Sony AR7IV vs GFX 50 for DSLR scanning.

[babouphoto]

In this example, I compare a single frame rendition (top) with a pixel-shifted version (bottom) of the same scene. Again I took a 250 pixel wide sample of each, combined them into a composite, and resampled the results 4x. The shot was taken with a Sony AxRiii + Loxia 50/2. The conversion was made with PixelShift2DNG, and the samples were taken from the AWG and DNG raw images respectively, without sharpening. The difference is significant, but only with a side by side comparison.

 

It is noteworthy that there is no visible color aliasing in either image, even in the neutral toned areas.

 

[ATTACH=full]1319881[/ATTACH]

 

It is visible. My camera rental shop is closed today (duh) so my test will have to wait until next week.

[Ed_Ingold]

Color aliasing is visible? In my experience aliasing appears as flags on repetitive structures like picket fences and ironwork near the limit of resolution. Debayering effects are not, strictly speaking, aliasing, but may appear as colored spots in neutral areas. I see neither in this example. Manipulations were done in PSD mode, directly from RAW images, with JPEG conversion only as the last step. JPEG artifacts (e.g., halos) are present, but only at the pixel level.

 

For a sense of scale, this is the the original image, with the sample area outlined in red.

 

[digitaldog]

Color aliasing is visible?

Yes, absolutely possible depending on many factors, some of which were shown in my article where a scanning back and tri-shot vs. single shot camera were used.

More examples here:

Resolution, aliasing and light loss - why we love Bryce Bayer's baby anyway

 

With a Bayer filter, you get a fun color component to this aliasing.

The above is visible to me..... ;)

[Ed_Ingold]

That falls in the category of "repetitive patterns near the limit of resolution." If you were to photograph that pattern with film at the same magnification, the lines would converge into an indistinct blur, and would be faithfully copied as such with a camera or scanner, without aliasing. I'm not saying color aliasing doesn't exist, but it's pretty rare outside of a manufactured situation.

 

Nor would you see aliasing unless the resolution of the lens were comparable to or finer than the pixel spacing. The onset of aliasing is how you use the target above for focusing a video camera. (most cell phones resolve better than 4K video).

Edited December 1, 2019 by Ed_Ingold

[digitaldog]

That falls in the category of "repetitive patterns near the limit of resolution." If you were to photograph that pattern with film at the same magnification, the lines would converge into an indistinct blur, and would be faithfully copied as such with a camera or scanner, without aliasing. I'm not saying color aliasing doesn't exist, but it's pretty rare outside of a manufactured situation.

 

Nor would you see aliasing unless the resolution of the lens were comparable to or finer than the pixel spacing. The onset of aliasing is how you use the target above for focusing a video camera. (most cell phones resolve better than 4K video).

 

No, it's not as CLEARY illustrated by two articles. It's color aliasing! A well known attribute. Don't be foolish suggesting other.

 

"There are two ways to be fooled. One is to believe what isn't true; the other is to refuse to believe what is true.” -Søren Kierkega“

[frans_waterlander]

I don't want to rain on your parade, Andrew, but the name is Søren Kierkegaard.

[digitaldog]

I don't want to rain on your parade, Andrew, but the name is Søren Kierkegaard.

Well look who showed up to go off topic.

Yes Franz (Frans), that did get clipped here's the appended version:

“There are two ways to be fooled. One is to believe what isn't true; the other is to refuse to believe what is true.” -Søren Kierkegaard

Not an author of political fiction like someone just now posting OT here. :D

You didn't rain on anything; but perhaps you don't believe in color aliasing as well....

[Ed_Ingold]

The effect demonstrated by the focusing target would be better described as Moire effect.

[digitaldog]

The effect demonstrated by the focusing target would be better described as Moire effect.

By you. But that IS NOT how its describe by multiple authors of technical articles.

Nor here:

The Effect of the Color Filter Array Layout Choice on ... - MDPIhttps://www.mdpi.com › pdf

Or here:

Moiré, False Colour & Anti-Aliasing Filters

Or here:

http://www.foveon.com/files/Color_Alias_White_Paper_FinalHiRes.pdf

Or here:

Demosaicing - Wikipedia

Or here:

https://www.red.com/red-101/resolution-aliasing-motion-capture

Get the idea?

