Legacy ImageMagick Discussions Archive

@Anthony, especially:
Do you have some line drawing images that you consider good to test line drawing downsizing?

I found one:
A first attempt. (I've not tried anything else. Just a first "go".) This gives:

Are there tests of this kind of operation somewhere? So I can tell what's usually done and how well it fares?

Well: First attempt -> Thumbs down.
Here is the same filter through linear light as before, but without sigmoidization:

With less sigmoidal contrast, namely 6 (down from 25, which is rather large, in my first attempt):

Not particularly impressive either.
-----
OK, I'll do something else for a while.

One last try: Ginseng through linear light, no sigmoidization:

I do not know if any of these will help.

The rings image, which you probably have.

These I found by searching Google

http://v7.tsemtulku.com/images/stories/ ... RAWING.jpg

http://fc04.deviantart.net/fs43/i/2009/ ... queror.jpg

http://www.peppertop.com/polygon_web.svg

http://www.online-utility.org/image/Ima ... rawing.svg

http://www.online-utility.org/image/Ima ... rawing.svg

http://golancourses.net/2010spring/wp-c ... s-1963.jpg

http://richworks.in/wp-content/uploads/ ... actals.jpg

http://www.wussu.com/fractals/images/wd950403.gif

http://www.wallpapergate.com/data/media ... al_028.jpg

http://beatrizortiztrapote.files.wordpr ... tail13.jpg


or at http://sourceforge.net/projects/testimages/files/ (testimages_hdtv.zip or testimages.zip)

@Fred:
These are nice.
Thank you.

It takes a long time to download, but you might find that the color images in the testimages.zip set at http://sourceforge.net/projects/testimages/files/ to be useful for color images to downsample

I prefer linear light with EWA lanczos3Sharpest

henrywho wrote:I prefer linear light with EWA lanczos3Sharpest

Thank you Henry for your opinion. Here is the result:

Not that different, on this test image, than linear light (orthogonal) Ginseng, and also to linear light (orthogonal) Lanczos and Cosine (with the new 3-lobe Sinc default). It's different. Just not very different.
I've actually removed the suggestion that one may want to use sigmoidization when downsampling from the current draft of my recommendations: http://web.cs.laurentian.ca/nrobidoux/m ... colas.html
(And EWA Lanczos3Sharpest, which fares well in the tests of my grad student Adam Turcotte, was there from the beginning in the downsampling section, as an example of lowering the blur to get additional sharpness.)

EWA Lanczos3Sharpest offers more pronounced darkness differences in the gray "areas".
EWA Langrange, at this round, hurts the picture because its strong dark-halos are darkening the lines too much.