Towards an explanation of the Italian Flag Phenomenon

[Lindolfi]

Advertisement (gone after registration)

Again thanks very much for the contributions. The birefringence option posed by perb is certainly something to consider. But if it would occur due to strain, you would expect wilder patterns than the smooth function of the discolouration with respect to the distance to the center of the sensor that we see. Anyway, I can see what a polarizing filter does with the effect as you suggest, perb.

 

The same holds for the suggestion posed by Pop: using strong monochromatic filters to see how the intensity distribution works, but I expect the RGB curves that we already get from the white light experiments.

 

There is an other possibility on which I am working:

 

1] From the kodak site we know that the microlenses are arranged so that each micro lens covers a single colour sensitive pixel (red or green or blue) and not cover a cell of four colour pixels. ( Kodak: Plugged In - On Biological and Electronic Blind Spots )

 

2] These microlenses are not achromats, but consist of one refractive material and so exhibit chromatic aberration. In the Kodak photographs of them they have an enormous front curvature and thickness compared to their width, so a considerable amount of colour dispersion can be expected on a very small area.

 

3] That would mean that at increasing obliqueness of the light rays, more and more red component is lost, since that is the component that is bent the least by the microlenses as shown in this figure that I have drawn:

 

 

 

This is all fine and it does explain cyan discolouration (lack of red) at increasing distance from the center of the sensor, but it does not explain the magenta discolouration.

 

So the explanation does not hold water yet, but it is another principle to consider (the chromatic aberration of the micro lenses combined with the Bayer pattern).

 

To be able to exclude this, I need an Italian Flag Phenomenon image coming from a sensor without micro lenses. Anyone?

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[otto.f]

Again thanks very much for the contributions. The birefringence option posed by perb is certainly something to consider. But if it would occur due to strain, you would expect wilder patterns than the smooth function of the discolouration with respect to the distance to the center of the sensor that we see. Anyway, I can see what a polarizing filter does with the effect as you suggest, perb.

 

The same holds for the suggestion posed by Pop: using strong monochromatic filters to see how the intensity distribution works, but I expect the RGB curves that we already get from the white light experiments.

 

There is an other possibility on which I am working:

 

1] From the kodak site we know that the microlenses are arranged so that each micro lens covers a single colour sensitive pixel (red or green or blue) and not cover a cell of four colour pixels. ( Kodak: Plugged In - On Biological and Electronic Blind Spots )

 

2] These microlenses are not achromats, but consist of one refractive material and so exhibit chromatic aberration. In the Kodak photographs of them they have an enormous front curvature and thickness compared to their width, so a considerable amount of colour dispersion can be expected on a very small area.

 

3] That would mean that at increasing obliqueness of the light rays, more and more red component is lost, since that is the component that is bent the least by the microlenses as shown in this figure that I have drawn:

 

 

 

This is all fine and it does explain cyan discolouration (lack of red) at increasing distance from the center of the sensor, but it does not explain the magenta discolouration.

 

So the explanation does not hold water yet, but it is another principle to consider (the chromatic aberration of the micro lenses combined with the Bayer pattern).

 

To be able to exclude this, I need an Italian Flag Phenomenon image coming from a sensor without micro lenses. Anyone?

 

Maybe a stupid idea but could this happen in combination with the speed of reading out the analogue signal on the CCD-sensor in relation to it's capacity to retain the signal before it has been read out and translated into a digital color-pixel? This reading out is the only asymmetrical thing I can imagine, a delay could cause a difference in relation to what you just said here Lindolfi

[wlaidlaw]

I think folks suffering Italian Flag will be pretty happy when the production version of firmware currently in beta test with various members of this forum is released. I would suggest waiting until this happens. 1.138 FW is getting close to yesterday's news. No idea on timetable but don't think it can be long. Leica is working hard for you.

