We don’t get the Pixel range in South Africa and that’s okay. We get every other leading smartphone camera and, no matter what the marketing says, they all take pictures in much the same way. The iPhone, for instance, has been doing HDR trickery since the iPhone 4 and Samsung delivered live view HDR all the way back on the Galaxy S5. Google’s version of stacking different exposures of the same scene to better balance shadows and highlights is, and I can confirm, the best implementation of the tech so far. This focus on software and optimising of that software is the point of the recent Google announcement.
You see, the Pixel has neither the biggest sensor nor the brightest, or even the clearest, lens. It does, however, have a bit of custom silicon built into the image sensor chip called the Pixel Visual Core. First debuted as an initially unused feature on the Pixel 2, the Visual Core is an SoC (system on a chip) that contains a few image processing cores, a coprocessor core and RAM. The RAM is for buffering images – which helps enable the Top Shot feature that uses AI to suggest a better image in the buffer if someone maybe blinked when you hit the shutter key.
Building memory into, or rather onto, the image sensor is a Sony trick that they first used on the excellent Xperia XZ1 (the best device Sony has ever produced, in my opinion) and Samsung copied on the Galaxy S9. We were led to believe that sensor design was to enable 960 frames per second super slow-mo, but having a large image buffer allows for much more processing. HDR stacking and capturing fast motion is very easy for AI to execute with a large buffer.
Huawei took the path of most resistance in its approach to computational photography, relying instead on the extra AI processing powers of the NPU (neural processing unit) inside the Kirin 970 to do the heavy multi-frame lifting. This allows the company to achieve similar time warping frame rates, as well as get its automatic scene recognition on in the Master AI mode. Apple does something similar with the neural engine inside its A11 and A12 Bionic (nomenclature for its AI-powered silicon) processors, but without the 960fps or scene recognition functions, so actually not that similar at all.
Machine learning and AI is huge when it comes to making the tiny image sensors punch way above its weight. Qualcomm’s Hexagon DSP (digital signal processor) is the most widely distributed version of these co-processors and Samsung uses its own riff on that technology on the Exynos family to achieve the same effects. Like I said before, all the major phones use the same artificial intelligence tricks, so differentiation comes down to the actual software image processing and image output that the manufacturer prefers.
Huawei enlists the help of Leica to lend its signature texture to the final products. Samsung used to be known for its vibrant colours but seems to have nailed a more natural look on the Galaxy Note 9 and further enhanced the processing with a surprising software update. LG tries to play the neutral game but has followed its compatriot into sizzling vibrant territory on the G7 ThinQ. The less-popular Korean company relies almost solely on the Qualcomm Hexagon and also uses the tiniest sensor of the bunch, but offsets that loss with a glass lens element when all the others use plastic throughout.
I happened to have the P20, G7, iPhone 8 and Note 9 on hand while watching the Google event and thought a tricky lighting situation test would best show the differences.
Immediately obvious is the brighter f/1.5 and f/1.6 respective apertures of the Note 9 and G7. I threw the iPhone SE into the mix to show old school processing and it landed in at least the same colour reproduction WhatsApp group as the P20 in that handheld long-exposure Night Mode. The iPhone 8 (running iOS 12) did far better than I expected and I can’t see a massive improvement coming with the updated X family (XS, XS Max and XR), although the sensor size is bigger.
My conclusion is that the current crop of flagships run each other very close with regards to overall quality, so it’s more about prioritising what’s important to you than naming one phone an outright winner.