For David Millier

DMillier wrote:

How abut this: there a series of converging parallel line below nyquist. Above nyquist, the lines break up into a series of blotchy, patchy fragmented collections of B&W splinters.

Our perceptual system tries to make sense of this and partially recreates the lines (that aren't really lines) because the converging lines above give enough of a hint for us to guess. It works for this exact pattern to an extent. However, when you try the same trick with a rectangular plate punched with regular holes, you get a series of diagonal stripes that may be lines but definitely aren't holes.

And this is the problem with using aliasing to simulate detail: sometimes by pure chance the result looks passably like the real subject and other times it looks nothing like the subject whatsoever.

It's surprising that someone who is so philosophically dead against the "guessing" that is Bayer reconstruction is relaxed about the "guessing" that is aliasing false detail.I would have thought that a properly anti-aliased Foveon image would be the holy grail: no Bayer interpolation and no aliasing artefacts = a pure, accurate, realistic image.

Does it really matter that a few percent of resolution would be sacrificed? It's not as if the Merrill sensor is short of detail

ps.

See Mike's post in the "magenta blotches" thread for a great example of what aliasing does to damage the integrity of Foveon images.

--
Galleries and website: http://www.whisperingcat.co.uk/
Flickr: http://www.flickr.com/photos/davidmillier/