Waterfall stacking: Can you leave your tripod and ND filter at home?

What can you do to give the illusion of flowing water (or other similar motion) if you encounter a waterfall in broad daylight but left your ND filter and tripod at home? Or what can you do if the scene calls out for a UWA like the 7-14 or the Samyang/Rokinon 7.5/3.5 fisheye, neither of which takes filters (without a special holder) in the first place?

Inspired by discussions in this thread about the pros and cons of MFT WAs and in this thread about stacking I proposed a solution where you shoot a burst at a shutter speed that you can manage without the help of an ND and tripod and then merge and align the images in PP.

Does it work? And if so how well? I had to find out so went out in the beautiful (but cold, brrr) winter weather we had today to find about the only place in town with freely running water: the mighty "Iceland Falls" (Islandsfallet) of Uppsala.

First, I mounted my 14-45 on the E-M5 and shot the scene at ordinary "freezing" speed (f/5.1, 1/640 s).

Next, I stopped down as much as I could without going beyond the point (f/8) where diffraction starts to become a problem, lowered the shutter speed to 1/250, AFd and then switched to MF, set the camera to high-speed burst mode (9 fps) and fired until the buffer was full. This yielded 16 shots that were subsequently merged and aligned, with the fringe benefit of increasing DR by two EV.

Finally, I mounted the camera on a tripod, put on my home-grown variable ND filter (a linear polarizer on top of a circular one) and adjusted it until I got the shutter speed down to one second (which was close to, but not beyond, the point at which it starts to generate an undesirable blueish tint). But you do see, as expected, the polarizer effect in the sky.

Well, what do you think? I have some ideas myself but I like to hear yours first.

I prefer the blur in the 1 sec exposure. It's smoother.

Even though the total time for 16 shots at 9 fps would be longer than 1 second, I wonder if having more of a delay between the shots in the stacking case would have improved the blur. You could automate this using an anti-shock setting.

How did you process? Aligning is clear, but exactly how did you merge?

--
http://fruminousbandersnatch.blogspot.com/

The water does not look smooth, more like quick boil. I don't think it has the tranquil quality that the single long exposure does.

It works.

But

Why not use a built in camera program like fireworks that slows the shutter speed.  That or switch to shutter priority and choose the slowest shutter speed you can do.  You can prop the camera on the camera bag in lieu of a tripod if you have to and use the built in timer and or cable release.  Their is usually something even the roof of you car with the camera bag for a little height or a book to assist.

Of course it would be alot easier to leave the ND filter in the camera bag.  Planning should eliminate forgetting even create a list for the day or a master check list to leave in the camera bag, just saying if you remember a polarizer why not the ND.

Not bad at all, but the water is better in the ND pic.  The stacked version looks good in the falls, but you see some ghosting in the river where it's moving slower.

This is an interesting idea and in parts (the "waterfall") it worked well enough but you can see a big difference in the part where the river is flowing. The stacked version simply isn't smooth enough.

If you are working in Photoshop, you can duplicate the layer, apply a motion blur in the direction the water is flowing and apply the blurred layer where necessary via a layer mask.

Sabine

--
________________
www.acahaya.com

It works, but it is a different effect...a bit more thread-like and violent--like crazy hair in the wind.  The long exposure is a softer, creamier blur. Both are cool, but they get different results.

--
Fike, Trailpixie, or Marc http://www.trailpixie.net http://www.marcshaffer.net

Sorry, I like the first,with natural water flowing but to each his or her own.

Best Regards Mike.
http://michaelgoodes.blogspot.com/
http://ekim.zenfolio.com
'Live every day as if it was your last, one day it will be'.

Looks pretty good

Interesting experiment ... but I would stick with tripod and ND filter / Polarizer for now, its easier and work like charm without much time really needed , and I can shoot at low ISO together with optimal aperture for that too.

As for lack of UWA. I would say its pretty much a issue with all mirrorless today. M4/3 had 7-14, NX have nothing, NEX now had the 10-18. which is about as wide as they goes, but being zoom they are sacrificing speed and image quality for that I shall say where typically UWA really ask for top quality image quality from the lens native.

I really do think the M4/3 need a superb 9mm ( or shorter ) with speed at the very least at f/2.8 ( and NEX / NX need the same coverage / speed )

--
- Franka -

I would sure like to see your images at (at least) 1200 pixel-height. The new viewing-interface would allow even larger pixel-heights to be viewable. Having viewed these images at "native" size as well as at odd-numbered enlargement multipliers, I think that the sampled one looks noisier (to my eyes) for more reasons than being the combination of exposures sampling only 0.4% of the total recording time.

The concept of random noise cancellation depends on the scaling of the random noise being uniform within the averaged samples. Under that condition, reduction by the square-root of N is the best case (where a 2-stop factor of 4 SNR improvement would be the goal of averaging 16 exposures).

The dominant (image-sensor based) random noise present is Photon Shot Noise (as opposed to Read Noise) - which varies as a function of the illumination of the individual photo-sites. When sampling only 0.4% of the "action", the illumination of the individual photo-sites will be essentially random in nature - and with a magnitude that clearly exceeds the Photon Shot Noise itself.

Those two individual random components will combine vectorally (as the square-root of the sum of their squares), but the larger magnitude component will dominate (due to the effects of squaring). Thus, I don't think that this sampling concept is sound for subject-matter with non-stationary luminance (such as any moving subject-matter that varies in the scene lighting that it reflects).

