Whatever your feelings about Microsoft, their R&D operation employs some of the smartest minds in the business – and has done so for a long time. TechFest is the event that shares some of the best experiments. Some turn into products, others don’t, but the events always have some gems, and can provide plenty of inspiration.
What I’m personally most excited about is the pace of progress in image analysis and processing. You’ve already seen me marvel at projects that merge and stretch images in 2D and 3D, seamlessly, and each time, I wondered about video:
Microsoft Live Labs Photosynth Transforms 2D Photos into 3D Worlds; Hands-on
Savvy Stretching: Free Pixel-Resizing Tools, But What About Real-Time, Video?
Well, video has arrived, as pictured at top:
Real-Time Stitching of Mobile-Generated Videos
The ubiquity of mobile phones in today’s society enables the capture of multiple videos of the same event, each with a small field of view. This Microsoft Research technology unveiled during TechFest 2009 enables those small videos to be stitched together into one higher-resolution video in real time.
The project is the work of Ayman Kaheel, Motaz El-Saban, Mahmoud Refaat, and Ahmed Abdul Hamid of the Mobile Multimedia Content Services Group at the Microsoft Innovation Lab, Cairo. That’s proof that great R&D can and will happen in cities that aren’t called Cambridge (Massachusetts or England).
Microsoft, if you’re listening, I’d love to see this as a developer framework. I’m sure there are uses for end-user projects, but imagine how many other ideas couldn’t be realized by any single developer. Live visualists could do amazing things with this, stitching together bunches of different cameras.
In another visualist-ready achievement, Microsoft is projecting onto a really big dome with a single projector, thanks to a wide-angle lens, as seen below. It’s a simple project, but given the complex, multi-projector setups I’ve seen doing the same thing, it’s a very useful project. (Now here’s a case in which I wish they had done more documentation, as this is probably not something they’re going to patent.)
I have some contacts at Microsoft Research, so I’ll see if I can’t find out more.
