Re: [APML] Scaling Images (was: Cocoon Nebula)

[Mikkel Steine <mikkel@messier45.com>]

Matt,

> The pixels patterns resulting from GF are slightly different than Photoshop
> bicubic. The GF stars are sharper and less smeared. The JPG size of the GF
> image is larger, which may be an indication that GF preserves more information
> from the original image.

As Thor Olson explained about the JPEG comression, that would be true, yes.
But... read on...

> It is very curious how the image annoyingly shifts around when you toggle
> between the two. Could PS or GF be screwing up its scaling?

It seems that both bicubic downscaling and GF downscaling both eliminate certain
rows and columns in the image. This is due to the different way in which these
techniques work. I'm not very much into these techniques on the math side, but I
have several books that explain in detail problems with downscaling. In fact,
due to the difference in the techniques and as Jerry pointed out, that GF is
primarily for scaling up images, they will produce different results that look
horrendous when comparing the results. From your results I would not recommend
GF for downscaling at all. I guess bicubic scaling is a good all-round scaling
method. The apparent difference in detail preserved is probably only due to a
slight difference in edges in the images. Signal processing is NOT easy! :)

On the other hand, what I initially actually was looking for was a way to reduce
the brightness of the faintest stars before downscaling so that the image would
look more natural afterwards. Some sort of low-pass filtering perhaps to make
the faintest stars fainter and the brightest stars almost the same?

Any ideas?

Best regards,                              mailto:mikkel@messier45.com
Mikkel Steine                                     http://Messier45.com
----------------------------------------------------------------------
Search 500,000 Deep Sky objects:  http://Messier45.com/cgi/dsdb/dsb.pl
----------------------------------------------------------------------


--  APML Archives at <http://astro.umsystem.edu/apml/>  ---
             Unsubscribe at <majordomo@seds.org>