Running With Foxes

TechCrunch recently posted on MotionDSP, the Alger Associates backed San Mateo starup using patented algorithms to enhance low quality video. After looking at Digg and TechCrunch comments, I think some people were missing the complexity of what MotionDSP was doing.

MotionDSP is not just despeckling and cleaning up artifacts through anti-aliasing. MotionDSP is using pixel information from the other frames of the video to enrich the quality of the current frame. This works because as a low definition camera pans a scene, it picks up more visual information about the real world objects in each successive frame. Sean Varah, the CEO, told me that MotionDSP uses pixel information from up to as many as six frames to add information to the current frame. This yeilds a more accurate depiction of the object, as seen in the book shelf video, where you can now read the titles. The desktop version currently enhances one minute of video in ten minutes.

Yes, there are some other competitors like Salient Stills and Cognitech, but these firms focus on generating still images for security applications. MotionDSP is focusing instead on the consumer market, hoping to etch out a niche enhancing notoriously low quality cell phone videos (they are currently providing some examples for a very large online video company).

I see this really having some traction as online video becomes higher quality (YouTube doesn’t currently do the service justice). As to the threat of ever higher resolution phone cameras? Sean says you can’t just throw pixels at the problem because the size constraints will always hinder picture quality by forcing phones to have tiny apertures. MotionDSP can actually deliver even better results when they know the optics of your phone camera and when companies like Shozu make uploading video to the internet easier I see a company like MotionDSP poised to unlock all of that cell phone content.