PDA

View Full Version : down-sampling for BPN



John Chardine
10-26-2008, 12:03 PM
I found this write-up of down-sampling at:

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

Seems to suggest a progressive down-sampling with USMs at each stage. I have tried it and it produces a nice result but not sure if any better than the old way I used which was (in Ps): 1. Image-Image Size-96dpi, Resample image unchecked, 2. File-Automate-Fit image 800dpi x 800 dpi.

I'd be interested in what people think about the method described in the URL above.

Alfred Forns
10-26-2008, 02:39 PM
John there are no secrets and don't think that will give you better results Have seen that method many times.

The dpi will make little difference since its all about the total number of pixels. I would suggest to use bicubic sharper when downsizing and when saving do use save for web ..... and select the max size of 200kb

David Thomasson
11-15-2008, 01:15 PM
I think there's a fundamental misconception in the method you linked to. If the goal is to get a smaller, faithful representation of a large image, dpi and ppi are irrelevant. If an image is 800 pixels x 800 pixels, it will look the same on screen no matter what the resolution is. One of these color wheels was saved at 72 ppi. The other was saved at 10,000 ppi.

http://img505.imageshack.us/img505/9051/5colorwheel2aph6.jpg

Suppose your original image is 3300 px on the long side and you want a screen sized version that is, say, 800 px on the long side. In Photoshop, leave "resample" checked and simply change the dimension on the long side to 800.

If you use the bicubic sharper algothrithm, there's no need to downsize incrementally and apply USM at each stage. Bicubic sharper does the sharpening, and you'll get the best result by downsizing in one step.

http://img84.imageshack.us/img84/6339/bicubicul3.jpg

John Chardine
11-15-2008, 02:49 PM
Thanks David. I know of several methods of down-sampling for the web including the one you illustrate, and it is unclear whether any of them work better than any other. I guess some tests are in order. I used to use Bicubic Sharper in Ps but it is rather a blackbox into which you throw the image with the assumption that what Adobe is telling you ("best for reduction") is in fact correct. Maybe it is. Another method I used for a time, which I think is advocated somewhere on BPN, is to run File-Automate-Fit image to 800 x 800. The essence of the method described in the link I gave was to down-sample the image in stages with small amounts of sharpening in between, rather than do the whole process in one giant step. I find the logic of this approach compelling but ultimately as you and Alf point out, it may get you no where.

David Thomasson
11-15-2008, 05:33 PM
When downsizing and saving for the web, I think the calibrated eyeball is the final arbiter. A difference that can't be seen is really no difference at all.