Decouple image and mask in model.

[hajicj]

This is a design issue. Currently, the image is also its own mask. However, we might want to have the ability to load an image and keep its mask separate (also for switching certain parts of the mask on and off, or highlighting unmarked pixels, etc) - the mask could be dynamic, the image is static.

All the tools like TrimLasso and ConnectedComponents should work with the mask.

What needs to be figured out:

• How to represent the mask (Part of the model? Or: isn't the image the new thing, and we've been working with the mask all this time, just calling it "image"?)
• How to display the image and mask? (Where does it go in the .kv file? Transparency...?)
• How does loading work? How do we package the two images, so that loading is comfortable?
• How to synchronize moving & scaling both?
• How to deal with mask (binarization) errors? Do we allow users to edit the mask?

[hajicj]

Representation

I'd say that the most logical path is calling what we now have the "mask" and saying that the "image" part (non-binary) is new.

Also, this terminology is confusing. Let's call the binary thing still the "image". The non-binary version could be called... dunno, photo? Or original?

[hajicj]

Note that this issue does not really address #77 -- that is a question of making some dynamic overlay of the annotated image.