Tuesday, September 17, 2013

Understanding morphology

Most machine vision software comes with a host of image processing tools. If you look through the menus you’ll see all kinds of filter and morphological operators.

Now usually I tell people if they need to use these tools their image isn’t very good. I’d rather see time and money spent on creating a better image than on trying to clean up a crappy one. But that said, sometimes it does help to do some image processing.

Morphology isn’t generally understood by those who actually use machine vision, (as opposed to grad students who do image processing every day.) If that’s you, take a look at Ben Dawson’s article “Image Morphology” published in the Vision & Sensors supplement to Quality Magazine, July 8th, 2013.

Ben does a very thorough job of explaining the principles of morphology and why you might want to use it. He also provides some references that I think are well worth following up, (if you have time to read textbooks, that is.)

What Ben’s a little light on though are the two weaknesses. Morphology can be computationally intensive, meaning it’s slow, and it changes the shape of the objects in the image. (The clue should be in the “morpho” term.) But if you can live with that it can be a very helpful tool.

Once again, click, read and learn.

No comments: