Luminous Landscape Forum
Equipment & Techniques => Medium Format / Film / Digital Backs – and Large Sensor Photography => Topic started by: ErikKaffehr on November 26, 2013, 12:31:30 am
-
Hi,
Someone suggested the image below shows spherochromatism, I don't think so but what is your view? Actual pixel dumps below.
(http://echophoto.dnsalias.net/ekr/Articles/MFDJourney/RawImages/AliasingSamples/20130804-CF043750.jpg)
Zeiss Distagon 40/4 FLE at f/8 on a P45+ (it is not the new IF version of the lens).
it goes away at f/16 - f/22:
(http://echophoto.dnsalias.net/ekr/Articles/MFDJourney/RawImages/AliasingSamples/20130804-CF043748.jpg)
Best regards
Erik
-
Someone suggested the image below shows spherochromatism, I don't think so but what is your view? Actual pixel dumps below.
Great, I am about to learn a new word...
What is "spherochromatism" and is it bad to have been taking pictures for 25 years without knowing the word... doctor? ;)
Cheers,
Bernard
-
Hi Bernard,
I am looking into aliasing from large pixels compared to small pixels. What I see in the image is a Bayer interpolation artefact, thin lines not covering all RGBG pixels. Luminosity is calculated correctly but color is not. That is my guess.
Some suggested that is something called spherochromatism and caused by a bad sample of the lens.
This is taken from the LensRental's site:
"There is one form of spherical aberration that you may have seen if you’ve used very wide aperture lenses: spherochromatism. Like chromatic aberration, some lenses have corrected spherical aberration at some wavelengths better than others. This may result in a magenta tinge for out of focus areas in the foreground and a greenish tinge in the out of focus portion of the background (or vice-versa)."
Best regards
Erik
Great, I am about to learn a new word...
What is "spherochromatism" and is it bad to have been taking pictures for 25 years without knowing the word... doctor? ;)
Cheers,
Bernard
-
I would think at that F16-22 range, it is diffraction that takes it away?
Would this be the same on another copy of the lens? Or perhaps a newer type of lens?
I'm not familiar enough with Ziess lenes to know if this is an older lens or most current. Visiting the Ziess website didn't help as I couldn't find this.
But I'm with Brenard, although in general he is far more technical than I, I too have not heard of this.
Is this what someone mentioned on the other thread being the Apo coating issue? Also sounds like a logical reason.
-
Hi,
The lens has significant chromatic aberration that is not fully corrected in Lightroom. I don't think this is chromatic aberration as it is on the pixel level, and follows distinct lines, vertically and horisontally. If it would be chromatic aberration I would say it would be more persistent. But I am asking for an opinons, so I don't argue that much.
Just to mention, there is nothing like 'Apochromatic coating', that term is true nonsense. Apochromatic means that lens focuses three colours in the same point. That can be achieved by a combination of three lenses with different indices of refraction. There are very few true apochromats. The Zeiss Otus may be one and there is the Coastal Optics 60 mm UVIR macro lens. Most lenses labeled Apo don't fulfil the requirement.
Zeiss is out of MF-business so it is clearly an older lens, known to be not very good. It has been replaced with a newer version called Distagon 40/4 FLE IF. The new IF version is a different animal. I was testing at f/16 - f/22 because I knew the lens had bad corners.
Now, just to put things in perspective, this lens makes great images:
Full size: http://echophoto.dnsalias.net/ekr/Articles/MFDJourney/RawImages/Samples3/20130907-CF044069.jpg
(http://echophoto.dnsalias.net/ekr/Articles/MFDJourney/RawImages/Samples3/20130907-CF044069.jpg)
Ful size: http://echophoto.dnsalias.net/ekr/Articles/MFDJourney/RawImages/Samples3/20130822-CF043962.jpg
(http://echophoto.dnsalias.net/ekr/Articles/MFDJourney/RawImages/Samples3/20130822-CF043962.jpg)
The Zeiss data sheet can be found here: http://www.hasselbladhistorical.eu/pdf/lds/CF40.pdf
I guess that the HC 3.5/35 is the closest one in the recent Hasselblad system: http://www.hasselblad.com/media/6640/hc35%20v3.pdf
I enclose the MTF plots.
I would think at that F16-22 range, it is diffraction that takes it away?
Would this be the same on another copy of the lens? Or perhaps a newer type of lens?
I'm not familiar enough with Ziess lenes to know if this is an older lens or most current. Visiting the Ziess website didn't help as I couldn't find this.
