Reducing Images - best practice?

Just wondering what other folks do to get the best possible results
when *reducing* images - mainly, but not solely, for web use...

For me, I've found that bicubic is definitely *not* the way to go on
many images, due to added 'artefacts'. I use Thumbs Plus and
Irfanview most often for reducing, and I have found that either TP's
simple 'resampling' routine, or IV's Lanczos, give much better results
than the other methods, especially on slanting lines/fine details (I
hate jaggies!!!). But it depends on the image content (I'm not sure
yet exactly what the criteria are, but I'm working on it..).

I don't generally use steps for reducing, as I can't see any advantage
- anyone else?

Once I have a good looking un-artefacted image, I use PS USM with a
small radius - maybe 0.3 to 0.8, about 100% to 200%, and levels at
around 3-7. Again, these vary depending on image content. I just
play until it looks right..

There's a very good page on this on Bart vdW's site:
http://www.xs4all.nl/~bvdwolf/main/f...e/example1.htm


Comments and alternative methods welcomed..

(Crossposted to aus.photo and rec.photo.digital, hope no-one objects -
if you do, just remove the cross-...)

On Mar 16, 6:39 am, wrote:
Just wondering what other folks do to get the best possible results
when *reducing* images - mainly, but not solely, for web use...

For me, I've found that bicubic is definitely *not* the way to go on
many images, due to added 'artefacts'. I use Thumbs Plus and
Irfanview most often for reducing, and I have found that either TP's
simple 'resampling' routine, or IV's Lanczos, give much better results
than the other methods, especially on slanting lines/fine details (I
hate jaggies!!!). But it depends on the image content (I'm not sure
yet exactly what the criteria are, but I'm working on it..).

I don't generally use steps for reducing, as I can't see any advantage
- anyone else?

Once I have a good looking un-artefacted image, I use PS USM with a
small radius - maybe 0.3 to 0.8, about 100% to 200%, and levels at
around 3-7. Again, these vary depending on image content. I just
play until it looks right..

There's a very good page on this on Bart vdW's site:http://www.xs4all.nl/~bvdwolf/main/f...e/example1.htm

Comments and alternative methods welcomed..

(Crossposted to aus.photo and rec.photo.digital, hope no-one objects -
if you do, just remove the cross-...)

This may depend somewhat on the specific photo editor. I use PSP, and
have found the bicubic very good in general. However, PSP has an
option called "smar t size", and I have no idea which algorithm or
algorithms it uses. I get the idea it depends on the amount of the
resize job and selects what it thinks is the best algorithm for that
job. I ordinarily cop out and just go with that, but when for some
reason I HAVE used PSP's bicubic have been quite satisfied

"Don Stauffer in Minnesota" wrote in message
oups.com...
On Mar 16, 6:39 am, wrote:
Just wondering what other folks do to get the best possible
results when *reducing* images - mainly, but not solely, for
web use...
SNIP
There's a very good page on this on Bart vdW's site:
http://www.xs4all.nl/~bvdwolf/main/f...e/example1.htm

This may depend somewhat on the specific photo editor. I use
PSP, and have found the bicubic very good in general. However,
PSP has an option called "smar t size", and I have no idea which
algorithm or algorithms it uses.

If you want to be sure it does a proper job for a given subject, try
the
link on my other webpage article on downsampling:
http://www.xs4all.nl/~bvdwolf/main/foto/down_sample/down_sample.htm
which also allow to compare to several other properly prefiltered
ImageMagick methods.
The 'Rings1.gif' zone-plate target has image detail at any angle and
with virtually all spatial frequencies that can be represented in an
image of that size.

I made that page is specifically because of the increasing need to
down-sample multi-megapixel images (from high resolution film scans or
capable digicams) e.g. for web publishing. It gives guidance,
especially if one wants to batch process many images without the need
to inspect each and every image for artifacts. If the down-sampled
target is well behaved, normal images (even the tricky ones) can't be
worse than that extremely critical target.

--
Bart