360Photography

Many smartphones today take great images in broad daylight. That’s no surprise – when there’s a lot of light, it doesn’t matter so much that the small smartphone sensor doesn’t collect as many photons as a larger sensor: there’s an abundance of photons to begin with. But smartphone image quality can take a nosedive as light levels drop and there just aren’t many photons to collect (especially for a small sensor). That’s where computational techniques and burst photography come in.

Low light performance is a huge differentiator that separates the best smartphones from
the worst

Low light performance is a huge differentiator that separates the best smartphones from the worst. And Google’s Night Sight has been the low-light king of recent¹, thanks to its averaging of many (up to 15) frames, its clever tile-based alignment to deal with hand movement and motion in the scene, and its use of a super-resolution pipeline that yields far better resolution, particularly color resolution, and lower noise than simple frame stacking techniques.

With the iPhone 11, Apple launched its own Night Mode to compete with offerings from Android phones. It uses ‘adaptive bracketing’ to combine both long and short exposures (to freeze any movement) to build a high quality image in low light conditions. Let’s see how it stacks up compared to Google’s Night Sight and Apple’s own previous generation iPhone XS.

The set-up

‘Low light performance’ is difficult to sum up in one number or picture when it comes to computational imaging. Different devices take different approaches, which ultimately means that comparative performance across devices can vary significantly with light level. Hence we’ve chosen to look at how the iPhone 11 performs as light levels decrease from evening light before sunset to very low light conditions well after sunset. The images span an hour-long time frame, from approximately 500 lux to 5 lux. All shots are handheld, since this is how we expect users to operate their smartphones. The iPhone 11 images spanning this time period are shown below.

7:00 pm, evening light 1/60 | ISO 100 485 lux | 7.6 EV	7:25 pm, late evening light 1/8 | ISO 250 25 lux | 3.4 EV
7:50 pm, low light 1/4 | ISO 640 5 lux | 1 EV	8:05 pm, very low light 1/8 | ISO 1250 <5 lux | <1 EV

Note that Night mode is only available with the main camera unit, not the 2x or 0.5x cameras. And before we proceed to our comparisons, please see this footnote about the rollovers and crops that follow: on ‘HiDPI’ screens like smartphones and higher-end laptops/displays, the following crops are 100%, but on ‘standard’ displays you’ll only see 50% crops.²

Now, on to the comparisons. In the headings, we’ve labeled the winner.

Evening light (485 lux) | Winner: Google Pixel 3

Before sunset, there’s still a good amount of available light. At this light level (485 lux, as measured by the iPhone 11 camera), the option for Night mode on iPhone 11 is not available. Yet Night Sight on the Google Pixel 3 is available, as it is in all situations. And thanks to its averaging of up to 15 frames and its super-resolution pipeline, it provides far more detail than the iPhone 11.

It’s not even close.

Take a look at the detail in the foreground trees and foliage, particularly right behind the fence at the bottom. Or the buildings and their windows up top, which appear far crisper on the Pixel 3.

iPhone 11 (Night mode unavailable)	Pixel 3 Night Sight

Late evening light (25 lux) | Winner: Google Pixel 3

As the sun sets, light levels drop, and at 25 lux we finally have the option to turn on Night Mode on the iPhone, though it’s clearly not suggested by Apple since it’s not turned on by default. You’ll see the Night Mode option as a moon-like icon appearing on the bottom left of the screen in landscape orientation. Below we have a comparison of the iPhone with Night Mode manually turned on next to the Google Pixel 3 Night Sight (also manually enabled).

iPhone 11 Night mode (manually enabled)	Pixel 3 Night Sight

There’s more detail and far less noise – particularly in the skies – in the Google Pixel 3 shot. It’s hard to tell what shutter speeds and total exposure time either camera used, due to stacking techniques using differing shutter speeds and discarding frames or tiles at will based on their quality or usability. But it appears that, at best, the Pixel 3 utilized 15 frames of 1/5s shutter speeds, or 3s total, while the iPhone 11 indicated it would use a total of 1s in the user interface (the EXIF indicates 1/8s, so is likely un-representative). In other words, here it appears the Pixel 3 used a longer total exposure time.

