Maximum error correction and each QR Code "pixel" is 6x6, while each video pixel is 1x1. And, there is enough intentional "pixelation" to ensure enough codewords read correctly. (Repeated for each frame.)
Is the pixelation of the video a requirement for a successful scan? Would it be possible to embed a higher-quality video within a QR code, or does it have to be pixelated like it is in that one?
http://en.wikipedia.org/wiki/File:QR_Ver3_Codeword_Ordering....