It depends what the original filter did to it.
Even if it is a simple change of contrast and brightness, when you "reverse" the controls on the adjustment, you end up with an image that is more posterized -- do a test for yourself with the Levels. If you manipulate the Levels and save it, then do Levels again and push it back to the original, you notice that there are fewer lines in the histogram that are widely spaced, because you threw away the data on the first filter, and then have to stretch the remaining data out when you do the second.
There is a limited amount that you can do with interpolation, where the software 'guesses' the values to fill in the blanks. But this is putting in artificial data points, based on averages, and not the original data that was in the image pre-filtering. Try another example. Make a copy of an image, and reduce the resolution to about 1/5th the original. Then take the small copy, and increase the resolution again by 5 X to match the original - now look at the side by side. The blurriness you see is the effect of interpolation.