But I'm with Brenard, although in general he is far more technical than I, I too have not heard of this.
Is this what someone mentioned on the other thread being the Apo coating issue? Also sounds like a logical reason.
-
It cannot be spherochromatism as the entire line is at the same distance from the lens, yet the colors vary.
-
Someone suggested the image below shows spherochromatism, I don't think so but what is your view?
[...]
it goes away at f/16 - f/22:
Hi Erik,
No, it looks nothing like spherochromatism (spherical aberration that varies with wavelength, non-apochromatic correction) because that would result in resolution that varies with wavelength.
What this looks like is simple false color artifacting due to Bayer CFA demosaicing. Note the Magenta/Green and Blue/Yellow artifacting depending on horizontal/vertical orientation. It's a demosaicing artifact, a different Raw converter will probably give different results. The additional diffraction at f/16- f/22 will spread the signal over more sensels, allowing a better color interpolation/demosaicing. That's somewhat similar to the intended function of an OLPF.
Cheers,
Bart
-
This is taken from the LensRental's site:
"[...]This may result in a magenta tinge for out of focus areas in the foreground and a greenish tinge in the out of focus portion of the background (or vice-versa)."
This is more commonly referred to a Longitudinal Chromatic aberration.
Cheers,
Bart
-
Hi Bart,
That is my take, too, but there was a suggestion that it would something else, so I felt I would ask for opinions.
It seems a bit interesting that I see this but nobody else.
Best regards
Erik
Hi Erik,
No, it looks nothing like spherochromatism (spherical aberration that varies with wavelength, non-apochromatic correction) because that would result in resolution that varies with wavelength.
What this looks like is simple false color artifacting due to Bayer CFA demosaicing. Note the Magenta/Green and Blue/Yellow artifacting depending on horizontal/vertical orientation. It's a demosaicing artifact, a different Raw converter will probably give different results. The additional diffraction at f/16- f/22 will spread the signal over more sensels, allowing a better color interpolation/demosaicing. That's somewhat similar to the intended function of an OLPF.
Cheers,
Bart
-
It seems a bit interesting that I see this but nobody else.
Maybe they do see it, but camouflage it, and keep silent. A Raw converter like e.g. RawTherapee with AMAZE demosaicing offers user adjustable suppression of such false color artifacts. And when one anticipates such issues, one can always shoot a second image at a much narrower aperture, and use that in a layered (allows different deconvolution than optimal for the wider aperture image) masked edit to locally suppress the issues.
Cheers,
Bart
-
One of these days we might win you over to Capture One so you don't have to deal with (nearly as many) issues like this in the quality of your raw processing.
Maybe I need to suck it up and comp you a seat in our C1 class.
-
Hi,
The lens has significant chromatic aberration that is not fully corrected in Lightroom. I don't think this is chromatic aberration as it is on the pixel level, and follows distinct lines, vertically and horisontally. If it would be chromatic aberration I would say it would be more persistent. But I am asking for an opinons, so I don't argue that much.
Just to mention, there is nothing like 'Apochromatic coating', that term is true nonsense. Apochromatic means that lens focuses three colours in the same point. That can be achieved by a combination of three lenses with different indices of refraction. There are very few true apochromats. The Zeiss Otus may be one and there is the Coastal Optics 60 mm UVIR macro lens. Most lenses labeled Apo don't fulfil the requirement.
There is no such thing as an APO coating. That is a misunderstanding. Lets look at lens coating tech. Typically we are talking about anti-reflective coatings for photography. Some companies use coating chemicals that are highly tuned to specific wavelengths. They have to use a sequence of several to get the effect they want. Others use a broad band coating. The combination of coatings and lens curves are what makes the output of the lens what it is. The coating are an integral part of the lens design. The same lens elements without the coatings would look different, less contrasty. On a multi-coated lens obviously one coating is the most exposed so it will take the brunt of cleaning or anything else that could damage coatings. That will change the colors of the lens as the relative intensities of the colors change a small amount.
The lens element themselves can have color separation in the focus length direction or in the lateral direction. Lateral errors usually are pronounced at the edges.
Lightroom/ACR is the most common software. It has gone through many revisions. I would be surprised to hear it is causing very visible de-bayer errors at this stage. Do others also suffer this issue using manual PS brushing to remove it?
