How to Improve Low-Light Smartphone Photography? Google Has a New Breakthrough

Compiled from the Google Research Blog

Google — that top-tier tech company — never stops pushing the frontier of imaging research. After releasing what many called “the best camera phone ever made,” the Pixel XL, Google has recently made new strides in low-light smartphone photography. Here’s the story:

When Google software engineer Florian Kainz was showing off some nighttime landscape photos he’d taken with a Canon EOS-1 D X DSLR, a friend from the Gcam team wasn’t impressed. So the two of them set up a challenge: shoot the same low-light scene — smartphone versus camera — and see whose noise control held up better.

Let’s look at the Canon flagship full-frame DSLR result first:

Aperture f/2.8, shutter 4s, ISO 400, shot with a Canon EOS-1 D X Mark II and Zeiss Otus 28mm f/1.4 ZE

And how did the Google phone do?

Aperture f/2.0, shutter 4s, ISO 1600, shot with a Nexus 6P

Full moon rising:

Aperture f/2.0, shutter 1/6s, ISO 1600, shot with a Nexus 6P

Honestly, whether you look at detail rendition or noise levels — on a phone screen, the photos from the Canon EOS-1 D X and the Google Nexus 6P are hardly distinguishable.

Google’s black magic is pretty incredible! Let’s break it down:

These research results were just published on the Google Research Blog. Essentially, similar to Google’s HDR shooting mode, this new technique (the engineers developed an internal test app) bumps Google’s HDR pipeline from 10 consecutive frames all the way up to 64 consecutive frames, then applies a post-processing frame-merging algorithm to effectively suppress noise.

After pressing the shutter button, the app captures 64 photos within a few seconds and saves them in DNG format. For even better results, Florian Kainz offers an additional trick: besides the 64 normal frames, cover the phone’s camera lens with black tape and shoot an extra 32 pure-black frames. Each pure-black frame is used to simulate a black-level calibration circuit, eliminating grid artifacts caused by localized black-level variation.

Aperture f/2.0, shutter 1/10s, ISO 4800, shot handheld with a Google Pixel. The single-frame noise is unbearable — but after applying the black-level calibration (the shadows on the left), high-ISO performance improves dramatically.

Can this kind of algorithm actually improve a phone’s high-ISO image quality? Let’s first do a quick primer on the types and causes of noise.

Photography enthusiasts know: for a given sensor, the higher the ISO, the more noise you get — and the less clean the image. Specifically, noise falls into two main categories:

• Thermal noise: During long exposures (3 minutes or more), the sensor heats up due to inadequate heat dissipation, causing some pixels to stop collecting data or overflow. This type of noise typically appears as bright specks.
• Chromatic noise: At higher ISOs, electrical signal interference generates additional noise — usually purple, green, or other colored speckles that degrade fine detail.

Here are the results Florian Kainz achieved with this app:

• Aperture f/2.0, shutter 4s, ISO 1600, shot with a Nexus 6P
• Aperture f/2.0, shutter 2s, ISO 6400, shot with a Google Pixel, 64-frame merge
• Aperture f/2.0, shutter 2s, ISO 12800, shot with a Google Pixel, 64-frame merge

As you can see, after multi-frame merging (a full 32 or 64 frames), the results are stunning. The images are remarkably clean, and the detail is beautifully preserved. Each frame’s exposure is 2 seconds, meaning it takes roughly two minutes to capture a single photo — not counting the in-device post-processing time.

Zooming out to the bigger picture: across today’s major phone manufacturers, the hardware game is about pursuing ever-larger apertures and making dual cameras the new “flagship standard.” On the software side, algorithms for multi-frame merging, fine-grained subject isolation, and depth-of-field simulation are steadily advancing.

As long as you have powerful enough algorithms and a processor that can keep up, shooting clean, stunning nighttime photos with a smartphone is entirely within reach. Either way, for smartphone photography enthusiasts, this is genuinely something to get excited about.