NiSi Neutral Density Filters Versus Mean Stacking: Which Is Better?

In 2016, David Strauss wrote an article on Fstoppers about how purchasing an ND filter holder set might be a better option. I, being the smartass that I am at times, left the following comment, "Or you can be really cheap and just mean stack exposures :P, plus it prevents long exposure noise." Without doing any actual comparisons between the two, I had made up my mind about filters and decided against them. Recently, however, a close friend of mine, Imran Mirza, asked me to keep an open mind and give neutral density filters a try. For that reason, I have been testing some filters from NiSi over the last few weeks. In my latest video, I compare using neutral density filters to using Photoshop techniques such as mean stacking. 

Mean stacking can be a great way to get a long exposure type look with moving subjects like a body of water or clouds in the sky. The added benefits of this method are that, for one, you're not required to spend any money upfront, also your images will have significantly less noise. On the surface, this method seems like the perfect option, however with companies now producing significantly better filters, how do both of these methods compare? Do filters still hold a place in the photography industry, or are they just too expensive and not required to get the desired look? 

Check out the video for the full comparison.

Log in or register to post comments

22 Comments

Very interesting but, honestly, I expected the results you got.

Usman Dawood's picture

I didn't expect it, maybe that shows my ignorance. The detail and sharpness was really surprising.

I don't think so. My expectation was due to results I've gotten from other Photoshop automated processes. They tend to be hit and miss.

Edward Porter's picture

Clickbait article - you at least have to summarize your opinion, not force us to watch your youtube to find the content...

Usman Dawood's picture

Point taken, will do that from now on.

In this case, you're probably right but, in a lot of cases, there are pros and cons that may be weighted differently by various people. And, even if he gave his opinion, I would want to watch because I might disagree. Again, that's not really the case in this instance.

I believe if you take more shots on the mean stacking, you'll get more buttery smooth water. I'd at least double the shots.

Usman Dawood's picture

I've tried a 100 stack before and even that still leaves you with fuzzy and rough looking water. Long exposure is still long exposure and light doesn't travel in frames. Personally, I was rooting for mean stacking. This was a bit of an eye-opener for me.

barry cash's picture

Filters win I agree, also mean stacking 100mb files sucks....so the filter route is the only way to go for me!

Mike O'Leary's picture

I'd just like to add; Lee Filter's 82mm adapter ring is $56.00 or £41.99......for a piece of metal. NiSi's adapter rings come with the base holder system. This is the main reason that I went with NiSi. That and the fact that the kit comes with a circular polarizer which fits into the system before the ND filters and has a roller mechanism so you can spin it easily. Seriously, what the hell are Lee thinking?

david kidd's picture

Thanks man, I choose ND filter for long exposure look, but like the video.

Why mean stacking? Median generally yields better results when it comes to removing things like cars, waves or noise.

Usman Dawood's picture

Median is ok but it's still worse than using filters. It generally leaves weird and ugly artifacts in the images. There's no great substitute for long exposure, light doesn't travel in frames.

Tamas Nemeth's picture

Expecting that 23 expo will even out the same way a 1000 times ND filter would be is a bit optimistic.
Obviously it is PITA when it's about file size and post processing, but for me - often borrowing lenses with ridiculous filter size I don't have ND filter to fit - stacking is often the only option. And when enough sample (I would never go under 100 exposure when evening out clouds and water) is taken, it can give pretty good results. By the way, I use imagemagic to even out processed TIFFs. Photoshop cannot handle that amount of data on my maxed out macbook pro I need for this. Sometimes even with imagemagic I need to do subsets of images at a time, and then even out those batches.
https://500px.com/photo/233701767
https://500px.com/photo/235703193

Usman Dawood's picture

Showing an image individually can never truly express technical differences between methods. Everything is relative. I’m confident an actual long exposure will still look more natural and smoother than a stack because light doesn’t travel in frames.

Also shooting with this many images and spending this amount of time is impractical. The cost in time isn’t worth it when you consider the alternative.

Having said that, this is a fantastic image so in essence it was worth it :-).

Thank you for sharing.

Tamas Nemeth's picture

Originally I shared your opinion, but because of certain conversations I thought I'd give it a try (heck, I don't want to spend another $1500+ on filters I'd use with borrow lenses only), and the outcome was pretty OK.
I wouldn't call the posted samples fantastic, the scenes and the lights were quite mediocre, but for the testing purpose it was fine. I got better results than expected. Obviously it would have been different with the scene you brought up in your example, but that sort of scene is not my main interest anyway.
But as you pointed out in your video, the ND filters' frequency response is not constant (and often it is different on different area of the filter), which means that if someone really wants to take it seriously, one should make an LCC profile (which would mean at least 3-4 times longer second exposure) for the given setup, and then the time spent on the scene gets really long. For example I shoot quite ofter 10+ minutes, which mean an easy 1h+ with all the black frame and LCC shots.

So I'd say there is no magic rule. Some scenes might require ND filter, some can be done with mean (or even median) more easily. The computer can do the averaging while the photog is doing something else anyway.

Usman Dawood's picture

Already made a video about creating an LCC profile :P.

Thank you for your thoughts really appreciate the discussion.

I really liked the video for a few reasons. First, you were willing to try another method even though you had a strong negative opinion of it, second, you did real world tests in more than one situation, and third (and most importantly imo), you admitted that your previous assumption was wrong based on the results. More than just a video about filters, this is a good example of keeping an open mind and the process of self-improvement as a photographer.

Usman Dawood's picture

Brian, I really appreciate that kind comment, thank you so much. Sometimes it can be a struggle to keep an open mind and recently I've been trying extra hard to I remain as objective as possible.

Thank you again, please keep in touch.

It’s interesting to hear someone preferring a post production method refer to themself as “a purist.” Generally, those who refer to themselves in such a way are against “photoshopping” and believe as much as possible should be done in camera.

On the results.. has anyone tried using Both of the two methods together? With maybe four 3 second exposures mean stacked?

Usman Dawood's picture

Maybe purist was the wrong word to use.

I have and long exposures with even just 2 or 3 images stacked, takes significantly less time overall and looks a great deal better than anything a mean stack can do.

Justin Ho's picture

Thanks for the video. Appreciated the comparison (this is something I've thought a lot about before as well. There's a few things you didn't mention but are also worth mentioning. Mean stacking can give you an equivalent exposure. So if you want a 10s exposure, you take 10 x 1s exposures and mean stack them. The real problem with mean stacking is that if you get to the point of using a 10 stop ND filter, the number of exposures you'd need in order to get an equivalent exposure would be impractical for most photographers. E.g., in your freeway example, the exposure time with the ND filter was 25s. In order to get an 'as close as equivalent' exposure, you'd need 2000 exposures at 1/80s of a second (your mean stacked exposure time) to get 25s worth of light. That's not really practical anymore. Agree on the advantages of mean stacking for noise reduction. Another advantage you didn't mention is that if you have shaky photos, you can take out individual frames in mean stacking; if you're just using an ND filter then that entire exposure is ruined. To figure out how many shots you need to be equivalent to an 'x' stop ND filter, it's 2^x exposures. So mean stacking works if you want to emulate anything up to a 5, maybe 6 stop filter (32 or 64 exposures). Maybe 7 stops. But above that it's impractical. So the recommendation could be that for those interested in finding out what ND filters can do for their work, mean stack to see what the emulated results would be up to a 7 stop filter, but the first ND filter you should really buy is a 10 stop since it's impractical to emulate a 10 stop.