Forum Moderators: not2easy
I recently started a project that requires creating lots of book cover thumbnails. Lots of image text and no original PSD files with the font still active (the best way to get small image text).
My normal approach has been to take small steps in resizing down (30% or less each time), and to apply the Sharpen Edges filter between each step. That works pretty nicely for preserving image detail, but text starts to break up and can end up looking pretty funky.
This new approach I stumbled onto uses L*a*b color space. By converting the image to L*a*b from RGB, you end up with all the important detail in the L channel - which contains Lightness data only, no color data. Choose just that channel and then click the "eyeball" on the top again so your display shows the final result, but your actions are only affecting the L channel.
First, this is a better way to run the Sharpen Edges filter, becuase you don't introduce any color artifacts. But I discovered you can sometimes get remarkable clarity in the text by adjusting Brightness/Contrast (again, just on the L channel). With some combinations of text color against background color, it takes a combination of the two approaches to get best results, but my book covers are now looking better than anything I ever turned out before.
Using this technique you will get some color shifts -- hue and saturation will remain the same but the lightness or shade will change. When this is unacceptable in a given case, I use a mask to isolate just the letters themselves (including all their attendant anti-aliasing pixels).
I much prefer to reset the font using the text engine once I've arrived at the destination size for the image. But when I don't have that luxury, this approach works wonders.
L*a*b (A color space I never did use before...
A lot of people avoid it - it just seems so alien. And yet, it is Photoshop's native color space, and incredibly flexible. It also handles the widest color gamut.
Because it is the native space, you can convert into and out of the space unlimited times and not lose any image data - and that's not true for any other color space.
You can often enhance subtleties in images very nicely by running Curves on the L channel. This is also useful at times with image text, especially when it's superimposed on an image instead of a flat color background.