The more traditional way of taking a long exposure photograph is, as the name would imply, keeping your shutter open and exposing your sensor or film for a long time. Of course the longer you keep your shutter open the brighter your image will get, meaning that a long exposure in daylight is nearly impossible without a neutral density (ND) filter that reduces the amount of light going through the lens. There is another option though, taking several shorter shots and then blending them together with software. The idea here is that you can expose each image properly without any filter and when the blending software combines the images it will smooth out the differences giving you the same effect as a long exposure.
I've taken long exposures both ways, and I've seen and read tutorials on both techniques, but usually the blending (or sometimes called stacking) option seemed to be considered an option only if you didn't have an ND filter. So I decided I wanted to do a real world comparison of the two techniques and try to really get into their differences.
I wanted to keep it simple yet realistic. I thought the late afternoon would be a good time because the light is nice but not changing as fast as it does around sunset. I was going to be using my Nikon D7200 (crop sensor) with a Sigma 10-20 3.5 lens (I used this at 15mm, I love this wide angle but if it's all the way open at 10mm it can introduce some technical issues that I wanted to minimize). I had 2 screw on filters that I was going to be comparing to not using a filter, the Hoya PROND 100 6 2/3 stop (100 times darkening) and the Tiffen 3 stop (8 times darkening). I would be using my camera's light sensor to get the correct exposure for each set of images as the light might change slightly throughout this, kept my ISO at 100, and my aperture at f/8.
I started with the Hoya 6 2/3 stop filter and my camera read 1.6s as the proper exposure, which isn't very long but it should still smooth out the water a bit. I ended up taking 5 shots of this in case I wanted to do some image stacking of these too. Next was the Tiffen 3 stop and my camera read 0.2s, so theoretically to get the same effect I would need 1.6/0.2 = 8 images to get the same effect as the darker filter. And last I took the images without any filter, my camera read 1/40s (or 0.025s) which meant I would need 64 images to match the darkest filter. By the way, based on these numbers it seems like the Hoya was acting more like a 6 stop than a 6 2/3 stop - but that very well could have been some changing light.
And just like I planned it rain started coming down when I was finishing up, just in time for a mile long walk on the beach. I got back and went through each step of the editing process with each of the 3 sets of images. I shot everything in RAW format, which isn't absolutely necessary but there were some things I was able to correct easily that might not have come out as well in JPEG.
The first difference is pretty obvious, some filters dramatically change the white balance and/or tinting of an image. This shouldn't cause too much of an issue if you are shooting in RAW, or even JPEG if you remember to adjust your white balance accordingly. I started with some basic automatic adjustments applied across the board. I set them to the landscape camera profile, applied remove CA and lens corrections, and selected both auto exposure and auto white balance. The auto exposure responded identically to all 3 images but the auto white balance was notably different with the Tiffen image, making the tint much more magenta. But now all 3 images seem pretty spot on color-wise.
I then made some adjustments based on my personal taste. A little work on the overall exposure and tone curve, then a white balance adjustment across the board (because I think the world is bluer than Lightroom does). An interesting thing I realized when adjusting the white balance is that it's more of a relative value than an absolute, which makes sense but I had never really thought about it. I noticed that because at first I just reduced each image's white balance by 500 K, but then the Tiffen was still too yellow. I moved them back to auto and calculated what a 15% reduction to each one would be, and ended up with the images below looking reasonably uniform.
At this point I'm happy with the color and exposure of the images overall, and you can clearly see that most color issues with a filter can be resolved without much of an issue. Now for stacking the images into artificial long exposures. I used Photoshop for this, there are more involved tutorials about this out there but these are the basic steps:
- Import the images you want to blend as layers in a Photoshop document (Most of the time I do this through Lightroom by right clicking on the set of images and going to edit in -> Open as layers in Photoshop).
- Select all of the layers and go to edit -> auto align layers (this usually works great if there were some slight movements between images).
- With all of the layers still selected go to layers -> smart objects -> convert to smart object, this will take some time depending on how many layers you have.
- You will then have 1 smart object instead of a bunch of layers, select that smart object and go to layers -> smart objects -> stack mode -> mean. This will also take some time depending on the number of images there were, but when it's finished you should have a smoothed out image that you can save or export (If you're coming from Lightroom just hit save and it will create a .tif file and put it with the rest of your RAW images).
The first one is the same because that's the single 1.6s exposure, the second one is 8 1/5s exposures blended together, and the third one is 64 1/40s exposures blended together. The 8 images took about 3 minutes to blend, whereas the 64 images took about 20 minutes. By default when saving the output .tif file Photoshop is trying to preserve the layers, with 64 images the file was apparently too large so I had to discard those layers. I don't think that's a huge deal if you're still keeping the original images though.
Anyway, right away you can see that even though I calculated the correct amount of images required so that each image is technically a 1.6 second exposure the ones with more original images came out smoother giving an even longer exposure effect. This makes sense because even though the shutter was only open for 1.6 seconds total the waves didn't stop moving between each shot. So because the 8 shots probably took me closer to 15 seconds and the 64 shots took me 1-2 minutes Photoshop is blending whatever happened in the overall time using 1.6 seconds of information. It makes sense now, but it didn't occur to me until I did this.
Now I wanted to get in close and see the difference in quality. You might have to enlarge this to really see it, but stacking the images does seem to significantly reduce the grain (noise). I knew this would happen because you're effectively reducing the ISO with each additional image that the software is blending, but I never knew how significant the difference would be. You can also see that the Hoya image was probably exposed a little differently. At this point is seems pretty clear to me that you can get cleaner results from stacking than you can using an ND filter. Not to say that it's a huge difference, but it is a lot more than I thought.
Now I wanted to play with those extra images I had taken with the Hoya, I blended the 5 together and here are the results compared with the 64 1/40 second no filter image.
It seems like these 5 long exposures smooth out the water nearly as much as the 64 short ones, which is making me realize that figuring out what an exposure will look like using this technique is much more of a guessing game - probably just closer to the total time it took you to take all of the images but I don't think that's the whole story.
Here I just added some software noise reduction (in Lightroom) to the Hoya 5 image blend. I wanted to see how close I could get the image quality to the 64 image blend and at this point I think it's comparable.
In summary, this ended up being a lot more interesting to me than I thought it would. I don't think this shows definitively that one way is better than the other, it really depends on your workflow. Below I'll run through my opinions on what the advantages and disadvantages of each technique are.
Easiness: I think it's easier to throw on a filter and take a few shots than to take a ton of images and merge them later. It also has the advantage of being able to get a good idea of what the image will look like on the back of your camera. Even if you combine the methods having fewer shots will save you a time in post processing. For either method I would recommend using a tripod and a shutter release, but it is true that you can use the stacking method handheld if you want to put a lot of trust in software auto alignment.
Cost: Obviously not buying a filter will save you the cost of a filter, but it may be important to weigh the cost of additional hard drive space if you're doing this a lot as well as the cost of your time.
Quality: More images blended together will give you a cleaner image, and having your sensor exposed for less time at once also reduces noise. These issues can be overcome somewhat with noise reduction in post processing, but there is a limit and even a little noise reduction will make the image less sharp.
I personally plan to use a combination of these methods for the most part, this test actually did change my mind a bit. I think it's the best of both worlds and offers a little more flexibility, but the lessen here seems to be that using a filter isn't necessarily the better option, and it's not just about cost.
After all of that I did a little more work and decided I liked it better in black and white, I still used the 5 blended 1.6 second exposure though.