Allan Bruce wrote:
Obviously more data is going to yield better results over less but it just means that sharpening (particularly) doesnt work optimally with the higher density sensors unless you have really sharp glass!
That depends on how big you are printing (i.e. whether your printer can resolve as much as the camera), and how wide a radius you set in the USM controls. Sharpening radius is very important, as it determines what scale of thing gets sharpened. Sharpening with a 1 pixel radius means something quite different in 12 and 21 megapixel files. 1 pixel is a smaller sharpening radius in the 21 Mpixel file. To sharpen both files optimally for a given print, you might need to set a (28%) larger sharpening radius on the 21 Mpixel file.
I don't find any significant difference between Pixel Perfect's high pass method and USM as long as each is done right for the print.
Pixel Perfect wrote:
Luma sharpening using luminosity blend mode is a faster and non-destructive alternative to switching to LAB mode and sharpening the L channel and then switching back to RGB mode.
I do most of my post processing in LAB, and I don't see any ill effects at all from switching back and forth between RBG and LAB. In fact, I just did a test. I isolated one pixel in a full image in Photoshop, converted it back and forth between LAB and RGB 5 times (10 conversions in total), and nothing changed, not the HEX value, RGB value, nothing. I assume what is true for one pixel is true for them all.
Also, doesn't Smart Sharpen sharpen only the luminosity channel? I think I read that somewhere.
StevenPA wrote:
I assume what is true for one pixel is true for them all.
When switching between colour spaces or models, that assumption would sometimes be wrong. For instance if you switch a black pixel and a bright purple one from a wider gamut to a narrower, the black pixel might retain its value, whereas the purple one might not. In this case though, I agree with you that the RGB->Lab->RGB transformation appears to be pretty much conservative across the whole gamut. It ought to be, because Lab was set up to fully encompass human vision, and more. On the wikipedia page it suggests that there are noticeable rounding errors in 8bit conversion, but not in 16bit.
There's a difference when looking at weddings I shot in the past with my 5D compared to the 5D II I now use at weddings. I can shoot ISO 3200 without any hesitation whereas before I really didn't like using it.
I've found the 5D MkII needs NO sharpening applied in certain situations or a bare minimum (especially stuff shot with primes stopped down, or even zooms stopped down). Whereas the 5D required much more sharpening in my experience.
In fact it was really bothering me how "oversharp" the MkII stuff was looking until I really backed off the DPP default of "3" for Standard !
This thread (so far at least ) is a wonderful example of the quality of the posters on this forum - with sensible reasoned and informative comments. We should all congratulate ourselves for being here.
Where can I learn more about sharpening in general and luma sharpening in particular?
I am guessing, but I would instinctively imagine that most of the need for sharpening is due to the blur from anti-aliasing in digital cameras, the effect of which would presumably (by design) be roughly proportional to the pixel pitch, unless the lens resolution is little better or worse than that. In this case, wouldn't the best sharpening (of a technically sharp image) be more or less constant in terms of pixel size?
OTOH, if the image is less than sharp due to poor lens resolution, misfocus or blur, beyond that of ainti-aliasing, all of which would be maginified at the pixel level in the case of a higher-resolving sensor, then bigger sharpening radii would be appropriate ... or am I off on the worng track somewhere?
FashionBoy - Yes! That is true. Actually after my impression that I was getting more noise in images I stopped using one action of sharpening because it appeared to be unnecessary.
rbraun wrote:
Brainiac said: "To sharpen both files optimally for a given print, you might need to set a (28%) larger sharpening radius on the 21 Mpixel file."
Thanks Brainiac. I was looking for info like this. So if I were to run 100,1,1 on 5D then it should be 100,1.3,1 on 5D2 right?
They should be more or less equivalent in print. However, the presence of more detail in the 5D2 file might make you feel less need to sharpen. It's a subjective thing in the end. I think the best thing is to use your eyes, but remember that sharpening happens on a smaller scale the more pixels there are. I quite often make a hidden sharpening layer and paint it in where I want it. There's nothing to stop you making two or three hidden sharpening layers with different radii of usm on each. The smallest radius might be appropriate for areas of very fine detail, whereas if you want to get a bit more out of less than optimally sharp areas you can use a wider radius usm layer.
But by and large, you should be able to settle on a standard sharpen action that makes you happy. I apply such an action AFTER doing an image size to set the dpi to 360 and the dimensions to my print size. In other words, sharpening should be done for a particular print size, as much as for the image itself.
IMO the camera with more pixels resolves finer detail and often needs a smaller radius. In LR I use 0.5 - 0.7 pixels for the 5D and 0.7 - 1 pixel on the 1D II. Usually there are many scales of detail that needs different numbers. I'll run high pass twice often with 0.5-0.8 pixels and 1.5-3 pixels
!
) is a wonderful example of the quality of the posters on this forum - with sensible reasoned and informative comments. We should all congratulate ourselves for being here. 