Every object in your repository will be stored once and referenced by it's SHA1 sum. This means that placing three copies of a 100KB image in two directories of your repo will use 100KB, plus some inconsequential amount of overhead.
The same applies to pushing, pulling and branching: So long as the SHA1 sum of the image doesn't change, git will never store a second copy or have to move more than one copy across the network.
You'll be using twice the disk space on each machine though: Git maintains a copy of all it's data in a hidden .git directory in the root of your repo.