Apart from that, though, the fact that the iPhone result looks noisier than the same shot with Night Mode manually turned off (not shown) leads us to believe that the noisy results are at least in part due to Apple’s decision to use less noise reduction in Night Mode. This mode appears to assume that the longer overall exposures will lead to lower noise and, therefore, less of a need for noise reduction.

However, in the end, it appears that under these light levels Apple is not using a long enough total exposure (the cumulative result of short and long frames) to yield low enough noise results that the lower noise reduction levels are appropriate. So, in these conditions when it appears light levels are not low enough for Apple to turn on Night Mode by default, the Google Pixel 3 outperforms, again.

Low light (5 lux) | Winner: Tie

iPhone 11 Night mode (default)	Pixel 3 Night Sight

As light levels drop further to around 5 lux, the iPhone 11 Night mode appears to catch up to Google’s Night Sight. Take a look above, and it’s hard to choose a winner. The EXIF data indicates the Pixel used 1/8s shutter speeds per frame, while the iPhone used at least 1/4s shutter speed for one or more frames, so it’s possible that the iPhone’s use of longer exposure times per frame allows it to catch up to Google’s result, despite presumably using fewer total frames. Keynotes from Apple and personal conversations with Google indicate that Apple only uses up to 8-9 frames of both short and long exposures, while the Pixel uses up to 15 frames of consistent exposure, for each phone’s respective burst photography frame-stacking methods.

Very low light (< 5 lux) | Winner: iPhone 11

iPhone 11 Night mode (default)	Pixel 3 Night Sight

As light levels drop even further, the iPhone 11 catches up to and surpasses Google’s Night Sight results. Note the lower noise in the dark blue sky above the cityscape. And while overall detail levels appear similar, buildings and windows look crisper thanks to lower noise and a higher signal:noise ratio. We presume this is due to the use of longer exposure times per frame.

It’s worth noting the iPhone, in this case, delivers a slightly darker result, which arguably ends up being more pleasing, to me anyway. Google’s Night Sight also does a good job of ensuring that nighttime shots don’t end up looking like daytime, but Apple appears to take a slightly more conservative approach.

We shot an even darker scene to see if the iPhone’s advantage persisted. Indeed, the iPhone 11’s advantage became even greater as light levels dropped further. Have a look below.

iPhone 11 Night Mode On (Night Mode Off)	Pixel 3 Night Sight On (Night Sight Off)

As you can see, the iPhone 11 delivers a more pleasing result, with more detail and considerably less noise, particularly in peripheral areas of the image where lens vignetting considerably lowers image quality as evidenced by the drastically increased noise in the Pixel 3 results.

Ultimately it appears that the lower the light levels, the better the iPhone 11 performs comparatively.

A consideration: (slightly) moving subjects

Neither camera’s night mode is meant for photographing moving subjects, but that doesn’t mean they can’t deal with motion. Because these devices use tile-based alignment to merge frames to the base frame, static and moving subjects in a scene can be treated differently. For example, on the iPhone, shorter and longer exposures can be used for moving and static subjects, respectively. Frames with too much motion blur for the moving subjects may be discarded, or perhaps only have their static portions used if the algorithms are clever enough.

Below we take a look at a slightly moving subject in two lighting conditions: the first dark enough for Night mode to be available as an option on the iPhone (though it isn’t automatically triggered until darker conditions), and the second in very dim indoor lighting where Night mode automatically triggers.

Night Mode	Standard Mode

Although I asked my subject to stay still, she moved around a bit as children are wont to do. The iPhone handles this modest motion well. You’ll recall that Apple’s Night mode uses adaptive bracketing, meaning it can combine both short and long exposures for the final result. It appears that the exposure times used for the face weren’t long enough to avoid a considerable degree of noise, which is exacerbated by more conservative application of noise reduction to Night mode shots. Here, we prefer the results without Night mode enabled, despite the slight watercolor painting-like result when viewed at 100%.

