If Cameras ate Carrots, We wouldn’t have to Worry about Good Lenses…

Before we purchased one of those fancy HD video boxes that allow you to pick and choose the channels you want, my partner and I watched whatever we could find on basic cable.  Typically, that meant settling for shows like CSI or Criminal Minds – not bad shows in and of themselves, but not the kind of stuff I really like. One of the things that always drove me crazy about those shows was the amount of disbelief you were expected to suspend any time they would play around with video footage or still images.

My favourite scenario was the one where the CSI “tech” lifts an image of a license plate taken by a grainy low-res security feed from a car speeding off into the night. The license plate has perhaps 5-10 pixels to its name and yet, by some magic of image processing, he or she is able to transform a seemingly rectangular blur into a pristine set of numbers and letters that wouldn’t look out of place in an IMAX movie! By this time I am usually no longer able to contain my incredulity, and exclaim something along the lines of: “Oh, come on!!! That’s impossible!!! There’s no way they can extract that much….”

(Maybe you recognize yourself in my behavior.)

And it is always at this point that my clearly better half retorts: “Stop doing that! You ruin it for me every time!” Sorry dear…

Here’s the thing. Even if producers of TV shows and movies stretch the limits of what is possible with image processing, it doesn’t mean that we in the machine vision industry should.

In fact, it is quite the opposite. It’s probably common knowledge for many of us, but it bears repeating: Before even thinking about trying to improve image quality with processing, make sure you have the best quality image to start with. In other words, make sure you have enough pixels to resolve the smallest features you need to see and make sure you buy the best lens you can afford. The last thing you want to do is put a mediocre lens between your precious pixels and what you’re looking at. If you do, you can be sure that fine details will get smeared across a bunch of pixels instead of the one or two they are supposed to show up on.

And that’s a wrap.

About Bob

Bob is in charge of Custom Product Development at the Montreal office of Teledyne DALSA where he leads the OEM Applications Group in the development of custom products and amazing imaging advancements. He enjoys exploring new ideas in Machine Vision, giving customers more than they expect and loves the fact that most of what Teledyne DALSA develops would make great content for the Discovery Channel!
Posted on by Bob. This entry was posted in Cameras, Image processing, Machine Vision. Bookmark the permalink.

Comments are closed.