A video demonstration of the Xiaomi Mi 11 night mode has been published by BlinkAI. What to see all the strength of the Snapdragon 888.
Xiaomi presented in China the Mi 11, its new flagship for the year 2021. It is a very high-end smartphone, offering the best of what is done in terms of hardware and many features additional software. Among them, we can mention the heart rate monitor integrated into the fingerprint sensor, the sharing of sound to two Bluetooth devices simultaneously or the night mode in video.
BlinkAI presents the video night mode of the Xiaomi Mi 11
This video night mode was designed in partnership with BlinkAI, an American start-up specializing in artificial intelligence. Xiaomi and BlinkAI have thus taken advantage of the new features of the Snapdragon 888 to apply the same treatment to each image of the videos that can already be experienced on photos.
In a few milliseconds, the algorithms of machine learning will push the sensitivity of each frame and improve its rendering by removing noise and adjusting colors. BlinkAI posted two demos on its YouTube channel:
The result is particularly impressive in the second video showing a statue. We see that the scene clearly gains in brightness while retaining colors that seem natural and without adding particularly noise to the image.
On the park maps, however, there are some limitations, mainly related to the small size of the sensor and the lenses on a smartphone, which make the result rather blurry from a distance (on tree branches in particular), as well as some concerns of algorithm that generate artefacts and other chromatic aberrations linked to demosaicing of the image.
We can see a purple halo in certain areas of the image, but also a contrast which seems to have been pushed to the level of the movements of the water, creating an impression of “glitch” in certain places.
A promising first generation
Remember that this is the first manufacturer to unsheathe its 2021 vintage smartphone and that many others will follow in the coming months. Without a doubt, each will then offer their own algorithmic version of video night mode, with varying degrees of success. These algorithms will then be refined from generation to generation in order to achieve an ever better result.
Remember the first photo night modes offered by Huawei and Samsung, which tended to over-saturate the photos and push the brightness to excess. If the result is still not perfect, the road traveled in a few years leaves a lot of hope for the future of these algorithms for video.