On the resolution and sharpness of digital images...
Moderators: rjlittlefield, ChrisR, Chris S., Pau
Thanks Rik. I have trouble following threads too sometimes. The text communication on forums always seems disjointed to me. If you aren't sharing the same point, the view can seem strange. This topic which we are trying to figure out is a very complex one, as I mentioned before. I guess we may not be just "tuning in" to each other. I am thinking "x" and you "y."
Will
- rjlittlefield
- Site Admin
- Posts: 23603
- Joined: Tue Aug 01, 2006 8:34 am
- Location: Richland, Washington State, USA
- Contact:
Yep, this stuff is hard to figure out and communicate even you're sitting in front of the same computer, pushing buttons on the same camera, and sketching on the same whiteboard. In some sense it's a miracle that it works at all over the internet. On the other hand, there's probably no more than one other person within 10 miles of me who cares about this stuff, and I have no idea who that individual might be, if in fact they exist! So we make do.
I'm glad we got that confusion settled down. I'll be offline for the next couple of days -- it'll be interesting to see what's come up when I get back.
--Rik
I'm glad we got that confusion settled down. I'll be offline for the next couple of days -- it'll be interesting to see what's come up when I get back.
--Rik
- rjlittlefield
- Site Admin
- Posts: 23603
- Joined: Tue Aug 01, 2006 8:34 am
- Location: Richland, Washington State, USA
- Contact:
rjlittlefield wrote:Max Lyons reports that when confronted with a 300 dpi wall-sized print of his giga-pixel Bryce Canyon
BTW, just to be fair to Max, his interests are in generating beautiful images that also happen to be highly detailed. Check out his web site for examples. As far as I know, Max has no association with the Gigapxl project, which I'll agree has goals that sound very much like what you describe.Epidic wrote:The GigaPixel project is not really about making good images, just really detailed ones to WOW the audience.
On the bright side, though, the Gigapxl project has done some incredible work in lens design and implementation. Each of their lenses is unique by intention. They design to the glass specifications, then measure the actual glass properties for their particular melt, reoptimize, grind the lenses, measure what they ground, reoptimize, and make a custom barrel to hold those particular elements exactly the right distance apart to be optimal for them. The process and performance is really quite impressive; see https://web.archive.org/web/20060408133 ... alized.htm for a description.
--Rik
Edit 11/12/2021, replaced broken URL http://www.gigapxl.org/technology-realized.htm with corresponding Internet Archive URL.
Interesting thread and comments.
I guess once I took a K25 slide and placed it under a low powered microscope to view a car number plate taken from a good mile away, I've never even worried about the resolution of digital, why can remain a digital shooters problem and also why I've never entered into whats best, film or digital. Personally I'll have slow film any day and yet I shoot digital now and probably always will. Full sized sensor would be nice in a smaller camera though
All the best.
Danny.
I guess once I took a K25 slide and placed it under a low powered microscope to view a car number plate taken from a good mile away, I've never even worried about the resolution of digital, why can remain a digital shooters problem and also why I've never entered into whats best, film or digital. Personally I'll have slow film any day and yet I shoot digital now and probably always will. Full sized sensor would be nice in a smaller camera though
All the best.
Danny.
Worry about the image that comes out of the box, rather than the box itself.
- rjlittlefield
- Site Admin
- Posts: 23603
- Joined: Tue Aug 01, 2006 8:34 am
- Location: Richland, Washington State, USA
- Contact:
It's been quite a while since I shot slow film. ISO 200 was my old standby. Turns out that there's not a big difference between my 6.3M pixels DSLR and ISO 200 film. As I wrote here (just this weekend):nzmacro wrote:Personally I'll have slow film any day and yet I shoot digital now and probably always will.
The difference is due to the gradation. Now if I was just feeling rich enough to shell out for 12 Mpixels instead of making do with 6...Overall, this 6.3M pixels DSLR is approximately equal to the tested film, when imaging the same subject area. The DSLR has less resolution than film for high contrast features of the subject, but it captures more low-contrast features.
