3D stereo images with two cameras

As new camera models are being introduced, the old ones can be bought at discount prices. There is a rumour currently that Panasonic is soon going to release the Lumix GX8, and I was able to get a pair of Lumix GX7 at a reasonable price. But why would anyone want to get two of the same camera?

Mounting them to a Desmond Mini Dual Camera Bracket, it is easy to set them up for 3D stereo photography. The stereo distance becomes about 140mm here, which is a bit wide, but quite usable:

Note the lenses used: I used an old and new version of the Lumix G 20mm f/1.7. You would want to use the same lens on both cameras, to make sure the images come out comparable. However, even if these two lenses do not look alike, they are optically identical, just having a different body design. See my comparison of the old/new lens here.

Panasonic Lumix DMC-3D1 review

The Panasonic 3D1 is one of the very few twin lens compact 3D cameras available. It is shown below with a Seiko diver's watch for scale:

Nowadays, many cameras feature 3D in their specifications, but they achieve this by letting you swipe the camera horizontally while shooting several images, and then stitching the images together for a 3D effect. While this does indeed give you a 3D effect, it is not a true 3D capture in the sense that the same image is captured at the same time from two different angles. Rather, if there is movement in the image while you are sweeping, you may be capture different images for the left and right frame, which will look bad. Also, this technique does not support video recording.

The Panasonic Lumix 3D1 has got two separate, identical lenses, each covering a 12.1 MP sensor. Sliding down the front cover reveals the lenses, and also powers on the camera:

In my review, I will focus mostly on the 3D features of the camera. When used as a 2D camera, it is nothing special at all, and I cannot imaging that anyone would buy this camera for 2D shooting anyway.

3D video recording with the Lumix 12.5mm f/12 3D lens

I have already written a review of the Lumix G 12.5mm f/12 3D lens. In the review, I concluded that you should not buy the lens. One of the reasons was the lack of the possibility to record 3D videos, which is a strange omission.

However, it is of course possible to fool the camera into believing that the 3D lens is not attached, in which case video recording becomes a possibility. The way to do this, is to put some plastic tape over the electrical contacts of the lens. Then, the camera will not be able to communicate with the lens, and fails to notice its 3D properties, treating it like any other lens.

For this to work, you must enable the "shoot without lens" option on the camera:

Here is a video illustrating how to do this:



Be careful, though. If pieces of the plastic tape comes off and falls into the sensor, that can cause big problems.

Video

Using this method on the GH1 camera, I was able to use much more of the sensor than what is being used when taking 3D still images. As illustrated by the image below, I cropped off the centre 120 pixel column, and used the rest to make an anaglyph 3D video:

This uses a total of 94% of the sensor area.

As I wrote in the review, when the 3D lens is used in the conventional way, only 35% of the sensor area is used. Used this way, there is an additional crop ratio of 2.4, meaning that the lens becomes a short tele lens of 30mm, just like the Sigma 30mm f/2.8 EX DN.

The way I use the lens as a video lens, though, the crop ratio is only 1.6, giving a focal length equivalent of 20mm, a slightly short normal lens. So the field of view of the 3D lens becomes more useful with the plastic tape trick.

On the other hand, the resolution of 900x1080 pixels is quite unusual, and it is likely that you need to crop this to actually use it in any sensible way.

Here is an example video. The video is anaglyph, and must be watched with red/cyan glasses.



There is still some vignetting in the corners, but this proves that it is possible to record 3D videos with the Lumix G 12.5mm f/12 3D lens.

Still images

You can use this trick with still images as well. It gives you a larger image to work with, rather than the 1920x1440 3D image pair you normally get. Here is an example image taken when using the plastic tape trick. The areas used by the 3D lens in the conventional way are highlighted with the rectangles:

Using the lens in the conventional way yields this image:

When using the whole sensor frame output, though, you can get a much larger field of view, using 1930x3016 pixels per 3D image:

This shows that when fooling the camera into not knowing that the 3D lens is connected, you can get a larger sensor area to work with. The disadvantage is that you need to compile it into 3D manually, and also, you are stuck with the portrait format, unless you crop it down to landscape format, though. You could also use the RAW file, this way.

Alternative lens

Loreo also produces a 3D lens for Micro Four Thirds. It uses the same principle: Two lenses project an image on the left and right part of the sensor, and they must be combined into a 3D video. Unlike the Lumix 12.5mm 3D lens, though, the Loreo lens has a more sensible distance between the lenses, the stereo base, giving a more pronounced 3D effect:

The Loreo lens uses the same principle as the Lumix G 12.5mm f/12 3D lens. It projects two images side by side on the sensor. The big difference, though, is that the Loreo lens has a much larger stereo base, and uses mirrors to project the images into the relatively small sensor. The Lumix 3D lens, on the other hand, does not use mirrors at all, and places both lenses inside the sensor area.

