Artificial Intelligence
How better smartphone video is getting, well, smarter
Innovative video enhancement technology combined with artificial intelligence is making smartphones the video platform of choice, says Andreas Lifvendahl, CEO, Imint
T
oday, smartphone cameras are becoming even smarter. Current leading smartphones include artificial intelligence (AI) capabilities that can take what their cameras see and turn it into better, more engaging, more useful video. Imagine, when you’re video chatting on your smartphone or recording yourself talking and walking through a museum, that your face is always perfectly centered. No matter how your hand moves, a combination of artificial intelligence and facial recognition programmed into the smartphone ensures you stay in-frame.
That technology already exists, part of a video optimisation software solution my company makes called Vidhance. By integrating AI into smartphone software, consumers can create more compelling video content and enjoy richer communication with friends, family, and co-workers. And so far, we’ve seen just the tip of the iceberg.
Three phases of AI video in smartphones
To date, companies like mine have made great strides in developing sophisticated software algorithms for enabling enhanced video performance in smartphones. We’ve created proprietary technologies to reduce motion blur, stabilise video, correct the field of view, and more.
We can now upgrade these core technologies through AI so that the smartphone can predict how best to improve video and make it searchable. In the context of smartphone video, AI is a powerful technology that enables improved real- time video enhancement processing on the device—so it’s immediate—and in a way that anticipates the user’s desire or intent. As developers of software to improve smartphone video, we see AI being applied in three general phases:
Phase 1: Basic machine learning: The technology to keep users aligned in a video frame—what we call Vidhance Selfie Mode—is available now. It and other solutions exploit AI technology from mobile hardware developers such as Qualcomm to interpret what smartphone cameras see and apply algorithms that optimise the experience.
Another existing video enhancement solution that can be automated through AI is horizon correction. For anyone capturing video on a smartphone—especially when they’re walking or panning—a skewed view of the horizon, one that tilts or wobbles, negatively affects the video output. It’s easy to apply algorithms to compensate and adjust, but with AI, the smartphone camera auto-corrects, making the enhancement technology much more useful.
Phase 2: Object recognition and analytics:
New high-speed, low latency 5G wireless communication is about to bring big data to smartphone video enhancement. Although not yet on the market, the technology now exists to
42 May 2021 Components in Electronics
www.cieonline.co.uk
use AI and video data stored in the cloud to recognise what the smartphone camera sees—the family cat, for example—and to apply metadata that describes the content. Say you take 10 minutes of video and the cat enters at the three-minute mark. With AI, object recognition, and automated indexing, the smartphone knows the cat was there—and for how long—and could lead you to the exact spot in the exact video file should you ever want to create a compilation of cat videos.
Perhaps even more exciting is the coming field of multi-camera, multi-smartphone video. When several AI-enabled smartphones capture the same event—a soccer game, for example—phase 2 capabilities will allow users to automatically edit together disparate video streams into a single video. Through object recognition and analytics, each phone will know what it’s looking at (and when) and be able to communicate that information to an automated or manual editing solution, which will compare data from other video feeds to create the final cut.
Phase 3: A subjective, AI-based feedback loop:
The first two phases of AI-based video enhancement are fairly objective, supported by metrics, distances, data, etc. But how does the AI engine know a video is good or interesting? If, for example, a video platform
is taking phase 2 object information to create a new video, how does it know one clip—maybe a goal being scored—is more compelling than another?
In phase 3, not only will smartphone video benefit from object information delivered over 5G, but it may also learn from cloud- based qualitative information. Which shot is the best? How can the smartphone decide what algorithms to apply when? Well, which kind of video earns the most “likes”? Which draws the most clicks?
We know how video enhancement technology can be applied to improve content—centering images, leveling the horizon, reducing blur—but in phase 3, a crowdsourced, big data feedback loop may be able to weigh in on whether content is actually compelling, engaging, funny, or whatever quality the user is aiming for.
Democratising high-quality video Although AI can help automate the creation of better smartphone video, the technology itself is an enabler. The goal shouldn’t be to create a smartphone that users can point at anything and expect a polished, professional-looking video. Instead, advanced video enhancement software, combined with AI, furthers the democratisation of video creation.
weareimint.com
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54
