It’s very clever, but is it machine vision?

If you ever find time to watch the various cop and forensic investigation shows on TV you’ll have seen how the poor underling gets sent off to review hours of surveillance video. Invariably, he or she returns some hours later with the key piece of video evidence needed to facilitate the “takedown.”

Well this might work on TV but in the real world it takes many, many man hours to go through such tapes. And as we vision professionals know, since visual inspection is mind-numbingly tedious work, there’s a very high risk of missing the crucial few seconds.
This is where “Video Synopsis” from BriefCam could score big. Computer vision people at Hebrew University in Israel have come up with a way to compress surveillance video without leaving out any of the action. As best I can describe it, the approach is to segment background from foreground, then overlay the foreground events on what is effectively a static background image.

But rather than trying to make sense of my mangled and somewhat inaccurate explanation, I suggest you mosey on over to http://briefcam.com/ and see for yourself.

And when you’ve taken a look at the videos presented, come back and answer this question for me: I’m pretty sure Video Synopsis qualifies as computer vision, but is it machine vision?