Saturday, December 17, 2005

Misrepresentation of A/V Forensics by Hollywood

Eventhough I must admit that I watch very little television these days, I do keep my eye out for news reports concerning representations of media forensics (audio, video, computer) on such programs as CSI, The Wire, and, in its day, the X-Files. I also take in the occassional cinema blockbuster.

From these various glimpses into mass-market entertainment, two things typically catch my attention: the misrepresentation of electronic media forensics and (blatant) vendor product placements. My focus in this post is the former. I must say that I particularly enjoy the scenes where the characters take a single (still) image from a surveillance camera and zoom in on a subject's face. Of course, the subject is so far from the camera that his face takes up maybe five or six pixels horizontally and not many more vertically. On some of the most amusing examples, the illumination on the face is poor on top of it. I'll ignore, for the moment, the fact that in real-life the video they are working with is almost certainly heavily compressed.

Then, a technological miracle occurs and through the application of digital filtering the face is a near perfect match to the suspect and the actors race for the door (or flip open their mobile phone to alert their partner, take your pick). It makes for eye catching drama, but that is not the way it happens in real life. Zooming in on a few pixels in a still image just leads to really big pixels - not a clear image - no matter what filters you apply. In my experience, minor improvements in a still image's edge contrast are possible, but not major ones. Major improvements imply recreating information that is not in the recorded image.

I might should add the caveat "using today's technology", but at the time of this writing I can't conceive of how we can ever see major improvements in restoring & enhancing low resolution still images - particularly to the level of what Hollywood (mis)represents. I hope that I'm proved wrong, of course.

(Writing this post makes me think I need to address the use of compression when storing surveillance camera video. I will in the near future).

No comments: