best resize technique down to thumbnail

[gondwana]

Hi

I am new to this forum so hello everyone. I have a good digital photographic experience as well as a few years of photoshop behind me.

However... I would like to know from all the experts on this forum what they consider as the best technique to resize a clean original digital or scanned photo (2-3K x 3-4K px) down to a web size and a thumbnail size.

In a few words, here are the steps I follow.

0/ Adobe RGB fine jpg or raw file from the camera or psd file from scan = original0
1/ run an interactive action to resize, rotate, adjust level & curves & saturation, very little unmask sharpening and save a psd (temp work file to backup on dvd) + a jpg (10/12) (= original1 on hd)
2/ from the temp psd file, run a fully automatic action that does
- resize to 600 x 400 px
- increase saturation 6%
- add sharpening 60/1/3
- convert to sRGB
- save for web 45 (= image admin/client version)
- merge copyright layer on image
- save for web 45 (= image copyright version)
- remove sharpening & saturation
- resize to 150 x 100 px
- increase saturation 6%
- add sharpening 60/1/3
- save for web 60 (= thumbnail)
So I end-up with 3 images all for web.

Here my questions...

1/ What is the best technique to resize? Is it different from original1 to image than from image to thumbnail? One step or more than one step? I always use bicubic sharper (unless resizing up of course).

2/ What are the best settings for unmask sharpening (considering that it is a compromise as it must be fully automatic)? Should I consider different values for the 2 different size? What about smart sharpen?

I would be glad to discuss the matter further with anyone interested...

[Kevin Connery]

Bicubic Sharper should be good for downsampling, though the combination of that plus the extra USM isn't one I'm familar enough with to recommend.

If you used straight Bicubic, the 60/1/3 seems to have a lower amount (I usually run 100-200%) with a much larger radius (for web sized images to about 400x600-ish, I usually work with .3-.6 pixels). A 1 pixel radius will be pretty noticable at the nominal* 72 ppi display resolution (* actually anywhere between ~50 and 150 pixels/inch, depending on the viewer), which may be why the small amount is used.

A 3 threshold is also fairly high for downsampled images, as downsampling tends to remove most noise, and threshold is mostly used to avoid sharpening noise. 1-2 would probably be better, and if the images were clean to begin with, zero might work also.

[gondwana]

Thanks Kevin

In my research for a better sharpening workflow I found that very useful article from Bruce Fraser :

http://graphicssoft.about.com/gi/dynamic/offsite.htm?zi=1/XJ&sdn=graphicssoft&zu=http%3A%2F%2Fwww.creativepro.com%2Fstory%2Ffeat ure%2F20357.html

Applying his receipes have improved a lot some photos. The only thing I am still not happy with is the last step when I need a final size of 150px wide thumbnail.

[pixel8]

Best?

For what. Speed with OK quality for general purpose batch processing or quality alone. If it's quality you're after then each image should be hand sharpened on its own merits.

[gondwana]

Best?

For what. Speed with OK quality for general purpose batch processing or quality alone. If it's quality you're after then each image should be hand sharpened on its own merits.

I have a good original file 3000x4000 px that I resize down to 400x600 px following the technique described above and it is working great in most cases (I have 20000 images to re-processed). However, when it come to go down to the thumbnail size, I think better can be done AUTOMATICALLY then what I use to do. Is it better to resize the small file again, or better to start again from the original down to the thumb in one step? What value of sharpening on a small image of 95x150 px? My photos are mainly landscapes.

[pixel8]

well from my experience in reducing very large files to web size then thumnails I think a better result is had by reducing the smaller version further and resharpening again accordingly. But really I find it difficult to believe that it really matters.

[Kevin Connery]

I use PhotoKit Sharpener myself, and agree with the 2-3 stage sharpening philosophy, but USM can do a very nice job as well.

Nevertheless, at 100x150 pixels, there's not much detail TO sharpen, and I'm not sure there's much than could be done with a photographic image at that size to make it look its best other than on a very case-by-case basis.

[Ammar Midani]

I would show the naivigator palette, zoom out to the size i want my image to be, read the percentage value, and then enter it in the image size for Percentage, use Bicubic sharper.

This also could be used as an action (in steps, like 85% for each step) when downsampling from a very big photo.