The new iPhone 7 Plus introduced at Apple’s September product event sports two cameras, set closely side by side, as was widely rumored. Philip Schiller, Apple’s senior vice president of worldwide marketing, described the system, which pairs a 28mm wide-angle lens, just as in the iPhone 7, with a new 56mm telephoto lens.
Apple will release the phone with hardware-based 1x and 2x zoom, and a software synthesized zoom between the two. An improved digitally interpolated zoom mode handles images above 2x to 10x. A software update that will ship later will add a Portrait mode that creates photos that look like they came from much more expensive digital cameras.
Two cameras don’t just take two pictures: Their lenses collect data that software shapes into images or video. Simultaneously captures allow for a range of possibilities, many of them involving computational photography, a field that mines image data or image sensors to produce a photograph as a possible outcome, rather than a straightforward picture.
The most common use is high-dynamic range (HDR) photography, available in the iPhone for years, in which pictures taken in quick succession with different exposure settings are processed into a single image that has detail in both the brightest and darkest areas. The most prominent hardware example was the Lytro, a now-discontinued camera series that computed light rays—the direction of each captured pixel—and could use that information to produce an image that could be refocused after the picture was taken.
At launch, Apple will use the two cameras to allow photos to be taken with the native 28mm and 56mm cameras, but it will also synthesize zoom ratios between 1x and 2x by calculating data from both lenses. These won’t be manufactured images, like a digital zoom, but will extract and combine data to produce something likely indistinguishable from a true variable optical zoom.
The digital zoom has been improved—and extended to 10x—because the phone will be able to interpolate missing data by capturing more information between the two cameras. Digitally zoomed photos will still look somewhat artificial, but the results should be substantially crisper and more realistic than single-camera iPhones.
Schiller also said that developers worked on an “extra credit” project that will ship as a free update for iPhone 7 Plus owners later this year. This new Portrait mode will simulate “bokeh” (boh-keh), the sharp-focus foreground paired with blurry background that’s a hallmark of very shallow depth-of-field expensive lenses and camera systems. “The higher the quality of the bokeh, the more advanced and higher quality the photo system,” Schiller said.
The iPhone 7 Plus can accomplish this through machine-learning recognition of objects in a scene, which let it identify people, and by using images captured from the two cameras, to differentiate focal planes. The result is synthesized, but could appear fairly realistic.
The Android-based HTC One M8 introduced in March 2014
had two cameras and simulated bokeh, which had mixed performance depending on the scenes in question. Google and Apple’s object-recognition has improved dramatically in the intervening period. A pioneer in this field, Marc Levoy, released a proof-of-concept iOS app years ago,
SynthCam, that captured video frames to produce this effect; it’s much more simple with multiple cameras in a single device. Levoy retired from a professorship at Stanford to
go to work for Google.
This should be just the beginning, regardless of what Apple chooses to implement in the Camera app. With two cameras with different lens focal lengths, an iPhone can venture into Lytro territory, allowing for multiple focus points that can be selected after shooting. It also opens the potential for a flood of apps that make interesting use of the two cameras far behind Apple’s built-in Camera app, given that Apple is also offering third-party developers access to RAW image data and the cameras’ wide-color data.
Side-by-side cameras made stereoscopic photos and videos possible. While these are called 3D, they’re created just like our binocular vision: two slightly separated images that allow us to reconstruct a receding set of objects and distances. The same approach allows true 3D scanning of a static object, moving a camera around it until an app recognizes it has enough imagery to stitch together a model.
It’s also possible to compute motion details with two cameras shooting simultaneously and remove blur. That’s right: A photo seemingly ruined by someone turning their head could be snapped back into static focus.
Multiple cameras also make it easier to create automatically stitched 2D panoramas. Apple’s current Camera panorama mode is fussy and requires a stable hand and careful motion. Two cameras makes it easier for enough information to be captured that a photographer can be messier.