Generally I think of a line on the screen as represented by a sum of
underlying basis functions which are completely determined by the value
at the known sample points (say the result of doing a a big DCT).
But this interpretation means you also can make assumptions about the
varying values between those points, with exact upscaling, etc. If the
original image was Nyquist limited then those assumptions will even be
correct, else not.
- Tom
Tom Barry wrote:
Craig -
I'm still not sure why you single out non-nyquist filtered images as being a special problem for non-square pixels.
To me, the main problem with non-nyquist images is the values at the designated sample points are correct but there is an ambiguity about what the values are between those points.
There is no ambiguity. The sample points are essentially squares, (or rectangles), each of which is filled with the
designated color. Non-Nyquist images IN EFFECT assume infinite resolution, with each "pixel" a fully, perfectly,
resolved square.
If you think of it this way, you immediately see the problem with rescaling. It only works correctly with up-scaling to 2, 3, 4, 5, .... times as many pixels in each direction.
Doug McDonald
---------------------------------------------------------------------- You can UNSUBSCRIBE from the OpenDTV list in two ways:
- Using the UNSUBSCRIBE command in your user configuration settings at FreeLists.org
- By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word unsubscribe in the subject line.
-- Tom Barry trbarry@xxxxxxxxxxx Now seeking new full time position as video software developer Find my resume and video filters at www.trbarry.com
---------------------------------------------------------------------- You can UNSUBSCRIBE from the OpenDTV list in two ways:
- By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word unsubscribe in the subject line.
