Home

Please support mflenses.com if you need any graphic related work order it from us, click on above banner to order!

SearchSearch MemberlistMemberlist RegisterRegister ProfileProfile Log in to check your private messagesLog in to check your private messages Log inLog in

Analogue lense resolution in pixel by circle of confusion
View previous topic :: View next topic  


PostPosted: Tue Jan 26, 2021 5:02 pm    Post subject: Reply with quote

visualopsins wrote:
The question is How much less magnification is necessary to enlarge 24mp image to normal print size as, say, image from 8mp sensor -- and, is the depth-of-field in the two prints identical? (The coc of 24mp is smaller but gets enlarged less)


I am not shure whether I get you right. The idea of CoC is based that the human eye can resolve 1500points or lines in the diagonal of the picture. That is why e.g. the CoC of fullformat is sqrt(24^2+36^2)mm/1500=0.029mm. Thus depth of field should be always the same when looking at a picture from "appropriate" distance.
Best regards
Andreas


PostPosted: Tue Jan 26, 2021 6:10 pm    Post subject: Reply with quote

pandreas68 wrote:
visualopsins wrote:
The question is How much less magnification is necessary to enlarge 24mp image to normal print size as, say, image from 8mp sensor -- and, is the depth-of-field in the two prints identical? (The coc of 24mp is smaller but gets enlarged less)


I am not shure whether I get you right. The idea of CoC is based that the human eye can resolve 1500points or lines in the diagonal of the picture. That is why e.g. the CoC of fullformat is sqrt(24^2+36^2)mm/1500=0.029mm. Thus depth of field should be always the same when looking at a picture from "appropriate" distance.
Best regards
Andreas


However, the coc in 8mp image covers fewer pixels than in the 24mp image. For same size prints viewed at the same distance, the pixels inside the coc of 8mp image get enlarged more than inside the coc of 24mp image.


PostPosted: Wed Jan 27, 2021 3:19 pm    Post subject: Reply with quote

visualopsins wrote:


However, the coc in 8mp image covers fewer pixels than in the 24mp image. For same size prints viewed at the same distance, the pixels inside the coc of 8mp image get enlarged more than inside the coc of 24mp image.


Yes of course you are right. If I calculate for fullformat then I achieve 2.85Px CoC diameter at 8MPx, while at 10MPx some 3.1 and at 24PMpx some 4.8px and depending on that different magnification for a certain print size is needed.

I am not an expert in that sensor stuff, but I'd say for filter algorithms the neighboring pixels may be always involved, so that for my "technical intuition" it does not make sense to use a resolution which leads to less than 3 Pixel in the diameter. Therefor the minimum resolution I'd see at 10-12Mpx to have at least one neighboring pixel. 24MPx may also make sense to have some 2 neighboring Pixels. Much more than 24MPx should not bring much quality enhancement at least not when not cropping while relying on the standard CoC diameter. Of course, the more Pixels in the CoC diameter the more resolution is possible beyond CoC thinking maybe just as in times of high resolution black white films. While the additional resolution of 24MPx may be still easily seen by many I doubt that additional resolution can be noticed beyond 50MPx as long as the picture is not cropped and is watched from appropriate distance. Best regards, Andreas


PostPosted: Wed Jan 27, 2021 3:51 pm    Post subject: Reply with quote

Ernst Dinkla wrote:

on SQF:
https://www.imatest.com/docs/sqf/
If


Thanks a lot for that link! Now I have a lot to read (and understand...). Via your link I also came to

http://www.bobatkins.com/photography/technical/mtf/mtf1.html

There I'll start...

Also interesting that there is an IEEE Camera Phone Image Quality Group thinking about that.

I think I also understand now your suggestion. In my words, you mean that for a lens and a certain f-stop a certain paper size can be provided which can be printed in 450dpi, and from this it could be upscaled respectively downscaled to each size respectively resolution according to change of picture height, don't you? Thinking in picture height instead of diagonal also simplifies the things for me.

Best regards,
Andreas


PostPosted: Thu Jan 28, 2021 2:56 pm    Post subject: Image Viewing Distance Reply with quote

I've read some of the stuff. I think I understand the philosophy and the ideas behind acutance and subjective quality factor (sqf), but I have some difficulties with some details.

Maybe the most important detail is the viewing distance. When coming from the rule that a picture should be watched from 1,5 times its diameter at a distance, the viewing angle stays constant, and therefor the needed print resolution decreases as the image size increases while the necessary camery respectively lens resolution stays constant. (eg. refer to https://www.lightscapecreations.co.uk/blog/2019/10/the-perfect-viewing-distance).

So far I've understood the sqf as described in http://www.bobatkins.com/photography/technical/mtf/mtf4.html the viewing distance is to be assumed constant with 34cm. This explains why the necessary resolution of camera respectively lense increases with the necessary magnification. Unfortunately I could not access the original article from Granger.

In https://www.imatest.com/docs/sqf/ however there is stated that viewing distance is not fixed in sqf calculation while also not known at least regarding the sqf calculations for lens quality tests of Popolar Photography.

For acutance viewing distances are used as e.g. defined by distance = 30cm*sqrt(picture height/(10cm))

When plotting these two assumptions it looks like this:



So somewhere at a print height from 50cm we'd need to double the resolution according to half viewing distance.

I could not yet find where the square root approach comes from. Maybe from IEEE spec IEEE 1858-2016 - IEEE Standard for Camera Phone Image Quality, which I can't access for free.

Hm. Does it make sense to think about other viewing distances as the typical 1.5 times diagonal?