Credits to UnConeD (Steve Wittens) for the explanation of Bilinear Filtering (contact info is at http://www.acko.net/)

The AVS display is an image, and so it's composed of a grid of pixels.  So when you wish to do a rotation or distortion, you can't do it perfectly, because the new grid doesn't match the original one perfectly. To get around this, AVS supports Bilinear Filtering. 
I'll explain through an example: suppose a movement specifies that a certain pixel needs to have the value at coordinates (10.4 , 15.8).  Without Bilinear Filtering, we would have to settle for (10, 16), the closest alternative.  But this would cancel out subtle movements and introduce obvious artifacts.  With Bilinear Filtering, AVS will take the 4 surrounding pixels (10, 15), (11, 15), (10, 16) and (11, 16), and blend them together using weights calculated off the fractional part of the coordinates (0.4, 0.8).  Of course this is a little bit slower, but you get a much cleaner picture.  Note that, because of the blending and rounding errors, Bilinear Filtering tends to fade out your image a tiny little bit.