Digital Light & Color

I tried a search on "resize sharpen" and didn't find anything that directly addressed my question. When I take largish images, of order 5600x3700, and re-size them for projection at our club to 1024x768 bounding size, I get mixed results. If the image has 'simple detail' such as a macro of a flower, them resulting downsized image will typically look satisfactory and will respond well to a bit of sharpening. But if the image has 'complex detail, say a field of flowers, the re-sized image looks soft and does not respond well to simple sharpening. (The sharpening I use is some combination of LCE via unsharp mask with large radius (10-40 pixels) and small amount (5-25%) and bilateral sharpen.) I generally use the Lanczos 8x8 to re-size because, in theory, it should work best and I'm a sucker for theory. But I've tried the other methods provided in PWP5 with not necessarily better results. I really don't know if my problem is in re-sizing or sharpening technique. I recall reading somewhere that one should never size more than 10% at a time. I haven't tried that because it seems impractical to me for the size of change I'm talking about. Especially for lots of images. Has there been a discussion or work-flow presented on this forum that addresses this specific question and if so could someone point it out to me? And, if not, has anyone any suggestions?

Thanks, Peter

Resizing, especially by large amounts by necessity blurs small details in the original image since each pixel in the result must combine all the information in a range of pixels in the original. This lost detail cannot be recovered by sharpening - the best you can do is to produce a sharp smaller image. Personally, I usually prefer the simple Sharpen or Sharpen More for this task. Resizing by increments is, in my opinion, a worthless technique.

There are a couple of nuances that you may wish to consider/experiment for your down sized images:
(1) Try adding a Gamma correction to the Resize workflow of 48-bit color images described in this previous message board thread here: http://www.dl-c.com/board/viewtopic.php?f=4&t=240

(2) When sharpening and/or local contrast enhance-ing the down sized image, use an 'active' mask with the default diagonal Brightness Curve applied keeping the mask white amount half to two-thirds of the mask black amount... this often times allows more aggressive adjustments without strong white edge pixels or obvious light halos.

For the following illustration, a 1728x2304 pixel dimensioned, 48-bit color image was Resized using default Bicubic settings to 200x267 pixels [an 8.64 reduction]. The left half has no gamma correction and Sharpen Amount = 65 with no mask. The right half has gamma correction and Sharpen with a mask described in (2) with mask white amount = 40 and mask black amount = 75.... the left/right differences are difficult to see at the required posting jpeg compression but become readily apparant when an AbsoluteDifference comparison is made of the tiff image versions...

A bit off topic: Den reminded me of another simple technique of keeping white halos at bay when sharpening using USM: composite the original image with the USM-sharpened one in Darken blending mode.