Image Segmentation

6

So

So all we have to do is define and implement the

similarity predicate.

But, what do we want to be similar in each region?

Is there any property that will cause the regions to

be meaningful objects?