So you end up with the same problems when using the Loreo lens as well: You are restricted to the portrait format, both when using still images and video. Of course, you can crop the images down to a horizontal landscape format, but then you lose quite a bit of resolution when cropping.

Poor man's 3D

3D is popular nowadays. Panasonic have launched their first 3D lens, which is probably not very useful. It only works on newer G-series cameras, and can only be used for stills, not video. And the stereo base, the distance between the two lenses, is only 1 cm. This is suitable for macro images, but too small for most other uses.

If you're photographing still objects, you can take 3D like images with your normal gear. It's just a matter of taking two images with the camera spaced some distance apart.

In fact, this is quite easy if you have an Arca Swiss style quick release plate. This enables some sliding from side to side when the camera is mounted to a tripod. Here is a picture of the Panasonic GH1 with a quick release plate mounted. On the table is another similar plate:

The plates on the picture are not actually produced by Arca Swiss. They are named Benro PU-60.

Using the quick release plate to offset the camera, I obtained these two images of the same subject. The stereo base is about 3cm.  The video shows how easy it is to slide the camera sideways for taking the stereo images:



To avoid having different exposures in the two images, I set the white balance (WB) manually. I also used Manual (M) exposure, and set the shutter and aperture manually.


Now, the next task is to compile these two images into one single image suitable for stereo viewing.

One way is to use different colours.  The image must then be viewed using 3D glasses with different colours. I used the Gimp script "script-fu-make-anaglyph", which does this very quickly. Here is the stereo image for cyan/red glasses:

For red/blue:

And for red/green:

Another way to view the stereo image, is to place them side by side as a pair. This is called cross-eye viewing, and takes some practice. You need to look at the left hand image with your right eye, and at the right hand image with your left eye.

Click on the image below to see it enlarged on your screen.  Put your finger between your face and the computer screen, about 10-20cm from your eyes.  Then focus on your finger, and see that the two images merge and become three.  Focus on the centre image.  If you can do this, it can be quite rewarding, as the stereo effect is very good.

Yet another way, is to compose the two images into an animated GIF:



And here is an animation consisting of three separate images:



Conclusion

This method works well with images of still life. But for moving objects, it is of course impossible to take the same image twice. And the same goes for videos.

Just how to present the 3D images is, as far as I know, not well standardized.

Lumix 12.5mm 3D lens

Ok, so this lens has just been announced, and obviously I have not tried it. So this is based on what I have read about the lens so far.

Since writing this, I have made a review based on using the actual product.

Contrary to what some people say, I actually find this lens pretty interesting. Some say that Panasonic should rather spend the time developing "real" lenses. However, I think that this lens might have some real creative potential, and it also serves to bring more interest into the Micro Four Thirds system. So what's not to like?

Focal length

The lens is specified to have a focal length of 12.5mm. However, don't let that fool you into believing that this is a wide angle lens. Far from it. The two separate lenses project image circles that are smaller than the full sensor. Hence, there is an additional crop factor to take into consideration.

The field of view is comparable to a 65mm lens on traditional 35mm cameras, corresponding to 32.5mm on Four Thirds cameras. Hence, this is a long normal lens. Observing that 32.5 / 12.5 = 2.6, we can conclude that this lens has an additional crop factor of 2.6, beyond the crop factor of Four Thirds lenses, which is 2.

On the Panasonic Lumix GH2, the lens produces stereo images of 2048x1536 pixel resolution. With the GH2 having 4602x3456 pixels, this means that each of the two stereo images comprise 20% of the image sensor. Here is an illustration:



Aperture

With a fixed aperture of f/12, the lens gives a deep depth of focus. Which is good for a lot of applications. On the other hand, at f/12, you will need to push the ISO up unless being outside on a sunny day.

3D effect

Obviously, the 3D effect is achieved by having the two separate lenses seeing slightly different perspectives. The lenses appear to be very simple constructions, with four lens elements in three groups.

The stereo base, the offset between the two lenses, is 10mm, corresponding to about half of the sensor width. This looks like a very short distance, and I am unsure if any significant 3D effect can be achieved with this short lens offset. Perhaps the 3D effect can be boosted in post processing?

Usage on non-compatible cameras

When used with a 3D compatible Micro Four Thirds camera, you can probably use the lens in a seamless fashion. The viewfinder will probably not show a confusing picture like the illustration above, but rather one of the subsections, to make framing easier. Compatible cameras at the moment include Panasonic Lumix G2 and GH2.

With an older, non-compatible camera, like G1, GF1, etc, I'm guessing the lens can be used like all others, with the exception that the two lenses will not illuminate the whole sensor. You may need to set the autofocus point to somewhere off centre, for that reason. Perhaps you will also need to change the exposure mode, since parts of the sensor are not exposed.

Of course, if the lens is used with non-compatible camera, i.e., anything older than the G2 or GH2, then you must convert the image into 3D in post processing.