In my experience I have some kit zooms that I never use that have known color issues. Even some of my older primes have some in extreme light. I simply do not have this problem with my best lenses. I pretty much use RT for all RAW work. RT has one problem of being very slow to load thumbnails. I use ViewNX 2 to work the catalog then load RT from it with a right click. View NX2 does not have the color problem either.
-
ViewNX2 at 100%. No color separation. Some mild sharpening halos.
-
Maybe they do see it, but camouflage it, and keep silent. A Raw converter like e.g. RawTherapee with AMAZE demosaicing offers user adjustable suppression of such false color artifacts. And when one anticipates such issues, one can always shoot a second image at a much narrower aperture, and use that in a layered (allows different deconvolution than optimal for the wider aperture image) masked edit to locally suppress the issues.
Cheers,
Bart
Sorry Bart, I find that very unlikely. First of all if the software was having problems debayering, why would it only be on a small part of the image? It is not logical. Second of all if everyone was having to correct all their images they would have abandoned ship to other software long ago.
It is far more likely that it is an issue with the lens.
-
Sorry Bart, I find that very unlikely. First of all if the software was having problems debayering, why would it only be on a small part of the image?
Hi Arthur,
IMHO it's likely because it only shows on high contrast edges in the focus plane. Longitudinal Chromatic aberration produces colored edges in out of focus (OOF) areas also in the image center, often Green in front and Red in the rear of the focus plane. With a shallow DOF, due to relatively longer focal lengths in MFDB shots, or more diffraction blur due to narrower apertures, it does not manifest itself the same in all images.
Otherwise LCA is curable with reducing fringe colors on high contrast edges, false color artifacts have too many different colors for a simple cure in post-processing, so it is best handled when doing the Raw conversion/demosaicing.
It is not logical. Second of all if everyone was having to correct all their images they would have abandoned ship to other software long ago.
It is far more likely that it is an issue with the lens.
If Erik tries a different Raw converter and it goes away or change colors, that would prove false color artifacting, if it stays the same then it's something else. Not everybody uses a camera/back without AA-filter, so many would not be affected to the same degree.
Cheers,
Bart
-
It seems only Harry Potter gets to find the Download Now button.
FYI I have licenses for Elements 2, 3 and PS CS3. None are still on my computer.
-
Looks like some heavy supercolorfringielisticexposialidofism to me.
No way to save the image.
:P
-
Sorry Bart, I find that very unlikely. First of all if the software was having problems debayering, why would it only be on a small part of the image? It is not logical. Second of all if everyone was having to correct all their images they would have abandoned ship to other software long ago.
It is far more likely that it is an issue with the lens.
This is definitely debayering error. Promise.
-
Hi,
I checked a bit. The lightning conductor is about one pixel wide. The color shifts on the lightning conductor itself, but not on surrounding pixels. I have tested with:
- Lightroom 5.3 RC using PV 2012 (original sample)
- Lightroom 5.3 RC using PV 2003, this has much reduced color aliasing but it is still there
- Capture One 7.1.3 as bad or worse as 5.3 RC using PV 2012
Raw Therapie doesn't open IIQ file.
I am pretty sure this is color aliasing. I have a long experience of optics, having something like 25 lenses (Minolta, Sony, Pentax 67, Zeiss) I have never seen an aberration exactly one pixel wide in two orthogonal direction.
I have tested another of the images. It cleaned up well on LR 5.3RC with PV 2003, but was quite soft. On C1 it was pretty ugly. I have tested that image in RawTherapy and it was best of the bunch. C1 version below.
Best regards
Erik
Hi Arthur,
IMHO it's likely because it only shows on high contrast edges in the focus plane. Longitudinal Chromatic aberration produces colored edges in out of focus (OOF) areas also in the image center, often Green in front and Red in the rear of the focus plane. With a shallow DOF, due to relatively longer focal lengths in MFDB shots, or more diffraction blur due to narrower apertures, it does not manifest itself the same in all images.
Otherwise LCA is curable with reducing fringe colors on high contrast edges, false color artifacts have too many different colors for a simple cure in post-processing, so it is best handled when doing the Raw conversion/demosaicing.
If Erik tries a different Raw converter and it goes away or change colors, that would prove false color artifacting, if it stays the same then it's something else. Not everybody uses a camera/back without AA-filter, so many would not be affected to the same degree.
Cheers,
Bart
-
This is definitely debayering error. Promise.
+1
Spherochromatism would not cause a color change like presented.
It would most likely cause a color change becoming bigger towards the edge of the image.
The error presented changes with the slight position changes of the cable.
