Digithoughts

In traditional photography, you have to choose a focal point.  Everything in front and behind that point progressively moves more and more out of focus. This process can slow photographers down and sometimes cause you to miss the action, either as your camera hunts for too long or focuses on the wrong point, leaving you with a useless image.

But, recently Lytro announced their new Light-Field camera. With a light-field (or ‘plenoptic’) camera, you can take the shot without worrying about focus and then adjust the focus to your heart’s content, later. You can see this in action here. Just click on the part of the image you want to focus (requires Flash).

This is all cool stuff but we’re still in early days of this technology…

In order to gain the ability to store all possible focal depths, we’re essentially trading resolution. As an example, Stanford University researchers used a 16 megapixel camera and, after creating ~175 pixel microlenses to create the light-field data, were left with only a 90 kilopixel final resolution (back in the bad old days of 640x480 cameras, we still had 307 kilopixels, albeit with a single focal point).

There is a solution that allows for higher resolution images but it comes at the cost of unwanted aliasing artifacts. So, do you trade resolution for “truer” images or gain resolution at the risk of more distortion in certain cases?

What’s more interesting is where this is all going…

The megapixel race is all but played out in current photography. For most users, 16mp versus 18mp sensor makes barely any perceptible difference to final image quality. Even if we bumped to 40mp, that would just mean larger files to most users who still just want a cool picture for Facebook. But that same 16 to 40mp sensor bump pushes a 90kp light-field camera to 450kp or can significantly reduce the artifacts in a higher resolution version. Continued improvements to sensors will make all the difference to light-field cameras. Indeed, Adobe already has a 100mp prototype light-field camera that’s capable of producing 5.2mp images of the entire scene around the camera.

More interesting still that there’s quite a bit of 3D information available, just based on how a light-field camera stores its information. The basic browser Lytro only allows for changing focal points, but a future version is planned to add 3D support. As the images already capture this data, the pictures we take today will be available in 3D tomorrow, as soon as the viewer arrives.

Even better is the potential for stereo photogrammetry, the technique where multiple images are combined to build a 3D model (hypr3d.com is a free example). The problem with traditional images is that they lose focus away from the focal point and so the ability to resolve accurate 3D data diminishes. But a light-field image contains information for all depths. As an example, a set of light-field security cameras, each recording from different angles, could produce a 3D model of a subject that can be rotated, zoomed, posed, etc. later to produce the best possible image. Think Harrison Ford’s Esper photo analysis machine in Blade Runner in the real world.

So, today, we have a cool camera with some fun applications. Tomorrow? Light-field cameras may completely change how we think about photography. What do you think?

Nicholas Davison Director of Web Development

Read more from the Engineering category. If you would like to leave a comment, click here: Comment or stay up to date with this post via RSS, or you can Trackback from your site.