Light Camera: The Embodiment of Computational Photography

The Light Camera has been designed to provide the same image quality as a high-end DSLR (Digital Single Lens Reflex) camera, at the fraction of the volume. The concept is simple, but the implementation isn’t: by using 16 smartphone-like camera and lenses and combining the information they gather into a single 52 megapixel image, Light can produce excellent image quality that rivals much larger cameras, and it can also zoom in and out, despite having no moving parts.

We all know that modern smartphone can shoot excellent photos. Cameras such as the XPERIA Z5, the Galaxy S6, the LG G4 or Google’s Nexus 6P (by Huawei) can all snap excellent images in all kinds of lighting conditions. However, when you zoom in to look at the finer details, it’s easy to see that they haven’t reached the “DSLR-level” quality we would ideally want.

That has to do with how the physics of photography works. The common-wisdom is that when it comes to photography, bigger lenses and sensors are simply better. That’s somewhat true, which is why optical telescopes keep getting bigger and bigger, including the new 30-meter wide telescope from Mitsubishi.

However, bigger lenses and sensor equate to a (much) larger camera body. A higher zoom ratio means that lenses tend to get deeper/longer. And of course, all the high-quality glass built to insane accuracy specifications does cost money… lots of it.

"DSLR-LEVEL PHOTOS IN A MUCH MORE COMPACT FORM FACTOR" The premise of the Light Camera is that it can bring comparable, or competitive, DSLR-level photos in a much more compact form factor. And despite its $1299 price, it may also be more cost-effective in terms of quality/price ratio.

Here’s a photo that Light says was captured with its camera. The official gallery is located on their site.

Computational photography means that instead of trying to capture an excellent raw image from a single source, it would be possible to capture various type of information and use algorithms to reconstruct a picture that is just as good, or better. This is the idea that Light is based on, and a field in rapid expansion.

The Light Camera leverages the economies of scale of the smartphone world. Without that, it would have never been possible. Thanks to the rise of a Billion+ handset a year market, high-quality camera modules are one of the most researched technologies in the world.

The Light Camera features 16 camera modules with different focal-length, and 10 of them fire during a shoot to gather information which will be processed and combined by the on-board computer to create a final image. Because each module has different characteristics, and because their location relative to each other is known, it is possible for software to create an excellent picture. The camera can be used as a 35-150mm camera equivalent.

Also because it has different types of lenses, they acquire enough information for the Light Camera to create a depth estimation, which comes very handy to create depth-of-field effects, for example. There are tons of possibilities, but this is one of the obvious “regular-camera” equivalence.

At the moment, we have been able to play with a prototype, but Light wasn’t ready to let us take one in the field because they are still working on the software. That said, I think that the science is sound, and in theory, this can work quite well. Other companies such as Lytro have tried using computational photography technology in the consumer market with very mixed results – because there was too much friction in the user experience.

If Light can avoid falling into that trap, which apparently they will, then things are going to heat up. The good thing is that this technology can scale both up and down. With more and better camera modules, the quality will increase. But less of them, things can become smaller and affordable. In time, you could surely see smartphones with multiple lenses as well – expect more players to join this party.

Hubert Nguyen
Categories: Photo-Video
Tags: Digital Cameras, dslr