Please read the following articles.
http://smartbear.com/support/viewarticle/11145/http://smartbear.com/support/viewarticle/11136/Essentially, any change in how the target test system renders images on screen could result in image differences. Pixels may line up a little differently, they may be blurred, use more colors, etc. These two articles give a pretty good run down of what kinds of differences you may run into and what you can do to mitigate them.
To be truthful, unless there is no other way of doing so, I avoid region comparisons in automated tests. The reason being is that you spend a lot of time building code and mitigating differences when a screen comparison can be done in just a few minutes with a manual test. This comes under the heading of what I call "What I CAN automate is not always what I SHOULD automate."