Thats a strong indicator for a debayering error and not an optical aberration.
-
Yeah, and it is as bad on Capture One, I promise.
I would add it is a very good argument to go for a high res MF back.
Best regards
Erik
This is definitely debayering error. Promise.
-
Hi,
I checked a bit. The lightning conductor is about one pixel wide. The color shifts on the lightning conductor itself, but not on surrounding pixels. I have tested with:
- Lightroom 5.3 RC using PV 2012 (original sample)
- Lightroom 5.3 RC using PV 2003, this has much reduced color aliasing but it is still there
- Capture One 7.1.3 as bad or worse as 5.3 RC using PV 2012
Raw Therapie doesn't open IIQ file.
I am pretty sure this is color aliasing. I have a long experience of optics, having something like 25 lenses (Minolta, Sony, Pentax 67, Zeiss) I have never seen an aberration exactly one pixel wide in two orthogonal direction.
I have tested another of the images. It cleaned up well on LR 5.3RC with PV 2003, but was quite soft. On C1 it was pretty ugly. I have tested that image in RawTherapy and it was best of the bunch. C1 version below.
Best regards
Erik
I am fairly sure I opened a IIQ sample you provided before of a grassy area with a tree line then mountain background in raw therapee. Maybe it is convert IIQ to DNG first.
-
Yes, IIQ is supported.
-
Hi,
Yes, I am aware of that. I prefer DNG but Capture One does not handle P45+ DNGs well.
I may ditch C1 altogether and go DNG. I see some advantages with C1, but DNG support for P45+ is bad. I am also negative on user interface, handling of white balance, default setting and HDR facilities.
But, I have been with LR since beta 3 in 2006, also before LR arrived I was considering to write a program which would be quite close tom LR in most aspects, data base based and parametric editing based on raw. So LR and me made friends immediately.
Best regards
Erik
I am fairly sure I opened a IIQ sample you provided before of a grassy area with a tree line then mountain background in raw therapee. Maybe it is convert IIQ to DNG first.
-
Mind sending me the original raw?
DNG??? Why? What problem are you solving?
-
Hi Doug,
I will post a link to the raw files in a couple of days. I try to post raws for most images I discuss, but have problems to keep up with my ambitions.
Just to make clear, I am much interested in imaging theory and I really have an engineering view on things. But I am also a serious amateur landscape photographer. Aliasing is a long time interest for me, going back to 2005 or so.
Regarding DNG, I am pretty sure that it is a recommended archival format (by UGRA? Cannot find the link), IIQ is not. I found that some of the images I discuss process best in RawTherapy and it won't read IIQ but works well for DNG. DNG is well documented, are IIQ details published? Who will support IIQ on Mac OS X.11 if Phase One goes defunct? Anders Torger recently developed a HDR tool, and he feels strong about DNG. If Lightroom supports DNG, like Raw Developer from Iridium and almost any other vendor except DxO, while cannot C1 do it?
I don't user ARW either. You sure IIQ will be around in 20 years?
Best regards
Erik
Mind sending me the original raw?
DNG??? Why? What problem are you solving?
-
Sorry Bart, I find that very unlikely. First of all if the software was having problems debayering, why would it only be on a small part of the image? It is not logical. Second of all if everyone was having to correct all their images they would have abandoned ship to other software long ago.
It is far more likely that it is an issue with the lens.
"Debayering" color moiré artifacts are, unfortunately, rather common with MF backs. Everybody who used these should know and especially fashion photographers who have extended problems with fabrics. So: no, it is not always a small part of the image and: no, even if it is annoying people are not abandoning the ship.
-
Hi Doug,
Most of the raws discussed are here: http://echophoto.dnsalias.net/ekr/Articles/Aliasing/RawSamples/
I also added an image of the feather shot with my Sony Alpha 99 SLT (all images with 150mm at f/8 or f/9 at 3.5 m). My main concern in the feather image is not colour artifacts, but:
- Cross hatch pattern at the lower end of the feather and strains bending the wrong way.
- Discontinuous strains.
- Images indicate that small pixels are good.
I normally post raw images, full scale JPEGS and miniatures for easy navigation, but I don't have time and energy right now.
I know that there is a moiré brush in both LR and Capture One, but it has side effects, so it needs to be applied with care. Removing colour artefacts still leaves monochrome artefacts, like non continuos waves and broken strains on feather.
Raw Therapee, Amaze and false color supression (1-2 steps) gives best results.
