The evolution of smartphone photography

1/6The evolution of smartphone photography
The evolution of smartphone photography
Smartphone cameras have improved beyond our expectations. From the grainy VGA cameras the early feature phones came with to the latest Dual Camera stacks that can replicate DSLR-like portrait shots with perfection, smartphone photography has evolved in leaps and bounds. But there is still that one department, where smartphones still can’t match up to the DSLRs- low light photography. But Google intends to change that.

OEMs often tout low-light photography as a feature in their phone cameras, but more often than not, it’s little more than a gimmick. Usually, the smartphone tweak the camera app to keep the shutter open for longer to let in more light, which does result in well-lit night shots but even a little shake results in blurred objects. Also don’t forget the noise.

Gcam=DSLR?

2/6Gcam=DSLR?
Gcam=DSLR?
But at Google, engineers push the boundaries and get results. Similarly, Florian Kainz, an engineer working on Google’s Gcam team, the same team behind the legendary camera of the Google Pixel had been challenged by his colleagues to recreate a night shot of the Golden Gate Bridge that was originally shot by a DSLR using a smartphone camera. This is the original photo.

Is this shot from a phone?

3/6Is this shot from a phone?
Is this shot from a phone?
Naturally, he took on the challenge and managed to recreate the image using the Nexus 6P and some post-processing. This is the photo he recreated using a smartphone, giving us a glimpse of what Google could do to make night-time outdoor photography a real thing using smartphones.

Beyond HDR+

4/6Beyond HDR+
Beyond HDR+
Of course you can’t take photos like this with your smartphones yet. The Google Pixel comes the closest to taking such photos, but it’s still not close enough. Google uses the HDR+ feature which rapidly takes 10 short exposure shots and averages them to form a single image to remove blur and noise while collecting enough light to take surprisingly good photos at reasonably low-levels. But it still will be useless in very low-light, moonlight landscapes, for instance, or starlight photos.

The long and winding road to take a DSLR-esque photo

5/6The long and winding road to take a DSLR-esque photo
The long and winding road to take a DSLR-esque photo
To arrive at the photo, Kainz installed a custom Android app that allowed him to set the ISO, shutter speed and focus manually. He also used a tripod to keep the phone still. The custom app was to remove the maximum 4-second exposure time to take longer exposure shots, so as to reduce the ISO setting and thereby reduce the noise. Once the shutter button is pressed, the app waits for a few seconds and takes up to 64 frames with the selected exposure and ISO settings. It then saves the frames as DNG files which are then later exported to a PC to post-process them into one.

Apart from shooting multiple shots of the scene, Kainz also took equal number of black shots by covering the lens with an adhesive tape. Naturally, the individual frames turned out to be quite noisy, given the tiny sensor smartphones have. To stitch the photos together, he took the average of all the frames to clean up the noise and then subtracted the average of the equal number of black frames to remove the variations of the sensor’s black levels. The resulting image can give a DSLR a run for its money.

Other experiments

6/6Other experiments
Other experiments
Using the same technique, he even managed to shoot the milky way using the Google Pixel from the Californian coast where the nights get really dark and stars are the only sources of light on nights when there is no moon. However, at 9 to 10 megapixels, the resulting photos are not of as high quality as a DSLR, but the details are all there. Seeing the Pixel and the Nexus 6P achieve such results is a clear sign that the final frontier of smartphone photography is on the brink of being breached. It’s not a matter of how, but only a matter of when.

OEMs have to create a software to process the images internally- all the additions and subtractions and the averaging of the multiple photos, to make night point-and-shoot photography using a smartphone a reality.