If Mi 11 had not thrilled on the camera side, with Xiaomi Mi 11 Pro and Mi 11 Ultra the company has definitely done all-in. Not to mention the innovative Liquid Lens technology shown with the Mi MIX Fold , but that’s another story. Today we focus on non-folding, also because they have proven to have photographic capabilities of real camera phones, in the words of DxOMark . There are significant differences between the two models in terms of focal lengths, but both enjoy the same brand new primary sensor. The Samsung ISOCELL GN2 presents itself as an excellent sensor, boasting dimensions of 1 / 1.12 ″ and 1.4 µm pixels, hardware that would be wasted without a software up to it.
Xiaomi explains the Night Owl camera of Mi 11 Pro and Mi 11 Ultra
Precisely for this reason, Xiaomi engineers have created an algorithm designed specifically for the camera of Mi 11 Pro and Mi 11 Ultra. It is defined Night Owl , or Owl’s Eye, precisely to indicate the excellent capabilities in the nocturnal phase typical of this bird. This software component was created with the three most critical aspects to manage in mind: noise management , image stability and color accuracy .
Reduce image noise
When you think of a camera phone, nowadays we have devices that are easily able to return excellent images even at night. But there are contexts in which it is almost impossible to come up with a decent photo, especially when the ambient brightness is almost minimal and close to 0.1 Lux . When the light is so low, the first problem to deal with is the visual noise you get when taking a photo. In circumstances like these, traditional software tries to mitigate noise and chromatic aberrations. However, this process is accompanied by the so-called “watercolor effect”, with images having a low level of detail.
To prevent this from happening, the Night Owl algorithm used on Xiaomi Mi 11 Pro and Mi 11 Ultra takes eight consecutive photos with EV0 exposure. Then starts a software processing chain composed of various algorithms that deal with measurement, alignment, reconstruction and color correction. All with the aim of obtaining the best noise reduction effect, increasing the brightness and going as little as possible at the expense of details and colors.
Like all algorithms based on artificial intelligence mechanisms, Night Owl is also able to self-regulate and learn from the situations it faces. It is thus able to manage a noise calibration system when the framed scene is extremely dark, thanks also to the training it was subjected to by Xiaomi engineers.
The other gripe with night photography is image stabilization. As long as you can take advantage of a good dose of ambient light, the shot happens in a fraction of a second. But when the environment is practically dark, it is necessary to rely on the technique of long exposure , for which it is necessary that the electronic shutter of the camera be active longer. However, moving the camera at this stage means having a blurred photo, as the sensor captures all the information it gets while the shutter is open. Precisely for this reason, to have an optimal long exposure you need to keep the smartphone still, perhaps with the help of a tripod.
But not everyone always has a tripod at hand, therefore smartphone companies, as in the case of Xiaomi, use software systems to compensate for its absence. The Night Owl algorithm does just that, leveraging an optimized neural network to provide it with high quality data for noise reduction management. Specifically, an EV0 RAW multi-frame technique is used that compensates for those micro-movements of the hand that would risk returning a moved, blurred and less detail image.
Get the most faithful colors possible
Finally, the other parameter that the Night Owl algorithm monitors is that of chromatic fidelity . When dealing with a very dark scene, there are two risks: having an overexposed image with high levels of noise or an underexposed image with an out of phase white balance, usually tending to yellowish.
The algorithm that operates on Xiaomi Mi 11 Pro and Mi 11 Ultra used an AI system to measure and correct the acquired color, based on the information obtained from the scene and the sensor. In this way, the photo will turn out to have the right degree of illumination, without penalizing too much the color temperature obtained.