Say we have an image in which a human eye can distinctly see several "planes" of focus - a blurry background ("telelens"), the sharp(er) subject (a human, or an object), and again a blurry foreground (if any).
I would think that it is computationally possible and feasible to "separate" the subject from the rest of the image data. I.e. selecting the subject borders, by means of outputting a mask channel that can be imported into say Gimp (I don't use Windows, so unfortunately don't have Photoshop) as a selection.
I would review anything from a finished tool through a library (I am on Linux) down to links to articles and/or review papers that discuss means of realizing the algorithm I have outlined.