We tested the iPhone 11 vs. the Google Pixel 3 with very slightly moving subjects under even darker conditions below.

iPhone 11 Night Mode	iPhone 11 Deep Fusion	Pixel 3 Night Sight	Pixel 3 Regular camera

Here you can see that Apple’s Night mode yields lower noise than with the mode (manually) turned off. With the mode turned off, it appears Deep Fusion is active³, which yields slightly more detail at the cost of more noise (the lack of a smeary, watercolor painting-like texture is a giveaway that Deep Fusion kicked in). Neither iPhone result is as noise-free and crisply detailed as the Pixel 3 Night Sight shot, though. We can speculate that the better result is due to either the use of more total frames, or perhaps more effective use of frames where the subject has slightly moved, or some combination thereof. Google’s tile-based alignment can deal with inter-frame subject movement of up to 8% of the frame, instead of simply discarding tiles and frames where the subject has moved. It is unclear how robust Apple’s align-and-merge algorithm is comparatively.

Vs. iPhone XS

We tested the iPhone 11 Night Mode vs. the iPhone XS, which has no Night Mode to begin with. As you can see below, the XS image is far darker, with more noise and less detail than the iPhone 11. This is no surprise, but it’s informative to see the difference between the two cameras.

iPhone XS	iPhone 11

Conclusion

iPhone 11’s Night Mode is formidable and a very welcome tool in Apple’s arsenal. It not only provides pleasing images for its users, but it sometimes even surpass what is easily achievable by dedicated cameras. In the very lowest of light conditions, Apple has even managed to surpass the results of Google’s Night Sight, highly regarded – and rightfully so – as the industry standard for low light smartphone photography.

But there are some caveats. First, in less low light conditions – situations you’re actually more likely to be shooting in – Google’s use of more frames and its super-resolution pipeline mean that its Pixel 3 renders considerably better results, both in terms of noise and resolution. In fact, the Pixel 3 can out-resolve even the full-frame Sony a7S II, with more color resolution and less color aliasing.

Second, as soon as you throw people as subjects into the mix, things get a bit muddled. Both cameras perform pretty well, but we found Google’s Night Sight to more consistently yield sharper images with modest subject motion in the scene. Its use of up to 15 frames ensures lower noise, and its align-and-stack method can actually make use of many of those frames even if you subject has slightly moved, since the algorithm can tolerate inter-frame subject movement of up to ~8% of the frame.

If you’re photographing perfectly still scenes in very low light, Apple’s iPhone 11 is your best bet

That shouldn’t undermine Apple’s effort here which, overall, is actually currently class-leading under very, very low light conditions where the iPhone can use and fuse multiple frames of very long exposure. We’re told the iPhone 11 can use total exposure times of 10s handheld, and 28s on a tripod. Google’s Night Sight, on the other hand, tends to use an upper limit of 1/3s per frame handheld, or up to 1s on a tripod. Rumors however appear to suggest the Pixel 4 being capable of even longer total exposures, so it remains to be seen who will be the ultimate low light king.

Currently though, if you’re photographing perfectly still scenes in very low light, Apple’s iPhone 11 is your best bet. For most users, factoring in moving subjects and less low light (yet still dark) conditions, Google’s Night Sight remains the technology to beat.

Footnotes:

¹ Huawei phones have their own formidable night modes; while we haven’t gotten our hands on the latest P30 Pro, The Verge has its own results that show a very compelling offering from the Chinese company.

² A note about our presentation: these are rollovers, so on desktop you can hover your mouse over the states below the image to switch the crop. On mobile, simply tap the states at the bottom of each rollover to switch the crop. Tap (or click) on the crop itself to launch a separate window with the full-resolution image. Finally, on ‘Retina’ laptops and nearly all modern higher-end smartphones, these are 100% crops (each pixel maps 1 display pixel); however, on ‘standard’ (not HiDPI) displays these are 50% crops. In other words, on standard displays the differences you see are actually under-represented. [return to text]

³We had updated the iPhone 11 to the latest iOS 13.2 public beta by the time this set of shots was taken; hence the (sudden) availability of Deep Fusion.

Articles: Digital Photography Review (dpreview.com)