Attached Link: http://www.controlledvocabulary.com/imagedatabases/downsampling.html

You know...everybody talks about resizing images to make them larger for printing (upsampling), but not many talk about resizing an image for web viewing - usually smaller size or downsampling. While reading up on star interpolation, I came upon a really great article on a technique to downsample from film scanner resolutions (either 4000dpi or 2700 dpi scanners) for viewing on the web. I've only tried this technique once, but it appears to do a phenomenal job at keeping fine detail in while it's reducing the size. The really cool thing is that it only uses standard photoshop tools and filters (mainly USM and Image Size)to do it. I am very much impressed with this. The URL above is the link to the article. -=>Donald

Typo - Star interpolation should be Stair Interpolation

...hmmm. Looks interesting. I'll have to try this. Changing the resolution by 1/root(2), 500 to 350 to 247 to 172 to 122 to 72 will half the file size each iteration... Wonder if that is significant? Or if you can use any ratio that gets you to your target size in about five steps?

Attached Link: http://www.graphic-design.com/Photoshop/Tips/

Great info! I will be trying that one tomorrow. Thanks D! The above link is a site with PS tips that I get in E-mail once a month. I have no idea when I signed up for it but it is there religously. Looking for a job you sign up for everything. Here is the URL if you want to sign up, it has some great Design tips too. www.graphic-design.com The link to subscribe is at the bottom. Magick Michael

while im sure it gets mentioned on the page, simply reducing the ppi resolution (ie, 72 pixels per inch), doesn't change anything. It's changing the actual image size that matters. If you have an image that is say 2000x1250 pixels in size, at a ppi of 300 pixels per inch, changing the latter to 72 ppi doesn't actually alter the file itself, just the way it is displayed. It will still be 2000x1250, and thus, have the exact same amount of detail as it did before.

Fict - that's only true if you turn re-sampling off. With resampling turned on, changing the DPI of the image also resamples the image to a new resolution. That's why his first step is to make change his image to 500dpi with resampling turned off. Then he has a starting point without any change to the image. All successive changes in DPI are with resampling turned on. Misha - I'm still looking for any discussion of the math behind it, why they are choosing what they do. For example, many of the Stair Interpolation techniques recommend you change image size in steps of 110%. Why 110? Dunno. -=>Donald

basically, its probably something they like to use. Perhaps they used some heavy mathematical formuals to figure it out, but i doubt it. It's just one of many methods. Best idea is to find your own. Using those as starting points, yes, but otherwise find what works best for your eyes. Does Genuine Fractals do a good job at downsampling i wonder, or will it just get confused and ask, "Why? Why dear god why???"

Attached Link: http://www.humansoftware.com/pages1200/XFile/HSxfil0.html

Yea...GF will do up or downsampling. It's pretty good at it. There have been some testing that shows that bicubic stair interpolation may be better than GF though (don't know off hand if that applies to the new GF 2.5 or not). I made an action to do the downsampling that started this thread. Works pretty darn well too. Another resampling app/plugin I've been looking at is XFile (see link above). Can't find out a lot about it though. Made by a company called Humansoft (same guys that make Squizz and Photoweave). Apparently they don't believe in demos or something. *shrug* -=>D

I have always tried (back from my old MacPaint days) to downsample by an integer amount; along a linear dimension, one pixel in the final image was 2,3,4,... etc., pixels from the original. Back with ancient MacPaint this prevented pixels from migrating at the edges and really messing things up. This integer reduction does not seem necessary with the bicubic interpolation turned on. These articles are specifically NOT using integer reductions; (110% ???), wonder if that is significant, or as fuctious says, just convenient? [The Nyquist stuff gets pretty strange when you are re-sampling "boxcar" approximations. Maybe the original image started out band-limited, but as soon as it turns into discrete pixels it is no longer band-limited. The bicubic interpolation should be working like a low pass filter. I've seen photoshop plugins out there that use higher order interpolation, (sharper filter?).]

Star interpolation... hmmm... hehehe ;-) (*

Okay...I'm not even going to touch this one. :) hmmm indeed.. hahaha -=>D

With Star Interpolation.

With single pass (biqubic) and unsharp mask...hmm... see some difference in peacock feathers... guess its worth programming an Action script.