Enlarging images part three
Ivan | Tue, 2004-07-06 13:41A reader pointed out that after enlarging the image, sharpening done directly on the RGB image is not the best one can do.
Here is what he says:
"I would think that performing the sharpening on the L channel (Lightness) on an image in LAB mode would provide superior results. This will sharpen without color artifacts. This is the same idea as using the K channel in CYMK, but will be much smoother - LAB has a much larger gamut. Convert back to RGB or direct to CYMK when you're done."
I agree with him 100%. Compare the edges of the balloons on both images carefully. On the left side I applied Unsharp mask on RGB image directly. On the right, first I converted the image to Lab Color mode (Image/Mode/Lab color) and than selected the Lightness channel in the Channels palette and applied the same Unsharp mask filter. You can spot that on the right the edges are much cleaner and sharper.
Previous posts about enlarging and improving images are here: part one and part two.
Thanks Allan for the valuable tip!
- Ivan's blog
- Login or register to post comments
Hi, excelent article!
I've plugged it at www.redpixel.com.br
Direct link:
http://www.redpixel.com.br/newsdetails.asp?newsdate=06/07/2004&newstime=10:51:29
Best regards,
Filipe M. Deschamps
I'm glad you found it useful. Thanks for the link!
Oh, I can see the article being translated. Cool!
Since sharpening involves heightening contrast, when that's done on color channels the result is that adjacent colors more more complementary. Reddish-brown next to greenish-brown becomes red next to green, for example. If sharpening is done in RGB it can lead to color artifacts where what is wanted is heightened value contrast. That's why working just on the L channel in LAB is helpful, because all there is there is value. If you sharpen in RGB or CMYK, it's a good idea afterward to "fade" that effect, selecting 100% and luminosity as the parameters for the fade. This retains the sharpening, but converts the unwanted color contrast achieved in the sharpening into value contrast. Read Dan Margulis for more on this.