by Mike McNamee Published 01/04/2015
Solving the Sharpening Conundrum
Given our misgivings about judging sharpening levels, what is the printer/post-processor to do? At Epson Print Academy we have a suite of images sharpened for inkjet printing at full 20x16-inch size. A ring around made with an action in Photoshop sharpens at 12 levels of strength and the delegates are invited to judge the most pleasing level. A number of patterns have emerged from the 500+ delegates who have viewed the test prints.
Pairs of viewers rarely agree on the 'best' level between Amount 40/ Radius 0.5 and Amount 80/ Radius 1.0. In other words the optimum pre-sharpening lies somewhere between the two values.
We use the Pixel Genius method of sharpening which can have a High Pass layer opacity between 0% (no output sharpening) and 100% ('full strength' output sharpening). Delegates rarely stray above 67% opacity but usually do not agree between increments of 33% - in other words they could not decide between say 20% and 40% so there is little point in changing by increments of say 5%. The choices depend upon subject content and its detail frequency but usually lie between 33% and 67% opacity.
Architectural shots made with Nikon D800 produce a lot of confusion with delegates unable to differentiate between Amount 40/Radius 0.5 and Amount 150/Radius 1.5, the latter being towards the limits of pre-sharpening. In other words, for this type of image, the D800 can take more sharpening because of the inherently high resolution of the file (we look forward to working out what to do with the new Canon 50mp files!!).
Much of the above goes out of the window with images shot at 6400 ISO where noise is so great as to rapidly deteriorate image structure at quite low levels of sharpening.
While it is difficult to set out absolute values, we find that female portraits (especially with fine hair) require the least sharpening strength, followed by butterflies and then by low-detail frequency images such as the test 'microscope' shot.
None of the above is of great use to the printer poised to set their values so we include starting values for experimentation - they might well suffice for the majority of work without modification.
What to do with the D800?
The Nikon D700 has images 4,256 pixels wide but the D800 covers 7,360 pixels. In theory therefore if a scene value is 1.0 pixels Radius for a D700 then the D800 would require 1.73 pixels. However, the noise of the D800 should be proportionally more as it offers a smaller pixel area and this limits the amount of sharpening that you might apply to a D800 image! In trying to resolve this we made sharpening ring arounds at an even larger range of settings. The initial scoping trials showed that noise became obtrusive on the image for all pre-sharpening values of Amount 150 Radius 1.5 (ie the entire right-hand column). Additionally the 67% and 100% Opacity value for output sharpening at Amount 80/Radius 1.0 was slightly over-sharpened also. With this in mind we repeated the ring around using values of pre-sharpening of 40/0.5, 60/0.7 and 80/1.0 with the same output sharpening values. The optimum was then judged to lie in the zone indicated but we could not in all honesty differentiate between the five remaining value sets. Amid all this confusion we have only looked at two simple parameters and not even brought Clarity into the equation - if you are confused you are in good company! The message here is don't let anybody ever tell you that the optimum sharpening is 'values x-y-z', we doubt they have scoped things out sufficiently diligently to ever be able to make such a pronouncement. Faced with this confusion you are probably best to start at Amount 60/Radius 0.7 and about 40% opacity, proof out and work from there.
There are 0 days to get ready for The Societies of Photographers Convention and Trade Show at The Novotel London West, Hammersmith ...
which starts on Wednesday 15th March 2023