The full story is in this news piece on their web site, complete with video.
Here's what caught my eye ...
The new process begins by photographing buildings with a system the team developed to get high-resolution, long wave infrared images using an inexpensive, low-resolution camera. Normally, the cost of high-resolution far-infrared cameras is prohibitive for such widespread use — such cameras can cost $40,000 each. As a substitute, the team developed a novel patent-pending technology called “Kinetic Super Resolution” that uses a computer to combine many different images taken with an inexpensive low-resolution IR camera (costing less than $1,000), that produces a high-resolution mosaic image.Note that in this context 'high resolution' would be 640 by 480 pixels resolution for $40K. In this case the MIT system's Kinetic Super Resolution improves image resolution by using moving (presumably this means scanning) images. I would imagine that this is trading temporal resolution for spatial resolution by looking at how successive pixels differ when these pixels are displaced by less than pixel/image resolution. A patent is pending.