Let's Talk About...sharpening

We're all in student mode SD. This is why I started the thread, as a way to get some groupthink on the problems and maybe get a better understanding of sharpening in general. It is of course a very interesting subject, both in depth and in the implications it has to other areas of imaging.

Sounds like I need to get more familiar with downsizing algorithms. The damage they do to the image may be simply that I'm not using the best method. This definitely is one of the factors affecting sharpening...

What algorithm is used by various image processing software, etc to view the image using "shrink to fit page" type of downsizing?

If this were an earlier version of Windows, my immediate reaction would be to search the whole computer for .dll files which had changed today. Those are Dynamic Link Libraries, composed of functions which multiple softwares might reasonably be expected to use - Save As functions etc. - put in place to enforce commonality among multiple coders writing software for Windows.

People used to write software-specific .dll's, kinda defeating the purpose, but intended for use only in the software's installed directory so no other programs used it. A lot of times, though, they got out into the wild and created compatibility havoc all over the system. Sometimes the guys who write the installation software aren't aware of that need, and write the installer to put bespoke files into common directories. It could then function with perfect normalcy until you install that one_specific_program with which that new .dll doesn't play well, and havoc ensues.

I've had more than a few merry chases to corral those. Don't know if it's still a problem or not, because I haven't run into one in years. A good starting point, though, would be to immediately search for all files which have changed today, and puzzle through that a bit.

rpmsrpms - if you have the time, I would recommend running "Dependency walker" on the Nikon ViewNX2 application and see if it tells you anything is out of order. Send me an image capture of the dependency tree and I might be able to help.

Guys...thanks for the suggestions, but this latest mishap reminded me of the 5 other weird problems ViewNX2 has plagued me with in the last few years, so I dumped it. I had gotten used to using it with my Nikon cameras, but most of my editing work now is with Canon, so I am just going to use DPP. It has built-in RAW processing for the T2i and XS and a couple other nice features that I just discovered, so maybe this was a blessing in disguise.

Back to sharpening...I have also read that sharpening before and after is best. But at this point I need to say that the improvement seen by adding even more steps to the workflow would need to be justified for me to consider making a change. I plan to experiment.

Yes, thinking about it more I guess sharpening, being just a shift in contrast between specific pixels, could be reversed with an inverse algorithm, so it is probably not lossy.

I took a recent pic of a 2014 Dime Clash and processed it for sharpening in multiple ways. I did no levels adjustment, so these look a bit dark as I'm currently shooting -1EV to avoid hotspotting:

No sharpening, just downsizing 4x using DPP


No processing before downsizing, then sharpened 70/500 after downsizing


Sharpened 250/500 before downsizing, then sharpened 35/500 after downsizing


Sharpened before downsizing 500/500


I chose 500/500 before downsizing as the reference somewhat arbitrarily. It does not look oversharpened to me, though it's on the edge of what I'm comfortable with. The others are attempts to achieve the same level of sharpening with just post-downsizing and with a combination of pre and post.

Since it's hard to see much difference at 100%, I made an animation of a crop at 200%:



I have a hard time, even at 200%, seeing much difference between the 3 sharpened images. What this tells me is there is little or no disadvantage to just sharpening after downsizing.

Ray, my major takeaway from this is to feel the need to reevaluate my own definition of what these things *should* look like. I can see the differences clearly, I plainly like the one sharpened at full size the best, but I can't any longer say for sure that it actually is the best.

That, plus the certainty that none of this really matters for Forum posting. Nobody in the building would call the unsharpened shot anything but "wonderful," outside this subforum.