[babouphoto]

Will settle this by renting the A7Riv. It seems that there are benefits in my view but it will partially ately depend on how pixel shift handles grain and color aliasing.

Can't wait!

[Ed_Ingold]

If a focusing (or resolution) target should suddenly appear in the countryside, I will take care to avert my lens ;)

 

In practice, the only occasions I have found Moire or aliasing troublesome were with a relatively low resolution cameras without an AA filter - namely a Hasselblad with a 16 MP back. A 12 MP D3x exhibited no Moire on the same scene (corrugated steel grain silos). Revisiting the same scene more recently with an A7Rii (42 MP) produced no noticeable candy stripes. With high resolution, non AA sensors, objects which result in Moire tend to be small and relatively isolated. The most likely villains are fabrics, and only if in the plane of sharpest focus. I have not observed Moire in draped or folded fabric.

 

I see no signs of aliasing in either of my examples. Perhaps I'm not looking in the right place. If so, please enlighten me. I am using an iMac 5K display, which has pixels sized well below my threshold of vision.

[digitaldog]

I see no signs of aliasing in either of my examples. Perhaps I'm not looking in the right place. If so, please enlighten me. I am using an iMac 5K display, which has pixels sized well below my threshold of vision.

It's difficult to see an attribute you suggest doesn't exist. And at least two here see examples of this as presented so I'm not sure how to help.

Let's see the results of the OP's upcoming tests. If anything, we can agree to agree on his opinion of how he wishes to preceded.

[Ed_Ingold]

And at least two here see examples of this as presented so I'm not sure how to help.

Please copy an image and circle what you see as color aliasing.

[digitaldog]

Please copy an image and circle what you see as color aliasing.

You're requesting this from whom? Hopefully not me. Because in post #53, I provided an example of color aliasing from a site that provided and called a supplied image just that, and you suggested that isn't what you're seeing from that example. It was clear to them and me it was color aliasing. So I don't see how circling something is going to be of much help to you. Perhaps you're asking for visual guidance from the person who made post #51?

[bjorn_solberg1]

I regularly read about people who scan film with the Pentax K-1 and pixel shift, and they claim that pixel shift gives a better resulting scan than a non pixel shifted scan.

 

I have been following a thread over at "largeformat photography", where one member uses the Panasonic S1R and pixel shift to scan film professionally in his lab.

 

There are differences in the pixel shift technology between Pentax, Olympus, Phase One, Sony and Panasonic, but based on what I have seen on many photography forums I am convinced that pixel shift is beneficial when scanning film. Being able to bypass the Bayer filter and scan in full RGB is in my mind an obvious asset when scanning film. Then there is the notion of more resolution, less grain aliasing and a cleaner image resulting from a pixel shift scan.

 

I guess I will know for sure the day I purchase the Pentax K-1 mark II

[Ed_Ingold]

You're requesting this from whom? Hopefully not me. Because in post #53, I provided an example of color aliasing from a site that provided and called a supplied image just that, and you suggested that isn't what you're seeing from that example. It was clear to them and me it was color aliasing. So I don't see how circling something is going to be of much help to you. Perhaps you're asking for visual guidance from the person who made post #51?

I can see it in your example (the focusing target). What I'm asking is where do you see it in the examples I posted of slides scanned with and without pixel-shifting?

[rodeo_joe1]

This is a test, scanning a Kodachrome 25 slide with and without pixel-shifing, using a Sony A7Riii and Nikon 55/2.8 Macro lens. The upper panel is without.

The upper panel actually appears slightly sharper to me at a small scale. And I'm seeing no real difference in the rendering of the dye-cloud structure.

 

There's certainly no improvement in the macro detail of the balustrade.

 

BTW. I'm not a fan of the wasteful 2x green + blue + red of Bayer filtering. A true RGB triad array would seem to make more sense, and the geometry challenge isn't insurmountable. One example: Simply shifting alternate rows of RGB and BRG filters by a half-pixel would give triangles that can be sampled in multiple ways.

