For the Raspberry Pi, which is Linux-based, anything computer vision related you'd want to go to the OpenCV documentation and forums. It's not "obstacle avoidance software" but rather a set of software routines to simplify interfacing with a camera and performing the image manipulation algorithms common for computer vision.
If you're looking for a simple technique I'd suggest a line-generating laser mounted above your camera and pointed at the floor about a meter ahead of the robot. Any obstacles will show up with a red line on them, and it's relatively straightforward picking out that red line from the image.
Roborealm is good software for this sort of thing, but it's Windows-based. Still worth looking at for ideas.
