can you explain your averaging a little more please, as I’m not understanding your sequence. How do you arrive at 33% from 100% below, 25% from 100% and 33% below, etc?

      1. Mark Berry Avatar
        Mark Berry

        OK I think I’ve got it, from here: http://www.davidebarranca.com/2012/03/how-to-stack-layers-in-photoshop-as-mean/. I think you’d missed out 50% for the second layer up, and it’s not an average of the layers below, it’s 100% divided by the layer position, which if you add 50% does give the sequence you show, but which is not the average of the layers below.

        If I’ve got that wrong, please let me know, because it looks like a great technique :)

        1. David O'Sullivan Avatar
          David O’Sullivan

          yes, sorry i missed 50% – thats kinda crucial! I couldn’t think how to explain it! i just made an excel spreadsheet to do the math and copy from there each time. That link you included explains it better thank i could – good find

  4. sam Avatar

    I actually tried this technique a year ago for photographing waterfalls without a tripod (forgot to pack it!) with faster shutter speeds and taking bursts of pictures. I then fused them using photoshop (auto align and then averaged or withe the transparency ratios explained on the web). The results were fine, but not as good as with long exposure and a tripod of course. Unlike your results, it could be because I was using fast shutter that is, compared to the motion and lag between shutters, produce a slightly different effect and some ghosting. That said, I did not regret it! It still better that passing the shot and keep cursing myself for forgetting the tripod.

  5. bjarnewinkler Avatar

    Pat, one should then be able to do the same in camera via
    multiple exposures – have you ever tested that?

    1. David O'Sullivan Avatar
      David O’Sullivan

      There is a bloke who does it in camera with a 4 x 5, lots of shots at normal shutter speeds. With sheet film you just recock the shutter, easy :-)
      I don’t remember where I read about it though, but yes it works!

      1. Tim Avatar

        I may or may not be the bloke you have in mind, but I did it on my first ever excursion with an LF camera: instead of 1 frame, I took 5, to make http://www.shinyphoto.co.uk/pictures/crail-harbour-rocks

        Dumped LF years ago, now totally digital, but still using align+stacking for everything I produce.

    2. Pat David Avatar
      Pat David

      I haven’t tested it, but there are two main factors for consideration:
      1. Film – this certainly can be done with film cameras, but remember that the results are additive, not average (so you have to severely decrease the exposure to as little light as you can manage, then build up the layers to an equivalent exposure. If your digital camera allows multiple exposures to the same frame, then the same rules apply as for film.

      2. Digital – I had heard that some of the Sony cameras actually can do median processing from multiple frames to reduce noise in-camera.

  6. Ed Knepley Avatar
    Ed Knepley

    Ditto the folks who have doing it for a while. I use my Nikon DSLR’s built in multiple exposure system (in answer to an earlier comment).

    The theory for why this is a valid approach is summed up by reading about time and ensemble averages in ergodic systems. Read about it here in my 2012 blog post “Artful Blurred Water”.


  7. Sameer Avatar

    great technique!
    you can put camera to continuous shooting mode and use remote release to trigger series of photos. I do it for star-trails but never thought of doing it for long exp day time photos!! Thanks!!