Creating 3D Images for On-Screen Display

There's an odd disconnect in the world of 3D imaging. Many of us regularly enjoy watching visually stunning 3D movies both in cinemas and also on television. However if you try and research creating your own 3D images you are led either into the highly technical space of professional production using complicated equipment, or at the other extreme you end up reading a lot of rubbish about squinting at pairs of postage-stamp images to "try and get a 3D effect". While I don't want to be unkind, the latter is completely out of touch with the target environment I am thinking of, a 3D-enabled large screen television. Such devices are now relatively common, and there ought to be a recognised process for creating suitable images for them.

As it turns out, it's perfectly possible and relatively easy to create stunning 3D images which will display at the full resolution of the target television. With a little discipline and practice you can do so reliably with any camera, and even hand-held.

This article explains the process, as an attempt to bridge this odd information gap.

Setting Up the Camera

You can create decent 3D images with any capable digital camera, following the process below. If you are confident with your technique you can even work with JPEG-only sources, skipping everything that I say about developing the RAW files. However, for best results it's ideal if the camera shoots RAW, has a built-in electronic level ideally visible via the viewfinder, and directly supports multiple aspect ratios.

Camera set-up is fairly straightforward, and follows standard practice for capturing high depth of field static images. On my main Panasonic cameras I have a custom setting for 3D configured as follows:

• Aperture priority, with a starting aperture of f/8
• Level display ON
• RAW + fine JPEG output
• 16:9 aspect ratio
• A distinct "picture style" (used to enable automatic sorting of the images from others, as it doesn't affect the RAW files)

Choosing the Subject

While what you capture is very much up to you, if you're working with a single camera following the process outlined below, there are some constraints on the best subject and framing.

Although it's not impossible to create meaningful 3D images with limited depth of field, as a general rule they work best if sharp front to back. You will therefore need to ensure that your lighting, aperture, and primary focus point are such that you get all the image in focus, unless you explicitly want something at one of the extremes blurry. Also try to avoid foreground objects which break the frame on the sides or at the top, as this produces an odd effect which reduces the perception of being "in" the image when you view it.

This is a multishot technique, and is quite sensitive to movement, especially in the foreground. It works well for static subjects such as scenics and architecture, poorly for moving subjects, although it tolerates a small amount of movement in the background. One thing to watch out for is your own shadow, especially if you are working with sunlight behind you at golden hour, as this is one thing which will move between shots and needs to be hidden.

The process will work at any aspect ratio, but the viewing experience is best if the image roughly matches the target display, which will be 16:9. If your camera supports pre-visualisation of specific aspect ratios setting it to 16:9 will work best.

Taking the Shots

The actual shooting process is pretty straightforward, and not dissimilar to that for a stitched panorama, but there are a couple of differences and it's worth practicing the key move. The following assumes you are working hand-held on a scene where that's appropriate in terms of shutter speed:

1. Face the scene straight on if you can, with your body parallel and your feet perpendicular.
2. Frame the left-hand shot first, making sure that anything you want to include on the right-hand edge is comfortably inside the frame. Allow some space on the left. Generally it's sensible to zoom out (or position yourself) so that there's a little space for cropping all around the shot, as the alignment process will inevitably lose some of the edge of the images and it's easier to crop right at the end.
3. Take the first shot. Use the level to ensure that the camera is horizontal.
4. Now without moving the camera or your feet reposition your upper body about 6-7cm (3-4in) to the right. The important difference to shooting a panorama is that this is a sliding movement, not a rotation. You must still be absolutely straight on to the scene exactly as you were for the first shot. If you have got this right a tiny strip of the image will have changed at the left and right images, but nothing else. The actual distance is not critical, but essentially you want your left eye to end up roughly where the right one was previously. 
5. Use the centre line of the level as a marker to make sure the camera hasn't titled up or down between the shots, and to make sure it's still horizontal. If the camera doesn't have a level you can do the same thing by checking something near the top of the bottom of the frame, but this is slightly more difficult.
6. Take the second shot.

If you are working on a tripod what you ideally need is an Arca-Swiss-style clamp and a long plate, such as the Gitzo GS5370LC Quick Release Plate Long C. You set this up with the camera roughly centred on the plate and facing at right angles to it, then orient the tripod head so the plate is horizontal and parallel to the scene. The two shots are then simply taken with the clamp towards one end of the plate, and then slid through to the other end, although with a heavy camera you will need to check the level and may need to make a fine adjustment for each shot.

Developing the Images

Develop the two JPEG files as you would normally, applying any required adjustments for exposure, colour, sharpness and so on. The important thing is that the settings should be identical for both shots. Don't crop yet! Export to a pair of full-sized, top-quality JPEG files. If you want to rename them to indicate which is left and which is right that's fine, but if you have followed the "left first" shooting strategy then this is not strictly required.

If you haven't already done so, download the brilliant little program StereoPhotoMaker ("little" is right – it’s a single 2MB executable file and doesn't require any formal installation). It does what it says on the tin – takes a pair of images, aligns them, and saves the result to various 3D formats. This does all the "heavy lifting" of the process.

Start StereoPhotoMaker and select File/Open Left/Right Images.... Navigate to the right directory (the program is very good and remembers your previous location), and select the images (hold down Ctrl after selecting the first/left file to also select the right-hand file). Click OK and the images load.

From the Adjust menu, select Auto Align. If it can, the program works out what adjustments are required to align the two images. If your technique is good then the various adjustments should all come in at around 1 degree or less. Click Close to close the alignment dialog.

Now you can preview the image to see if it works. My preference is the "flashing" mode, which quickly alternates between the two images. You can invoke this by selecting Single Image View / Flashing from the Stereo menu, or press Ctrl+F11. At this stage you can see if you have any major problem such as a moving object or alignment problem. To return to side by side view, choose Stereo / Side-by-Side / Side-by-Side or press F9.

This is the point to crop the image. If you haven't done so before, from the Edit menu select Crop / Free Cropping Option..., and in the dialog set "Keep Aspect Ratio" and set an Aspect Ratio of 16:9, and click OK. Then to crop the image to that shape:

• From the Edit menu, select Crop / Free Cropping
• Move the mouse to the starting point of the crop
• Hold down the mouse button and drag the crop to the target opposite corner
• Click inside the selected area to apply the crop. If you want to cancel, click outside or press Esc.

Finally save the image. For most 3D televisions you want the .MPO format, which bundles the two images together in a single 3D file. From the File menu, select Save MPO File..., check the output directory and click Save.

You're done. The .MPO file should display just like a .JPEG file in standard image management software such as XnView, and you can copy or move it as you need to. If you want to keep the settings then as well as the JPEG files you need the files in the alignment sub-directory which StereoPhotoMaker creates.

Copy the .MPO file to your TV, initiate displaying it, and the TV should automatically switch into 3D mode. Pop the 3D glasses on, and lo and behold you're "in" the picture.

Examples

The best way to assess this technique is to try it for yourself. However if you want to check whether it will work for your TV before investing too much effort, feel free to download the following files and try and display them on your TV:

Santorini.mpo The Crane.mpo