also, for challenge images, generally I will not sharpen at full-size. i think that compressing the already sharpened image creates an output image of lower quality than one sharpened at 640. Still, at 640 my image will stay as a 16-bit file so that i can utilize as much pixel information as possible. The very last step for me before submitting is to change it from 16- to 8-bit.
With my previous cam (Canon 10D) I used some moderate sharpening in RAW conversion, since the AA filter on that cam was rather strong. No other sharpening prior to resizing, and image remains in 16-bit/channel mode. After resizing, I apply final sharpening prior to 8-bit conversion. Settings dependent on file, but usually either:
Radius = 0.2
Amt = 200 to 500
Threshold = 2 to 5 (depending on noise level)
or:
Radius = 0.3
Amt = 100 to 250
Threshold = 2 to 5 (depending on noise level)
With my current camera, the RAW files most often need *no* sharpening on conversion, only applying sharpening after resizing. Also note that the algorithm and steps used in resizing have a huge impact on the amoutnt and type of post-resize sharpening required.
In some cases, I mask off areas that I don't want sharpened, or use the history brush at 33% opacity to partially roll back the sharpening. For a images where a soft-focus effect is desired, I run higher sharpening amount.
There are also some special cases, like night-sky images, that require some real dinking around to optimize the look for a 640px (or 720px as case may be) environment. I'm still developing new techniques to improve the final results.
Can someone explain what the threshold level actually means please ?
"Threshold" defines the degree of contrast necessary for PS to consider something an "edge" and sharpen it. Try this: open an image up, magnify it to 100% on the screen, open up USM, and move the threshold slider all the way to the right, then back all the way to the left. See what's happening?
"Threshold" controls the subtlety of the edge-finding, "Amount" controls the degree of sharpening (contrast enhancement, basically) applied to the edges.
To add to what Robert said, open an image that has visible noise (shot at high ISO, for instance) and call up the USM dialog. Set the following:
Radius = 0.3
Amount = 250
Threshold = 0
Now increase the threshold, and watch the noise. You'll find that as you increase the threshold, the high-contrast edges still get sharpened, but the lower-contrast noise does not. This is why I always recommend a non-zero threshold for most all images. Usually, 2 to 5 for threshold works well, unless there is substantial noise.
Sharpening after resizing has the advantage that noise, which is usually very high frequency, is averaged out by the resizing process. This makes the sharpening process less prone to make noise more visible.
OK, a number of youse guys say that you re-sharpen a second time after you've resized your image for the challenges.
i've never done that...
Many, if not most, photographers sharpen their images the FIRST time after croping and resizing their images.
That is because the amount of sharpening needed for an image resized for a 640 pixel DPC web graphic is a lot different than for one resized for a 20 X 30 sized print.
OK, a number of youse guys say that you re-sharpen a second time after you've resized your image for the challenges.
i've never done that...
Many, if not most, photographers sharpen their images the FIRST time after croping and resizing their images.
That is because the amount of sharpening needed for an image resized for a 640 pixel DPC web graphic is a lot different than for one resized for a 20 X 30 sized print.
Are you not counting "bicubic sharper" during the resize?
Can someone explain what the threshold level actually means please ?
"Threshold" defines the degree of contrast necessary for PS to consider something an "edge" and sharpen it. Try this: open an image up, magnify it to 100% on the screen, open up USM, and move the threshold slider all the way to the right, then back all the way to the left. See what's happening?
"Threshold" controls the subtlety of the edge-finding, "Amount" controls the degree of sharpening (contrast enhancement, basically) applied to the edges.
I think I have that right :-)
R.
That's basically right. The "Diameter" setting controls how far to each side of the edge boundary the effect extends.
I also only sharpen after resizing to entry-size.
I usually run a "contrast-enhancing" USM pass at 12-16%, 48-64 diameter, and a threshhold of 0. These seems to tweak the contrast and brighten the image in a subtly way which the Curves control seems to miss.
If the details still need some sharpening I run a more traditional set of USM settings: 66-88%, diameter of 0.6-1.3, and a threshold of 5 (or 7 if there are fleshtones or other smooth gradients). For a print image, I'd use the same settings, but a slightly larger diameter setting, since the file is so much biggr and has many more pixels to play with.
Are you not counting "bicubic sharper" during the resize?
It's debatable whether using "bicubic sharper" is really sharpening. My experience says it is not, but it *does* greatly affect what's needed as far as post-resize sharpening. The number of steps used in resizing also impacts that. Given that you can choose to resize in a single step, or any number of discrete steps, and at each step use a different resampling algorithm, the possible combinations are unlimited. Some combinations will work better than others, but different strategies may be required for different images.
... Are you not counting "bicubic sharper" during the resize?
Correct, I don't and that counts for "bicubic smoother" to. All bicubic is intended to do, I believe, is to more accurately render the image to the resize.
