Instead of resizing to 800 pixels, try using the next-smaller number which is an even factor of your original number of pixels. For example, my camera takes an image 2816 pixels wide, so I usually resize to 704 pixels, exactly 1/4 of the original.
I have no proof, but I intuitively feel that you can get a more "accurate" result — especially if there's a lot of fine detail — if the algorithm isn't dealing with "fractional pixels" ... |