--Rik
Just came across this topic, here is my interpretation . . .
a B/W pair line pattern can be thought as square wave form, say of frequency f. To adequately sample it, by Nyquist-Shannon theorem, you need to have a sample frequency of 2f, and therefore, if you are using DSLR to sample it, you need four samples, ie, four pixels. and resulting four pixels, BB and WW.
a B/W pair line pattern can be thought as square wave form, say of frequency f. To adequately sample it, by Nyquist-Shannon theorem, you need to have a sample frequency of 2f, and therefore, if you are using DSLR to sample it, you need four samples, ie, four pixels. and resulting four pixels, BB and WW.
I am not sure what it means by above statement, for me, 2.05 pixes-per-line-pair is only sampling at same frequency (or maybe a bit more due to that 0.05) as signal, far below the theorem limit.So, to figure out how many pixels-per-line-pair yield acceptable contrast, we might want to worry not only about best case, but also average case, and maybe even worst case!
What the illustration shows, to my eyes, is that 2.05 pixels-per-line-pair clearly doesn't cut it. That's no surprise, since the Nyquist-Shannon sampling theorem says that 2.00 would be the absolute limit.
- rjlittlefield
- Site Admin
- Posts: 23603
- Joined: Tue Aug 01, 2006 8:34 am
- Location: Richland, Washington State, USA
- Contact:
Sorry, I just now noticed this almost 2-year old post.mjkzz wrote:Posted: Sun Apr 15, 2018
Just came across this topic, here is my interpretation . . .
a B/W pair line pattern can be thought as square wave form, say of frequency f. To adequately sample it, by Nyquist-Shannon theorem, you need to have a sample frequency of 2f, and therefore, if you are using DSLR to sample it, you need four samples, ie, four pixels. and resulting four pixels, BB and WW.
...
I am not sure what it means by above statement, for me, 2.05 pixes-per-line-pair is only sampling at same frequency (or maybe a bit more due to that 0.05) as signal, far below the theorem limit.
Unfortunately your interpretation is off by a factor of 2. Nyquist-Shannon says that you need to have 2 samples per cycle, not 4 samples as you've implied. With a DSLR, it translates to one dark pixel, one light pixel, for each black/white line pair.
The issue then becomes how dark and how light, that is, what contrast is captured, and that captured contrast varies as a function of how the B/W pairs line up with the pixel positions.
If you're extremely unlucky, at 2 pixels per B/W pair, you could end up with pixels positioned exactly on the B/W transitions, in which case you'll capture gray/gray -- no contrast at all! That pathological case corresponds to some of the fine print in the sampling theorem, which essentially says that for infinitely extended patterns, you need >2 samples per cycle at the highest frequency, not >=2 as we usually say. If you're into Wikipedia, see the Critical Frequency section of https://en.wikipedia.org/wiki/Nyquist–Shannon_sampling_theorem The 2.05 pixels per cycle that I used is an arbitrary number just a little greater than 2.
--Rik
I think your test results are pretty interesting Rik, even though I'm on the opposite side of the sharpness spectrum with my single frame images. I think that everyone has their own "pain level" where image sharpness and detail is concerned (even me) and at some point you have to start making trade offs to get the images you want, and is the casual viewer really gonna notice if something is missing. I can see someone who's into micro photography taking image sharpness and resolving power to an extreme though.
One more wrench to throw into the gears is light quality and angle. I've seen a lot of sharp images robbed of texture detail by poor lighting.
No matter how sharp am image is printing it poster size on matte canvas is a pretty good equalizer
As for post processing: I've been using Topaz Sharpen AI and Topaz Denoise AI. in that order, and they both do a really good job of image sharpening. I know that Topaz Labs recommends stripping out image nose first, but I see more detail (better image sharpening) by sharpening my images before pulling out the nose.
One more wrench to throw into the gears is light quality and angle. I've seen a lot of sharp images robbed of texture detail by poor lighting.
No matter how sharp am image is printing it poster size on matte canvas is a pretty good equalizer
As for post processing: I've been using Topaz Sharpen AI and Topaz Denoise AI. in that order, and they both do a really good job of image sharpening. I know that Topaz Labs recommends stripping out image nose first, but I see more detail (better image sharpening) by sharpening my images before pulling out the nose.