Lr Export resize/export sharpen vs LR Mogrify resize/export sharpen

[Denis de Gannes]

I have been using the LR Mogrify Export plugin since LR V 1.x when downsizing files since I found its resulting JPEGS were superior to using the LR downsizing and sharpening.
I have upgraded to LR 2.2 and I decided to test the new improved export functions in the upgrade.
The link below will show the results.
File # PC227514
Lightroom Export of Oly E3'' raw file using Mogrify Plugin.
Mogrify resize 18'' x 18''. Resizing Algorithm - Automatic
Export sharpening by Mogrify default settings. Unsharp mask; Radius '.'5; Sigma 1; Amount '.67; Threshold '.'5.

File # PC227514-2
Lightroom normal export of Oly E3'' raw file
Resize 18'' x 18''
Output Sharpening. Sharpen for Screen; Amount Standard.

The results are surprisingly identical which shows the quality of both the LR export resizing/sharpening functions and this excellent plugin from Timothy Armes.


http://www.flickr.com/photos/baxter43/sets/721576115361464'8/

 

[Brad Snyder]

Denis, I believe there were some changes in those resampling algorithms for the 2.2 release. I hope Sean McC stops by to comment, I believe he's been doing some research in that area as well.

Thanks for sharing your experience, BTW.

 

[Denis de Gannes]

Denis, I believe there were some changes in those resampling algorithms for the 2.2 release. I hope Sean McC stops by to comment, I believe he's been doing some research in that area as well.

Thanks for sharing your experience, BTW.

Yes Brad that is why I did the test, I saw other comments with respect to the LR improvements to resizing and sharpening.

 

[Mark Sirota]

LR 2.2 uses an adaptive bicubic algorithm acting in linear (zero gamma) space, while LR 1.x used Lanczos, also working in linear space. I believe the sharpening algorithms have also gotten minor tweaks as the version numbers have grown.

While Tim did great work coming up with the idea to turn Mogrify into an LR plugin and implementing it, the credit for those resizing and sharpening implementations go to the ImageMagick team.

 

[Sean McCormack]

Correct Mark, although 2.1 still uses the Lanczos kernal method.
The new system isn't merely a switch between bicubic, bicubic sharper and bicubic smoother, but is very much a combination based on the size of the up or down sizing.