This appendix covers a few of the more advanced segmentation and feature extraction tools in the SimpleCV framework. Some of these automate basic processes covered earlier in the book, such as looking at the differences between sets of images. Other elements represent more advanced tools for identifying and extracting features.
As we mentioned in Chapter 8, the purpose of feature detection is to identify the parts of an image that are “interesting.” It might seem counterintuitive, but one way to determine what is interesting is to first find what is uninteresting. Remove what is uninteresting in an image and the interesting portion must be part of what is left. This is a useful technique because sometimes it’s easier to identify the uninteresting parts of an image—namely the background. For instance, given video feed, the moving parts of the image are often the interesting parts and the stationary parts are less interesting.
In computer vision, this type of process is called foreground/background segmentation, and there are a variety of ways to accomplish it. The SimpleCV framework currently supports three different techniques:
frame differencing segmentation
running segmentation, and
All of these methods examine multiple images to find pixels that are changing (the foreground) versus the pixels that remain static (the background). Also remember that the terms foreground and background do not refer to how ...