 

Wilson

[pop]

The following is just an aid in visualization. I took - again - the first example from post #1. I separated it into the three colour channels. I then used the Gimp's colour curve tool to make the distribution of the brightness over the whole area of the picture visible, much as a relief map might show the hills in a terrain.

 

The blue channel:

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

 

The green one:

 

The red one:

 

Both the blue and the green one roughly exhibit the distribution one would expect: both mirror and rotational symmetry. There is, however, a "disturbance" reaching from the center of the lower edge towards the center of the image.

 

The red one shows the same disturbance much stronger. It also shows a darker border in the region of the upper left corner.

 

Even though it's only a visualization, the red channel reminds me of images I've seen of sheets of transparent material under tension as seen in polarized light.

 

Presumably, this could be made even more easily readable by compensating for the "normal" vignette of the lens.

[pop]

I described this earlier here: http://www.l-camera-forum.com/leica-forum/leica-m9-forum/153831-m9-red-edge-redux-part-ii-2.html#post1540181

 

I've applied the same processing to your sample:

 

 

Blue:

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

 

Green:

 

Red:

 

The "disturbances" seem much milder here.

 

If the phenomenon was indeed related to the distribution of the tension over the sensor, one of the logical next questions could be if there was any kind of correlation between the phenomenon and the sensors which cracked in some instances of this camera. (Grasping at straws, I know).

[lars_bergquist]

There is an other possibility on which I am working:

 

1] From the kodak site we know that the microlenses are arranged so that each micro lens covers a single colour sensitive pixel (red or green or blue) and not cover a cell of four colour pixels. ( Kodak: Plugged In - On Biological and Electronic Blind Spots )

 

2] These microlenses are not achromats, but consist of one refractive material and so exhibit chromatic aberration. In the Kodak photographs of them they have an enormous front curvature and thickness compared to their width, so a considerable amount of colour dispersion can be expected on a very small area.

 

3] That would mean that at increasing obliqueness of the light rays, more and more red component is lost, since that is the component that is bent the least by the microlenses as shown in this figure that I have drawn:

 

 

Colour dispersion in the microlenses was indeed one of the possible explanations I considered. But this effect too would have been symmetrical, so it won't do.

 

The red old man

[UliWer]

Advertisement (gone after registration)

...To be able to exclude this, I need an Italian Flag Phenomenon image coming from a sensor without micro lenses. Anyone?

 

Not exactly "italian flag" - though this example (link in my original postings and see ho_co's comment on it in #44) - from a Canon5d with a very unusual lens is surprising:

 

http://www.l-camera-forum.com/leica-forum/leica-m9-forum/161677-red-edge-redux-part-iii-2.html#post1585005

[Lindolfi]

Thanks Pop for the density plots of the channels. Again it is a combination of vignetting and discolouration and the downward "stream" in the red channel would also be visible in the bottom left corner if you would correct for vignetting.

 

Thanks UliWer for the example with the 5D. Here I don't like the test conditions in which there are colours (like the brown table) close to the white sheet of paper that is photographed. A white plastic opaque disk direct in front of the lens, illuminated with evenwhite light is much better.

[UliWer]

....If the phenomenon was indeed related to the distribution of the tension over the sensor, one of the logical next questions could be if there was any kind of correlation between the phenomenon and the sensors which cracked in some instances of this camera. (Grasping at straws, I know).

 

I don't think so: my M9 is from August 2010, when Leica said that the sensor-crack problem was solved - and I have the red edges sometimes with some lenses. I had them as well with the same lenses and the M8, where any tension on the filter above the sensor would certainly have cracked it - and the sensor is well intact.

[adan]

As to the question of whether this shows up on CMOS sensors - it did on the Kodak Pro 14n SLRs, which used Fill Factory (not Kodak) CMOS sensors.

