[bojan]

If data is lost (due to smear or bad optics), there is no way it can be retrieved in full (garbage in, garbage out - this is what data processing is in essence). It is like trying to recover data loss by heavy jpg compressing... It is perfectly well known how data is lost (jpg compression is a known procedure), but it is impossible to get it back, once it is done.. because of ambiguity of the final image (two completely different details in the original image can and will produce the same final result in the compressed image). The software can only guess about original data... and by trial and error, it can reconstruct the image with certain probability of accuracy. For full reconstruction, in ideal case, infinite processing time is required.

I remember a movie with Kevin Costner ("No way out") where he plays a American CIA officer but is in fact a Russian spy.
His blurred image was enhanced for days by a supercomputer, until finally it was finished and his face was shown.. Something like this can happen only in Holywood movies, not in real life, I am afraid.

With imperfect optics, it is a bit easier, i think, by performing some additional data collection.. to reduce uncertainties.
With couple images with known amount and direction of de-focussing, and knowing the exact figure of the mirror, I imagine more data can be retrieved.. but not all of it.
And this method should work better on images with known point-like sources (because we know what the final result of point-like source should look like)

Anyway, we shall see :-)