Edited December 2, 2019 by rodeo_joe|1

[Ed_Ingold]

In the slide scan, the results are limited by the medium (film), the lens, and (likely) hand-holding, although I see no signs of doubling typical of camera shake. Kodachrome has exceedingly fine grain. I underestimated the magnification in my previous post. It is way beyond the power of my 25x inspection microscope, probably over 100x. Secondly, in resampling in order to render individual pixels visible, the pixel boundaries were blurred due to cubic interpolation. In the Mt. St. Helens example, the finest details are only about one pixel wide, again probably blurred in resampling. The halos are probably due to default level sharpening, even if I didn't add anything deliberately.

 

I see some smudging of dark areas into lighter ones, which doesn't look like aliasing. It may be due to dye diffusion in processing. I don't have enough optical magnification to say one way or the other. The light area between balusters in the top panel is from a window on the opposite side of the dome.

 

The RGBG Bayer pattern is intended to reflect the relative sensitivity of the human eye. The green channel would have to be amplified in post, which would increase the noise compared to the present array.

 

It would be more than academic to see the effect of pixel-shifting on B&w film. B&W grain structure is actually dendritic rather than granular, with fine structure well into the realm of electron microscopy. What you see in a digital scan sre interpolated shadows rather than grain images. If color aliasing were to occur, the obvious solution would be to render the image in grey scale rather than RGB.

[babouphoto]

You should all go to the DP Review studio setup that I shared earlier on. It is in a controlled environment and much more telling. Compare the A7Riv to various cameras including the GFX100, Hasselblad and Phase One cameras.

[Ed_Ingold]

You should all go to the DP Review studio setup that I shared earlier on. It is in a controlled environment and much more telling. Compare the A7Riv to various cameras including the GFX100, Hasselblad and Phase One cameras.

Been there, done that.

 

The only clear advantage of pixel shifting is the resolution of high contrast resolution charts. I have yet to encounter one of those in the field. Pixel shifting has some advantage for landscapes and architecture, in which detail is virtually continuous well beyond the resolution of the camera and lens. It remains questionable whether it has any advantage when scanning film, in which nothing has a sharp edge..

[Ed_Ingold]

I repeated the experiments above, but resampled differently so that the pixel boundaries are retained. The straight image is on the top, the pixel-shiffted image is on the bottom. The halos are gone!

 

[digitaldog]

The halos are gone!]

Halos? :D

[andylynn]

I can see very small differences but I can’t say which is “better.” If I’m interpreting what you’ve said correctly, the magnification I’m seeing is equivalent to making a 5 foot wide print, so I can’t say you’d really be gaining anything by using this method. (But maybe an A7R iv has a better algorithm that would draw out more detail?...)

[Ed_Ingold]

Found one, lurking behind a bush ;) Top is straight, bottom is Pixel Shifted

 

Some chromatic aberration is present in diagonal lines. Since it is the same in both images, it is probably not Bayer aliasing. Judging from the uncertainty in the edges (# pixels from black to white), the pixel shifted version has roughly twice the resolution as the non-shifted version.

 

This target is only about 3"x5", suitable for video. There is ample evidence of aliasing in 4K video, but my living room isn't large enough to get the same effect at 42 MP with this lens.

 

 

Sony A7Riii + Sony 90/2.8 Macro @ f/5.0

[babouphoto]

Been there, done that.

 

The only clear advantage of pixel shifting is the resolution of high contrast resolution charts. I have yet to encounter one of those in the field. Pixel shifting has some advantage for landscapes and architecture, in which detail is virtually continuous well beyond the resolution of the camera and lens. It remains questionable whether it has any advantage when scanning film, in which nothing has a sharp edge..

 

Here are screenshots of the test focusing on the engravings. Sorry these are phone screenshots from the website since I don't have my laptop with me.

 

Unlesst the test was done improperly, this speaks for itself. Will still rent the camera and post comparisons later irrespective.