Transformations in 8 bit versus 16 bit

"reprints" of selected articles from Usenet and other forums
-default
Posts: 1916
Joined: Thu Mar 26, 2015 1:53 am

Postby -default » Wed Apr 08, 2009 5:00 am

Recently, on the alt.graphics.photoshop Usenet group, someone asked whether a 16 bit image would stand up to repeated rotations better than an 8 bit image.  Intuitively, it seems like this might be the case, since there is a lot of math involved in a rotation, and round-off error would seem to be an important factor.  To test this theory, I did a simple experiment involving repeatedly rotating and un-rotating the same image.

Images are as follows:

1) the original 8 bit image (287x278 pixels)
2) the original 8 bit image, rotated and un-rotated 16 times
3) #2, converted to 16 bits during the rotate/un-rotate
4) #1, resampled to 574x556, rotated and un-rotated, and resized back to 287x278
5) #4, converted to 16 bits during the rotate/un-rotate

Conclusion - when repeatedly rotated, 8 bit and 16 bit images degrade at about the same rate.  Bits per pixel is not important.  Image resolution is important.

If you'd like to do the experiment yourself, I've included the action file that I used to do the rotate/un-rotate operations.

Here are the results:

mikemeister_admin
Posts: 4927
Joined: Fri Sep 20, 2013 8:29 pm

Postby mikemeister_admin » Fri Apr 10, 2009 2:37 am

Thanks, Mike, for an interesting experiment! I know for a fact, from my own extensive tests, that 8 vs 16 bits can matter quite visibly for editing real-world photos (a position one D. Margulis doesn't quite agree with, but what does he know?) -- but I have never tested 8 vs 16 bits with transformations.


Return to “Articles from other forums”

Who is online

Users browsing this forum: No registered users and 1 guest