Bicubic Sharpen when using PKS

[soboyle]

When using photo kit sharpener should I avoid using bicubic sharpen when reducing images for web output or when printing small prints? Since it's an uncontroled sharpening of the image, it seems better to reduce image size using bicubic, and then rely on the output sharpeners to do their job.

[Robert Kinchin]

Soboyle,
Its my understanding that bicubic sharper is just a interpolation method that reduces aliasing when reducing an image. I dont think its a sharpener and I dont think you can remove the aliasing after the size reduction with a sharpener.

Similar, bicubic smoother cleans up jagged edges caused by up sampling

Robert

[Jeff Schewe]

There is no simple answer to your question. . .it all depends on your starting size and ending size.

I've found that downrezing beyond 50% should not be done in a single pass for optimum results. I've also found that multiple passes of bicubic sharper will build up artifacts pretty easy.

So, depending on your starting size and ending size, I would suggest several (or more) passes at 66% to get "close" to the final size and use a final pass of not more than 50%-meaning between 50% and 66%-of bicubic sharper to get the final exact dimensions. I would then apply the final output sharpening based upon the final output type and media.

When we designed PhotoKit Sharpener, Photoshop CS was still in development. And while we were of course testers, we simply could not find a fixed specific formula for suggested downrezing. The bicubic sharper option is very useful, but tricky to use since you need to be careful of multiple applications of that specific algorithm. It really depends on your starting and ending resolution.

On the flip side, uprezing has gotten better and simpler. Either do it in Camera Raw (if shooting raw) or using bicubic smoother in a single stage. You DON’T want to use step interpolation with bicubic smoother!

[soboyle]

I'm shooting with a 20D at 3504 x 2336 pix, and usually size for the web at about 500 pixels on the long side, so I am reducing to a size of about 14 percent of the original, it will be 4 full steps plus a tweeked step reduction to get there using the 66 percent method. I'll give that method a go and see how the results look compaired to a one step reduction. Hopefully the 5 steps are not enough to build up the artifacts you mention.

You mention uprezing in Camera raw, is there an advantage to doing it in camera raw vs photoshopCS?

[Jeff Schewe]

In general, the Camera Raw uprezing is considered superior to Photoshop Bicubic Smoother. However this is based upon over-rezing to slightly beyond your final needed size. Because Camera Raw can not resize to specific dimensions, you'll need to to final exact dimension and ppi resolution in CS.

[Scott.Southerland]

So, depending on your starting size and ending size, I would suggest several (or more) passes at 66% to get "close" to the final size and use a final pass of not more than 50%-meaning between 50% and 66%-of bicubic sharper to get the final exact dimensions. I would then apply the final output sharpening based upon the final output type and media.

Jeff, would you mind discussing the reasoning behind 66% steps? This is the first I've heard of 66% used as an optimal percentage for resampling.

[Jeff Schewe]

No particular reason for the 66% specifically except when doing a stepped downrez, I'm only killing 1 in 3 pixels and I have a lot of empathy for pixels. . .

A 50% downsize kills every other pixel.

Imagine a firing squad, at 50%, half your friends would be killed at a time. Terrible!!!!!

[Scott.Southerland]

Aha! Thanks for the heads up.

... now excuse me while I explain to my pixels that 1/3 of them are going to die, and the other two thirds have to watch. :D