PHOTOSYNTHESIS IN NATURE  Nature & outdoor photography



Stereo photography


Thanks to our binocular vision, we can percieve a sense of depth, but all of that depth is lost when we take a photo, since a picture is only two-dimensional. However, by making stereo photos, we can partly recreate the depth of a scene, even if it is shown as a flat image. Stereo photos are combinations of two separate images with each a slightly different perspective, and some examples of stereo photos can be found here.


Viewing stereo photos


Example of a stereo photo taken with one camera =>
[first photo] [second photo] [red/cyan anaglyph] [rocking]

Viewing stereo photos can be done by several techniques, and here are some of them:


• Colored anaglyphs. In this case, both photos are colored differently, usually red and cyan. By using special colored glasses, different parts of the anaglyph are blocked for each eye, resulting in a stereo image.


• Polarizer. This is similar to the anaglyph technique, but in stead of using different colors for the two images, different directions of polarization are used for the two images. Also in this case special glasses are needed (with the two glasses being two perpendicularly oriented polarization filters), as well as special devices to create the polarized images to look at.


• Rocking image. In this case, the two pictures are turned into an animation which continously switches between the two photos, creating a 3D effect.


• Parallel viewing. Here, the two images are placed next to each other, with the right image at the right side, and the left image at the left side. The trick is to get your eyes to point at infinity, but to focus at the image closer by. When that works out, you get a stereo image. Special devices exist to make this easier to achieve, by blocking the left image for the right eye, and vice versa. In this way, the eyes are less tempted to point at the image. A disadvantage of this method is that the images can't get much bigger than 65 mm, since bigger images would mean the eyes have to diverge to get the two images to overlap. Since eyes never diverge in normal situations, this is pretty difficult to achieve. In the case of bigger images, you can look at them from a bigger distance to prevent the need for diverging your eyes, but then you loose the image details because of the longer distance, so there is no advantage of using bigger images.


• Cross-eyed viewing. Again, the two images are placed next to each other, but now the right image is placed at the left side, and the left image at the right side. Here, the trick is to look cross-eyed at the stereo photos so that you more or less see three images, where there is a third one in the middle with an overlap. Keep your eyes at this position where you see a third picture in the middle. After a while your eyes will have adjusted focus and you'll be able to see the stereo photo with depth in it. It might take a while to get the hang of this technique, but once you have succeeded, it is quite easy to switch to stereo viewing without the help of any devices, which is why I prefer this method.

The same stereo photo => [parallel view] [cross-eyed view]


Creating stereo photos


There are several ways of creating stereo photos, and I'll describe the three that I have used below.


Creating stereo photos - one camera


In this case you simply take two pictures after each other with the same camera, making sure to keep the focus at the same spot in the frame, and keeping the distance between the two camera positions about 65 mm most of the time (more on the distance between the photos below). Succeeding with this is not always that easy, and I always take several pairs of photos to increase the chances of getting a good pair. But this becomes a lot easier the more you do it. Also, if your camera has a level indicator, this is a good opportunity to use it, so that the two photos are level. By cropping (and, if needed, rotating) your photos you can correct small mistakes on the computer. However, big mistakes are not easy to fix, so it pays off to get it done right as much as possible in camera. Of course this method does not work very well with moving objects in the photo (like moving branches or waves), since the two photos will be taken at different instances.

An example of a stereo photo taken with a single camera


With a single camera, it is also possible to take a stereo photo from a moving object. It is a bit tricky to get the duration between the two photos right, which of course depends on the speed of the object. Below is an example which was taken from a moving boat.

A stereo photo taken from a moving boat with a single camera


Creating stereo photos - two cameras


The method with one camera works pretty well, but, as mentioned above, it has some disadvantages:

My stereo set-up with two cameras

• It's not always that easy to keep the camera focussed at the same object for both photos. You can see this in the first example above with my hammock, were there is a very slight misalignment between the photos. (However, this is only visible when looking at the rocking animation. It is hardly noticeable when viewing these photos with the other methods.)

