Stacking Photo's to create an image

[SteveMi]

From Share Your Photo’s

An experiment stacking photo’s on Gimp2



The image is made of 33 consecutively taken photographs stacked together, each was then faded out to 3% visibility and the results merged together. I was hoping to take 100 consecutive shots but the batteries died on me

Quote AustinMN:

Quote Dreeamer:

The image is made of 33 consecutively taken photographs stacked together, each was then faded out to 3% visibility and the results merged together. I was hoping to take 100 consecutive shots but the batteries died on me

3% visibility may not be correct. I know it would not be in Photoshop, I'm not sure about The Gimp.

It seems obvious that if you had 4 images you wanted at equallly weighted averages, that you would set each at 25%. But that's not how it works.

To get equal weights, the bottom image would have to be at 100%, the next at 50% (half of the bottom image shows through this image), the third at 33% (2/3 of the image data should come from under this image), and the fourth at 25%.

Why? Because the transparency is multiplicative, not additive.

Let's follow what happens with the base image at each level.

First two images: 0.25 x (1 - 0.25) = 0.1875 (total would be 0.1875 + 0.25 = 0.4375; the remainder would be transparent)
Next image: 0.1875 x (1 - 0.25) = 0.140625
Final image: 0.046875 x (1 - 0.25) = 0.10546875, or less than 11% of the final image comes from the first image.

Instead, it needs to progress based on number of images, like this:

First two: 1.0 x (1 - 0.50) = 0.50 (both images have the same input to the final)
Next image: 0.50 x (1 - 0.3333) = 0.33335 (all three images have the same contribution, ignoring minor precision differences)
Final image: 0.33335 x (1 - 0.25) = 0.25 - which is what we want combining 4 images.

I found this part of stacking images so tedious that I downloaded and stacked with a program called Registax (a free download) by Cor Berrevoets. The interface is a bit clunky (OK, it's very confusing, even hostile) and the documetation leaves a lot to be desired, but I've found it works well once you get used to it's numerous quirks.

Some images I've stacked with it:


(more than 10)


(more than 30)

One more thing about Registax. I usually resize the source images to a smaller size, and I expect a deep stack (15-20 images) to run for hours, even all night.

Austin

Thank you Austin for your detailed explanation, much appreciated. I get the general idea of what your saying but i need to spend some time studying the maths in more depth, and also the technique of stacking photo’s. I accept your explanation which leads me to believe i haven’t made a stacked image of 33 photo’s at all as the process only took 15 minutes, so what have i got : as the image i ended up with is what i expected to get when i started the experiment.

Steve

[SteveMi]

Studying the maths further i can now understand how the individual photo’s don’t get equal weight, so although my image is a stacked image its not how it should be done. The correct way to stack photo’s will be as in your second description, which i need to fully understand and then try. It will be interesting to compare the results.

Thanks once again for enlightening me. steve

[SteveMi]

Quote Joe the Manatee:

Would one of you be so kind as to tell me the purpose of this process?

Sorry i can’t, i’m new to photo editing and this is the first time i’ve stacked photos. There may be many reasons for stacking photo’s, so far i’ve only used the process to merge photo’s together for effect.

Quote Joe the Manatee:

And since all three photos here are of running water, why that type of photo in particular?

The subject was water but i imagine any scene with movement in it would produce results, ie clouds, ocean waves, cars or people. I like time lapse photos of waterfalls and rivers but cannot take them with my camera (a compact digital) in the traditional way as i have no control over the shutter speed and can’t fix neutral density filters to the lens, so this was just an experiment to see what i could do.

[AustinMN]

Stacking was developed for astronomy. The purpose then was to reduce digital noise. When using film, astronomers would sometimes expose for hours at a time (occasionally, exposing the same piece of film several nights in a row, making the exposure equivalent to several days).

Trying to do the same with digital would result in 100% noise, 0% image.

By stacking many images of shorter duration, they can get the equivalent of a very long exposure, each of which has the noise of a shorter exposure. The signals combine in a complementary fashion, while the noise combines in ways that mostly cancels itself out. There are more advanced statistical techniques uszed that can reduce random sensor noise to virtually zero.

