DQE

Joined: 08 Jul 2008 Posts: 1433 Location: near Portland, Maine, USA
|
Posted: Fri Nov 11, 2011 7:01 pm Post subject: |
|
|
Rik,
Thanks for the extended reply.
You've given me much to think about in trying to merge my mostly disjoint radiographic imaging experiences and understandings with photographic imaging and macro stacking.
We also had to pay very extensive attention to issues related to visual perception, specific radiographic application requirements, etc, etc, including mandatory clinical requirements as well as clinically-perceived "wants". Not always an easy set of things to connect and merge! But it does make the field interesting. _________________ -Phil
"Diffraction never sleeps" |
|
Chris S.
Joined: 05 Apr 2009 Posts: 1084 Location: Ohio, USA
|
Posted: Sat Nov 12, 2011 12:14 am Post subject: |
|
|
In case anyone wonders, in the test Rik mentioned, "Skip=1" was 0.25 microns on a Mitutoyo 20x/0.42 objective. The test subject was moth wing scales. Optimal results came in at Skip=8, which was an increment of 2 microns. I could also live with Skip=12 (3 microns), with only slight chagrin. Skip=16 (4 microns) was a bit softer, but without obvious banding. By Skip=32, banding was obvious and quality was noticeably poorer. To emphasize: this was on a 20x objective, not 40x.
Eventually, this test will get posted, perhaps with a couple of others. Meanwhile, I'm working under a hypothesis that I think encompasses Rik's three-step strategy and the results I've seen so far: The point at which banding disappears represents the largest useful increment, but probably not the one that provides the very highest quality. Smaller increments can increase quality--at least for some subjects--until a point where going smaller only increases accumulated noise; this point represents the smallest useful increment. Any work done between these upper and lower bounds is likely to produce good results, with the optimal tradeoff depending on circumstances.
While I can't speak for another person's hardware or workflow, I suspect that if Des' specimen were shot on my rig, increments of 1.1 and and 1.5 would produce a small but observable difference in quality that would be real (that is, representing additional visual information, and not simply additional noise) I also doubt very much that it would dramatic. Would it be significant or worthwhile? That would depend on the photographer's goals.
Doggone it, Rik, did you have to go and write this bit:
| Quote: | | . . .shoot the quietest images you possibly can, average together a bunch of nominally identical images to further cut noise. . . |
Now I can't help but think of bathing my sensor in liquid nitrogen and shooting multiple images at each increment for pre-stacking noise cancellation. Maybe I could reduce the optimum from Skipfactor=8 to Skipfactor=4. Hey, a person examining a large print with a magnifying glass might notice a bit of improvement!
Cheers,
--Chris |
|