 

Example posted by a Kodak 14n Pro user on Photo.net: http://hannu.mallat.fi/scratch/italian-flag.jpg

 

Another: Re: What exactky is Italian flag syndrome: Kodak SLR Talk Forum: Digital Photography Review

 

However - the Kodak 14n's also did NOT use an AA filter, so if we are talking about effects possibly related to neighboring pixels under the Bayer pattern, the lack of an AA filter is a common thread between the Kodaks and the M9.

[Lindolfi]

OK Adan, thanks and Kodak claims there are no micro lenses in the 14n Pro: click

 

Presumably they did use some IR filter, so perhaps we should now look in the direction of the relation between the IR filter and the sensor without AA filter.

[mjh]

Presumably they did use some IR filter, so perhaps we should now look in the direction of the relation between the IR filter and the sensor without AA filter.

I believe the 14n had an absorption filter, just like the M8 and M9. (I’m positive that its successor, the SLR/n, had an absorption filter.)

[ho_co]

Does it occur on images made by CMOS sensors?

Not exactly "italian flag" - though this example (link in my original postings and see ho_co's comment on it in #44) - from a Canon5d with a very unusual lens is surprising:

http://www.l-camera-forum.com/leica-forum/leica-m9-forum/161677-red-edge-redux-part-iii-2.html#post1585005

 

Thanks for mentioning that image, Uli. It responds to Otto's question as well.

 

One thing to add--it hadn't occurred to me to mention it earlier--is that this particular 16464 focusing mount is a bit loose and in need of re-lubrication. There may be a bit of wobble in all my images with the 65 Elmar; that brings us back to Andy's "decentration" hypothesis that I had previously scoffed at.

[lars_bergquist]

As to the question of whether this shows up on CMOS sensors - it did on the Kodak Pro 14n SLRs, which used Fill Factory (not Kodak) CMOS sensors.

[ ... ]

 

However - the Kodak 14n's also did NOT use an AA filter, so if we are talking about effects possibly related to neighboring pixels under the Bayer pattern, the lack of an AA filter is a common thread between the Kodaks and the M9.

 

That is indeed a thought that struck me too. That lead should be investigated.

 

The bleary-eyed old man

[otto.f]

That is indeed a thought that struck me too. That lead should be investigated.

 

The bleary-eyed old man

 

Would that solve the asymmetrical question?

[giordano]

I'm just wondering (and please forgive me if this point has already been dealt with and I missed it). Given

(a) chromatic aberration in the microlenses,

( "shifted" microlenses (i.e. the pitch of the microlens array varies across the sensor while the pitch of the photosite array does not) and

© the rotational asymmetry of the Bayer array:

could the phenomenon be due to errors of a few nanometres in positioning the microlens layer when assembling the sensor pack?

[Lindolfi]

Yes giordano, it could if the whole layer of microlenses is shifted along the diagonal (left bottom to top right) by a fraction of the cross section of a single pixel relative to the pixel layer, given the principle of the loss of certain wavelength contributions as I have shown in my figure. I was already working on some modelling in that direction. A few nanometer is not enough, but I think the principle does work. Now, if the principle can be shown to work (and I think it does) and we have some information on the precision of the alignment of the microlens layer and the pixel array, we have an explanation, that even may explain the variation between M9 models.

[Erik Gunst Lund]

I agree that the Bayer pattern is the most likely cause, and given the steep angle of the photons hitting the well wall inside the sensor sites could it be the photons from other colors spilling into adjacent wells more towards the outer edges, than the center...

[Advertisement]

Hallo,

Look here: Leica M

[UliWer]

Would that solve the asymmetrical question?

 

The asymmetrie and especially the irregularity of the phenomenon are both strong arguments against any supposition that the reason is to be looked for in the cameras or lenses alone.

 

As we can say that the phenomenon is neither always there with the same cameras and lenses, nor always absent, nor does it show a constant pattern or degree of disturbance there must be some important external factor.

[lars_bergquist]

Would that solve the asymmetrical question?

 

No. But it would perhaps narrow the field a bit. Many mysteries go unsolved because we try to solve he wrong mystery.

 

The red-eyed old man