In praise of soft focus images

I’m told that actresses hate modern High Definition TV: it shows all those lines that they’d rather hide. Makeup is one solution, but even that tends to be visible so an alternative is to soften the focus just enough to blur out those lines.

The same trick can help in machine vision. If you have a surface with minor local variations – perhaps the grain structure on a sheet of stainless steel – a slight defocusing of the lens will even these out, giving a smoother, more homogeneous appearance.

It’s possible to achieve the same thing by applying various image processing filters, Gaussian for example, but why spend time on processing in software when a lens tweak works at the speed of light?

Of course, the risk with your ever-so-slightly defocused image is that a well-intentioned colleague will decide to sharpen things up, and suddenly the false reject rate goes through the roof. To prevent this happening you might like to lock the lens in position with a bead of silicon sealant and make a note in the documentation to record how the image is supposed to appear.