Running in circles in grepcode trying to find the source for how a rotated image is drawn with subpixel accuracy. I am specifically interested in the method drawImage(Image img, AffineTransform xform, ImageObserver obs)
which handles rotations but lacks implementation detail.
A rotation is rarely a mapping from integer (i,j) to integer (u,v) so at some intermediate level a sub-pixel view of the rotated image is created. This is not exposed to us via any API (that I know of) so I cannot inspect it, we are only given the imageraster. So how does graphics2d handle subpixel drawing when pixel coordinates are integers?