PDA

View Full Version : linear sharpen vs. resizing


sudaplatov
5th of March 2002 (Tue), 00:18
What is better in general: resize picture to smaller size
and then apply linear sharpen or first apply linear
sharpen and then resize?

Todd
5th of March 2002 (Tue), 00:27
Always, always sharpen last.

Todd Asher

Jonathan500
5th of March 2002 (Tue), 07:40
Hi,

How do you resize when using Linear Sharpen 3.11, there is a convert only, but then how do you use the excellent sharpneing actions without it trying to convert again?.

PS I would be upsizing using Fred's SI action

John Boyes
5th of March 2002 (Tue), 12:36
You could just resize the 16-bit Linear TIFF with SI first and THEN run LinearSharpen?

sudaplatov
5th of March 2002 (Tue), 13:00
Actually I meant not upsize but downsize - make image
smaller than its original size to put on web or so.
I noticed that some of LS users do one way and some
do the opposite way and I heard that resizing goes
first and sharpening goes last always.

twwilliams
5th of March 2002 (Tue), 13:11
Just as there's the Stair Interpolation process (upsize by 110%-120%, then sharpen, then upsize another 110%-120%, then sharpen, etc.) you can do the same thing when shrinking if you really, really want to maintain a lot of detail.

If you're going from, say, 2000 pixels wide down to 500 pixels wide, you might want to do it in a couple of steps: first resize to 1000 pixels wide and sharpen, then resize to 500 pixels and sharpen.

You can do it in even finer steps, but I haven't done any testing to see which is better.

I have seen an example somewhere of a photo of a brick house where resizing in smaller increments with USM at each step made a difference. But that may have been specific to that image.

Mike K
5th of March 2002 (Tue), 14:39
When I upsample I always sharpen last. however for downsampling for web sharing photos I simply do not worry about it as much. I find that if the image has lots of detail, that when I go to JPEG to compress my downsampled image (save for web command in PS6 or Elements) that the JPEG artifacts are much more severe than any differences in sharpening workflow. Thus for conveniece sake I usually just take my sharpened image (made for printing) then downsample and JPEG compress that. it usually looks fine for web display unless there is a lot of detail in the image. Since I am sending with a 56K line all of my web images are less than 100-150kb which limits the amount of detail anyway.
Mike K

Roger_Cavanagh
5th of March 2002 (Tue), 18:30
sudaplatov wrote:
Actually I meant not upsize but downsize - make image
smaller than its original size to put on web or so.
I noticed that some of LS users do one way and some
do the opposite way and I heard that resizing goes
first and sharpening goes last always.

I think it is best to sharpen then downsize. Certainly Fred Miranda does it this way as well as me. :) However, I do find that an additional light USM _after_ downsizing does add some "pop" back into the image.

I try to keep the image in 16-bit mode until after this final sharpen, then switch to sRGB and 8-bit before saving the JPG file.

Regards,

T a z
5th of March 2002 (Tue), 19:11
Roger...

Just to clarify, Fred Miranda actually recommends sharpening before AND after downsizing. Which is what you do also (except with USM instead of his action). Hope you don't mind the intrusion for the clarification, but just thought some folks who don't use his actions might like to know.

From his EdgeSharpenPro instructions:

AUTO (WEB) resized ~600x400: This option should be used after full size images are downsampled for web output. It's compatible with images resized to approximately 600x400 pixels in size

PS: Your full size images should be sharpened previously with any of the above "full size" actions. After resizing the image for the web (~600x400), use this action.

-Taz

sudaplatov
5th of March 2002 (Tue), 21:49
Many thanx to all!
I will try your recommendations today.

gmontem
6th of March 2002 (Wed), 01:19
Results will vary, but here's one example:

http://gmontem.virtualave.net/pics/forum/CRW_4836.jpg
For that image, I resized from 180dpi to 72dpi, used Pekka's LinearSharpen 3.11 (convert with Normal Sharpening HQ), and then converted to 8-bit sRGB.

http://www.pbase.com/image/1282153/original
Here the Convert with Normal Sharpening HQ was applied first before the resizing from 180dpi to 72dpi.

sudaplatov
6th of March 2002 (Wed), 08:16
I tried both ways and got similar results.
Resizing first and then linear sharpening
looks a little bit better where objects are big.
Sharpening first and then resizing looks better
for pictures with bunch of small details.
It least this is what I got - I resized images to
50% of their original size.