The extreme deep-space photos that have been taken with the Hubble have been a series of thousands (and perhaps millions) of short sequential exposures over a 2-week period that have been stacked using software similar to Registax. (They don't have the registration issues that I do).

The usefulness in daytime photography is that it allows the rough equivalent of a long exposure out of a series of shorter exposures. There are a variety of ways the technique could be used, such as photographing a busy street, but showing it without any cars or people. The way I have used it (and the way Dreeamer used it here) is to roughly approximate a long exposure of a waterfall.

When photographing waterfalls, a photographer will sometimes try to get an exposure of several seconds, causing all water movement to blur into just the "average" movement. The result, when done well, is a silky-smooth "bridalveil" or "drapery" effect.

The problem that can come up with digital photography is trying to get a long enough exposure. While my camera will go down to 4 seconds (which would usually be long enough), I can't stop it down far enough, even with a polarizer and the lowest ISO setting, to be able to use that 4 seconds in broad daylight.

Enter photo stacking. With stacking, you can turn 20 1/4 second exposures into the rough equivalent of one 4 second exposure, without blowing out the exposure.

The most important requirement is that the individual images have perfect registration. Every image has to line up perfectly with all of the others. This is the value of Registax, because it can find the proper registration, assuming the images are close to correct to start with. While my tripod is good, it's far from perfect, and I have proven to myself that my camera sometimes shifts images by a pixel somewhere in post-exposure processing.

Anyway, the result is something that isn't like either a stop-action shot or a long exposure - for a good deal of work. Whether it has photographic merit or not remains to be seen.

Austin

[SteveMi]

Quote AustinMN:

I expect a deep stack (15-20 images) to run for hours, even all night.

I’m struggling to reconcile this with the process i use which takes a few minutes. Is the time taken due to the Registax program, to computing power, or have i got the wrong end of the stick again and am not processing the data in the correct way.

The times and process i use

30 seconds to automatically load 33 photo’s into a stack.
10 minutes to manually set the opaque slider on each of the photos to the calculated values.
20 seconds to merge then save an image.

[AustinMN]

Quote Dreeamer:

Quote AustinMN:

I expect a deep stack (15-20 images) to run for hours, even all night.

I’m struggling to reconcile this with the process i use which takes a few minutes. Is the time taken due to the Registax program, to computing power, or have i got the wrong end of the stick again and am not processing the data in the correct way.

The times and process i use

30 seconds to automatically load 33 photo’s into a stack.
10 minutes to manually set the opaque slider on each of the photos to the calculated values.
20 seconds to merge then save an image.

Registax does a lot more than average the images together. Averaging them may be all you want or need.

Assuming 30 images, here's what I would expect:
* Converting each image to the Registax internal format (a minute or two).
* Choosing alignment points (anywhere from 1 to 15 minutes, depending on more and complexity)
* Making sure each image is aligned (registered) with all of the others. Sometimes, this requires several passes (anywhere from 15 mninutes to 2 hours).
* Performing a statistical analyisis on each pixel, so that transient items and excess noise can be ignored (30 minutes to 5-6 hours).
* Averaing together, on a pixel-by-pixel basis, the results of the statistical analysis (1 to 30 minutes).

The machine I am using is powerful, but is a little short on RAM. I can tell that much of the time is spent swapping things in and out of the swapfile, so I would not be surprised if more RAM improved the performance by a magnitude, and I would expect the time to drop by at elast half.

If you don't need all those steps (and your results so far tell me you probably don't), then there's no reason to try Registax.

Austin

[SteveMi]

Thanks for the explanation.

Quote AustinMN:

Averaging them may be all you want or need.

For the time being yes so i'll stick with Gimp, but will keep Registax in mind for the future.

steve

[arzy]

I like Panorama Maker 4, do pano with this programm is simply and Adobe Photoshop CS2 of corse.

[HaroldGarretson]

I've been using gimp and registax but so far I'm preferring gimp. It's a lot easier to use for photo stacking. Nice picture btw dreeamer.
Quality businesses for weight loss include lemonade diet, the diet solution too as the fat burning furnace. We even wish affiliate marketing as well as the tinnitus miracle since they're significant.