wing, 4x vs. 7x

[NikonUser]

EDIT: was "Technical Quiz"
4x objective vs enlarger lens, which is best?
I guess it depends on how one wishes to use the image, nevertheless:

Goal: to get a 1024px image of a mosquito wing

Subject: wing length 5mm, permanent Canada Balsam mount on slide

Camera: Nikon D610, 35x24mm sensor, 24.3 million pixels

(my) Possible routes:
1) use a 4x CFN Plan Achromatic Nikon objective; magnification on sensor=4x
2) use a Rodenstock 40/2.8 Apo Componon @ f/2.8 (fixed aperture); with long bellows such that wing fills entire frame width-wise; magnification on sensor=7x

Procedure: the 4x image has to be reduced from 2840px wide to 1024 px
the ex Apo Componon image has to be reduced from 6016 pixels wide to 1024 px

Question: does one get a higher quality 1024px image by compressing (if that is what is happening) a 6016px image from the Apo Componon;
or a better image from compressing the 2840px image from the Nikon objective?

here are the original images (on the 36mm wide sensor) from the 2 lenses:

[elf]

How many different downsampling routines did you try? Would combining detail from different routines give you a better image?

[rjlittlefield]

As an exercise in analysis, I'll offer the following.

At 7X, your f/2.8 Apo Componon will be operating at 2.8*(7+1) = f/22.4 effective. That may seem pretty far stopped down, but on the camera side it's NA = 1/(2*22.4) = 0.022321, which on the subject side is NA = 7*0.022321 = 0.15625 .

In contrast your 4X CFN Plan Acromatic objective is probably NA 0.10 .

So, with the Apo Componon you're running with a much larger NA, using more pixels on the sensor, and using an apochromatic lens instead of just achromatic.

Barring a dud lens or other defect in the setup or processing, I'd bet on higher image quality with the Componon, hands down.

That said, at only 1024 pixels wide, I'd expect that almost identical images could be produced with both lenses, albeit with different postprocessing such as filter parameters.

If your actual results are not consistent with this prediction, I would be interested to know more.

--Rik

[NikonUser]

Wow Rik, spot-on analysis. I wasn't sure which would be 'better' (bugs I can handle; math just simple plus and minus).
The Nikon 4x finite is actually 0.13 NA.
At 1024px both images look the same.
Looking at actual pixels the Componon wins (last image)

Each stacked with ZS PMax

[steveminchington]

This is an interesting article about Down-Sampling and explains how to get better results when reducing an image.

http://bvdwolf.home.xs4all.nl/main/foto ... sample.htm

To me, the more pixels you start off with, the more have to be removed in the down-sampling process which could increase the risk of getting unwanted effects. Maybe better to start off with a smaller image. That's my backward logic!

[rjlittlefield]

http://bvdwolf.home.xs4all.nl/main/foto ... sample.htm

That's a good reference and a trustworthy source. Bart van der Wolf knows his stuff rock solid forward and backwards.

To me, the more pixels you start off with, the more have to be removed in the down-sampling process which could increase the risk of getting unwanted effects. Maybe better to start off with a smaller image. That's my backward logic!

Yeah, that's a bit odd way of looking at it. From the standpoint of image quality, you don't lose by starting with more pixels, as long as you use an appropriate downsampling method. The risk in this case is that somebody will choose a method that is not appropriate.

Perhaps surprisingly, the downsampling method that I currently use in Zerene Stacker for pre-sizing is a good example of how not to do it, when the pre-sizing fraction gets small. That's because ZS does not anti-alias when resampling. As a result its method works OK for fractions close to 1, but when you get down to say 1/2 or 1/3, images can easily develop Moiré or "jaggies". There's an entry someplace well down on the "to-do" list to add anti-aliasing. The priority would be higher if users complained about it, but they don't. That may be because the most common use for drastic downsizing is to make stereo pairs, and in that case the increase in apparent sharpness caused by not anti-aliasing can actually make the stereo lock up better than it would if processed optimally for other purposes. It's a complicated world.

--Rik