• When doing this handheld, it's not easy to keep the distance between the two camera positions about 65 mm and keep the camera level for both photos.

• Moving objects are, of course, impossible to capture with a single camera.


So an alternative way is to attach two similar cameras and have them taking a picture simultaneously. I could get my hands on two affordable second-hand cameras, and attached them on a sturdy aluminum plate (see the photo to the left). The distance is a bit more than 65 mm (I think it is about 85 mm), but that is no problem as long as the subjects aren't too close to the cameras.


Unfortunately, these cameras don't take old fashioned cable releases, so I just have to press both buttons at the same time. But this actually works really well (especially in combination with the self timer on the cameras), so it is hardly a problem. Below is an example taken with this set-up.

En example of a stereo photo with moving objects. It was pretty stormy that day, so there were waves and the trees were moving around a lot


Creating stereo photos - distance between the photos


As stated above, 65 mm is usually taken as the standard, but when taking photos of more distant objects, larger distances between the photos work fine as well and can even enhance the stereo effect. Below is an example of this. The forest on the other side of the lake was pretty far away (about 100 meters), and taking a stereo with only 6 cm distance between the photos gives only a very moderate sense of depth. But the stereo effect gets a lot clearer by increasing the distance between the photos, up to a certain point. For me, 1 m works optimal, and 3 m still looks good, but my brain can't create a good picture from the 6 m version.

Of course, increasing the distance between the photos is only possible when having no objects in the foreground, otherwise the photos will look weird.

Increasing the distance between the photos => [6 cm] [30 cm] [1 m] [3 m] [6 m]


Creating stereo photos - points of overlap in the photos


Different points in a scene can be taken as the points of overlap, but the results will be similar since our brain will adapt to the different situations, as can be seen in the two examples below. In the first example the middle of the foreground is kept at the same relative place in the two pictures (red rectangle), which means that the background shifts in position between the left and right photos. In the second example the middle of the background is kept at the same relative position and the foreground changes position in stead. Both versions work equally well as stereo shots though, so there seems to be no reason to prefer one or the other.

Different points of focus => [foreground] [background]


Creating stereo photos - synthetic stereo (focus stacking)


Focus stack of a dead weevil => [focus stack] [rocking image]

With the technique described here, you can create focus stacks of very small objects by taking a bunch of photos with the focus slightly shifted for each photo. With this technique you can get large depth of fields were it otherwise would have been very small.


If you use Zerene Stacker for your stacking, then there is a neat extra function called synthetic stereo (I guess other stacking software probably also can do this, but I have no experience with that). The software will then shift each photo in the stack sequence very slightly sideways, creating an effect which is very similar to what it would have looked like if you actually would have taken the focus stack at an angle. Of course, this doesn't work perfectly in all cases, but it very often gives remarkably good results.


On the right is an example of this, which is a focus stack of a weevil with 35 pictures combined. The rocking image is a combination of 21 different focus stacks with each a slightly different amount of shift applied to it. And below you can see the stereo photo as a cross-eyed stereo.



Synthetic stereo photo of a dead weevil


As said, with this technique it is possible to vary the amount of shift in these stacks, and it is a bit of trial and error to find the optimal shift for each stack. Two factors play an important role when finding the right shift. First, you want a shift that is large enough to create a good sense of depth in the stereo photo. Second, using a shift that is too large will create stacks that have weird artifacts. So you have to find the optimum shift which gives enough sense of depth, but does not give artifacts, and this depends very much on the magnification and the number of photos used for the stack. In general, the more photos that are used for the stack, the better the results are.

I usually let the software create several shifts and simply combine each shift pair to a stereo photo and determine which one looks the best. Below is an example of a fungus with stereo photos with different shifts, where the last example starts to show artifacts.

Stereo pairs with different shifts => [smallest shift] [larger shift] [even larger shift] [largest shift]