SHOULD READ:

Having viewed these images at "native" size as well as at odd-numbered enlargement multipliers, I think that the sampled one looks noisier (to my eyes) for more reasons than being the combination of exposures sampling only 3.6% of the total recording time.

SHOULD READ:

When sampling only 3.6% of the "action", the illumination of the individual photo-sites will be essentially random in nature - and with a magnitude that clearly exceeds the Photon Shot Noise itself.

Well, I'm going to throw a spanner in the works and say that I like the first photo best where the water looks like water.  Personally, I think the long exposure on water to make it look smooth is an over-used cliche.

Just my opinion.

Nice photos though. 

Dan

1/30s would still be hand hold-able at that focal length, right? Maybe if you only left your tripod at home, and had a 3 stop ND filter with you, you could get an image closer to the long exposure with this technique?

Yes I know this doesn't help if you intend to do this with a lens with a built in hood like the Samyang.

Hi all and thanks for your comments,

I will respond to some of you individually, but first a few general points.

Yes, I agree of course, that the stacked image looks different from the one shot with long shutter speed. The water in the stacked image is rougher/coarser. This in turn gives rise to two questions.

First, is that a disadvantage? Many of you seem to think it is. You prefer the smoother look of the long exposure. But why is that? We are all familiar with that look since this is the conventional technique and I suspect that's part of the explanation. On the other hand, I have never found it particularly convincing as an illusion. The wild chaotic water is transformed into a peaceful flow of some rather high-viscosity liquid, like when you slowly poor milk into a glass. The effect can be aesthetically pleasing nevertheless (at least sometimes, it certainly is a bit of a cliche), but then, to my mind, usually because it looks intriguing and mysterious, sometimes a bit fairy-tale-like, not because it brings to mind the scene as I saw it with my naked eyes.

Second, is the rougher/coarser structure of the water in the stacked image a necessary consequence of the technique per se. AndyGM suggests it might not be and I had already arrived at the same conclusion. In this case I limited myself to the lowest shutter speed I could get without stopping down further than f/8 (to avoid serious diffraction) and without using an ND filter, i.e., 1/250. In retrospect, I regret that I didn't shoot a second burst using the ND filter and the lowest shutter speed I could safely employ handheld, which at 28 mm and the 14-45 OIS or the E-M5 IBIS would have meant 1/15, perhaps even 1/8. After all, what the technique can potentially buy you are three separate advantages (no need for an ND, no need for a tripod, and more DR) and you can choose to exploit one of them (e.g., no tripod) without exploiting the others.

Finally, let me point to an interesting, but not entirely unexpected side-effect of the technique. The bridge above the fall has quite a bit of traffic on it and it's pretty much impossible during the day to find a long enough moment without either moving vehicles, bikers, or pedestrians appearing on it. Such was the case during my burst as well. As you can see from the shot shown below, which is taken from the middle of the burst (image 8), there is a car in the middle, just in front of the yellow building, and two pedestrians a bit further to the right. Both are present in every image of the burst and both were moving slowly (the car was turning left), yet I am hard pressed to see a trace of them in the stacked image. So even rather slow-moving objects of this kind are likely to disappear in the stacked image, whether you like it or not.

Yes, I know I can regulate the time between shots in the burst quite precisely with the anti-shock setting, but no, I don't think a longer delay would have yielded a a smoother effect. If anything the opposite. However, as I pointed out in my previous post, a slower shutter speed (which would have been possible with an ND filter but without a tripod) might well have generated a smoother look. I'll give that a try if I find the opportunity.

I used this, an advantage of which from my point of view is that it works as plug-in to LR. It does cost something but very little (a donation of minimally 2 GBP). It does take some time. I didn't clock it but we are talking minutes rather than seconds on my brand new and quite fast PC. It's quite conceivable that there are better options. If anyone has ideas about this, please speak out.

As DM points out in your prior thread, there are free versions of the utility I used that work stand-alone rather than as a plug-in to LR.

I can understand the desire for creating the smooth "smoke in a wind tunnel" effect to portray tranquility, but the boiling sandstorm cloud effect has its uses too. I can think of some scenes where the juxtaposition of rough waters with a tranquil background would result in a pretty interesting photo.

I'm happy to have this tip in the back of my mind in case I ever find myself without an ND filter.

--
Photography - It's not what you look at that matters; it's what you see.
Galleries: http://www.photo.net/photos/teru
http://www.fotop.net/teruphoto

In this case, I used the slowest shutter-speed I could use without an ND filter since I wanted to see what you can potentially achieve if you don't have the option to use one. As I point out in a previous post, you can of course go slower than 1/250 if you use an ND.

I too, dislike the stacked image.  It is a valid reason to go for a rougher look on matters of preference, but it's not my personal preference.  I usually go for the sharp, frozen look, or the super-smooth look, and dismiss anything in-between.

You will have to shoot at a much higher framerate to get a smooth look similar to the one with a longer exposure.  That leads to a ridiculous number of images to merge in PP; say 30 images per second--more work and time than I want to spend.  Perhaps needing an even higher framerate if the subject is moving faster (like the lights of a car on the highway).  Since most cameras can not shoot this fast outside of video resolution, The application seems too limited to me to be of practical use for quality stills.

I appreciate the experiment though.

I think you are on the right track here. The thing I did isn't a full substitute for the effect you get with a low shutter speed. It's different. But I am not sure that the soft, creamy, tranquil (you name it) version is necessarily preferable. It's what we are used to since the low shutter speed is the standard solution, but why should a waterfall look like that. Shouldn't it be wilder and more chaotic?