DxOMark awarded the Google Pixel 2 the top smartphone stills camera award with a score of 99. Now that's a DSLR beating score. Or is it?
There's no denying, smartphone cameras are on an inexorable rise. Gone are the days when Sharp introduced the J-SH04 with a 0.1MP resolution camera so that you could instantly message your images. We're in a war — a megapixel, aperture, and post-processing war to produce the best consumer imagery possible in phone.
This has metamorphosed the opposing camps of in-camera and in-post so that there is no longer a division — it is both in-camera and in-post. What's more, it's largely performed via presets. In fact, talk to a smartphone user and it's likely they will consider photography to be: point (often at one self), click, tap (to process) and Instagram Facebook or Snapchat it. That is photography.
It may well be that the sheer exposure to a vast number of consumer images will change the subconscious of what we now believe to be a "good" image. Society will control what is considered acceptable — just wait for that pro 'tog offering "shot on iPhone" weddings as a package choice, rather than as a specialist one off. Of course the Achilles heel for smartphone cameras is sensor size. Larger sensors offer better low light performance (due to larger pixels), better bokeh, and better wide angle performance.
Let's take those in reverse order. Firstly, sensor size affects crop factor such that equivalent focal lengths means that a larger sensor will have a wider field-of-view. Of course the reverse is also true meaning that smartphone cameras can adopt relatively shorter focal lengths to achieve a useful field-of-view while maintaining a highly pocketable size.
Secondly, crop factor also affects the depth of field (DoF). Use any DoF calculator (my Android favorite is Hyperfocal Pro) to see how smaller sensors have much bigger DoFs, which conversely means that bokeh, and the quality of bokeh, tends to significantly improve with larger sensors.
Finally, low light performance is frankly poor for smartphone sensors. Shot noise is inversely proportional to sensor size (assuming the same quantum efficiency); a smaller sensor means more noise. The class leading Pixel 2 I noted at the beginning has a 1/2.6" sensor which equates to 5.5mmx4.1mm. That's 38 times smaller than a full frame sensor. To put that in perspective, the last time a compact camera had a sensor that small was around 2010 (e.g. Nikon Coolpix L21). And don't just take my word for it, look at the two shots below which show individual pixels (100% zoom) from a Nikon D700 and LG G5. The level of noise from the raw imagery on the smartphone is agonizingly painful.
Of course, smartphone manufacturers know they can't compete with the larger sensors in a camera, but they do have a range of strategies to make up for this shortfall. In terms of field of view and wide angle, LG have introduced several phones with 11mm equivalent focal length cameras. A limiting factor here is lens design and notable partnerships have been struck with specialist manufacturers, such as Huawei and Leica.
However it's the last two areas where manufacturers know they struggle when shooting in-camera, which means leveraging the benefits of computational photography to achieve better results. We are now on to second generation devices that simulate bokeh (such as the iPhone's Depth Control) and they are increasingly authentic. For dealing with noise, the most common approach is to combine multiple shots to produce the best single image. Just look at the upcoming Google Pixel 3 offering computational raw files, where in-phone processing directly produces a DNG file.
We do need to remind ourselves that smartphones aren't trying to produce the best possible images, but rather the best possible looking images for low resolution (a nice example from the Northrups). As soon as we start pixel peeping the deficiencies become glaring. That doesn't matter if you are shooting for specific media where you can't see the work-arounds. If there's an adage to work by for smartphone photography, then it's almost certainly don't crop in. There's a reason why they need all those pixels.
However if smartphones can achieve such startlingly good processed results, then it raises the question as to why traditional camera manufacturers can't. Yes, there is in-camera manipulation such as Nikon's Active D-Lighting, Fuji's raw processor, or Sony's panorama, but they're weak imitators of what Samsung, Apple and Google are achieving. In short we need cameras to offer fully computational platforms.
As a photographer, post-production is an equal part of the photographic process, so if this can be moved closer to the actual act of image capture then you can begin to take advantage of automated workflows to achieve outputs that are otherwise difficult. The platform is likely to be, at least initially, Android. In fact, Android cameras aren't new and have a history dating back to at least the 2013 with the Nikon Coolpix S800c and culminating in the rather good Samsung Galaxy Camera NX which used an APS-C sensor. The cost, functionality and performance weren't quite on a par with the best of the rest. In fact the reason they used Android wasn't as a computational platform, but a social media platform. The incorporation of WiFi, Bluetooth, and NFC by traditional manufacturers has largely negated this benefit. In short, users didn't see significant enough benefits to switch to a different platform, hence Samsung's subsequent change to the also very good NX1 which used a traditional form factor.
However I'm optimistic and excited about the Zeiss ZX1 because the incorporation of Lightroom shows the direction of travel. It might not be the camera that makes the breakthrough, but it is a statement of intent. Is computational photography the future? Will we see a proliferation of startups specializing in algorithms for an open camera platform that can take advantage of on board processing to produce computational raw? Or do Nikon's Z7 and Canon's EOS R mark the traditional camera's future?
Lead image courtesy of Fernando Stohr via Unsplash, used under Creative Commons.