Best regards
Erik
Mind sending me the original raw?
DNG??? Why? What problem are you solving?
-
Regarding DNG, I am pretty sure that it is a recommended archival format (by UGRA? Cannot find the link)
UGRA certainly not!
You sure IIQ will be around in 20 years?
Are you sure DNG will be around in 20 years?
The only future proof image file format is TIF... AFAIK.
Then again... as long as you store a current version of DNG Converter in your long time archive you can just as well use IIQ today and convert to DNG in the future if need is. You can also store a current computer with a current OS and a current RAW software.
Personally I store "flat" developped TIFs (full histogram preserved, 16bit, P1 camera profile assigned, no sharpening applied) from my RAWs in my long time archive (along with the respective RAW files). In addition I also store a current computer with current OS and the respective software to be able to open my RAW files in the future.
-
Hi,
Pretty sure that DNG will be around for 20 years, as the specification is open. DNG is a TIFF format by the way. But, TIFF is not raw. The data is demosaiced and converted to a different working space and gamma.
You may be aware that programs that work today may not work on a new generation of hardware, Hardware moves forward and software goes with it.
But, I agree that TIFF is safe. On the other hand, DNG can contain original raw. DNG is also small, about same size the raw file. Converting to TIFF increases image size about up to 6 times, so you have 4-6 times the data with less information.
The document I was referring to was not UGRA but UPDIG, sorry! It is here: http://www.updig.org/pdfs/updig_photographers_guidelines_v40.pdf
A shot discussion of raw formats is here: http://www.updig.org/guidelines/ph_file_formats.html
Best regards
Erik
UGRA certainly not!
Are you sure DNG will be around in 20 years?
The only future proof image file format is TIF... AFAIK.
Then again... as long as you store a current version of DNG Converter in your long time archive you can just as well use IIQ today and convert to DNG in the future if need is. You can also store a current computer with a current OS and a current RAW software.
Personally I store "flat" developped TIFs (full histogram preserved, 16bit, P1 camera profile assigned, no sharpening applied) from my RAWs in my long time archive (along with the respective RAW files). In addition I also store a current computer with current OS and the respective software to be able to open my RAW files in the future.
-
Regarding DNG, I am pretty sure that it is a recommended archival format (by UGRA? Cannot find the link)
format is one thing, software that converts to DNG is another thing... there were enough bugs and changes in Adobe DNG converter along the way in terms what data read off sensors were preserved and what data were not...
-
Hi,
The way I am set up I use DNG as the archival format, but it contains the original raw image. On the Phase One I use IIQ, as I want to have the option to use C1. Converting to DNG, I could still extract the IIQ file, of course but it is not really convenient.
I would say that I regard DNG support to be mandatory. Interestingly enough, Capture One handles DNGs for P45+, but does it badly while it works fine with DNGs coming from Sony ARW.
Best regards
Erik
format is one thing, software that converts to DNG is another thing... there were enough bugs and changes in Adobe DNG converter along the way in terms what data read off sensors were preserved and what data were not...
-
Thanks,
Last time I used it I couldn't find the images I was looking for, so I was jumping conclusions, sorry. Thanks enlightening!
Best regards
Erik
Yes, IIQ is supported.
-
Hi Erik,
It is simply colour aliasing, due to imaging a <= 1 pixel wide high-contrast object, onto a sensor without an AA filter.
You should be reassured really: seeing this effect confirms that you are in perfect focus and your lens is really sharp at the f-stop used! :D
It is unquestionably NOT due to any sort of aberration - chromatic or spherochromatic. Aberrations throw light out of focus, and spread it over multiple pixels.
I would also hesitate to call it a "debayering error". One cannot expect a deBayering algorithm to perform well on incomplete data, which is what we really have here, as I'll explain.
I get this effect also with my 9-micron pixel DB, in the right circumstances. I am glad to show you a sample:
(http://img854.imageshack.us/img854/8561/lakd.jpg)
This a 100% crop. No sharpening, no denoising.
ISO 400, 80/1.9 at f2.8, 60 seconds. Central area of Orion, for those who are interested.
This shows that my Mamiya 80/1.9 is already a critically sharp lens at f2.8. As each star image drifted over the sensor, virtually all of its light was focussed into a single pixel in the RGBG Bayer matrix. Most of the motion is horizontally left -> right, so each star spent a good while going R-G-R-G-R-G-R-G-R-G, or B-G-B-G-B-G-B-G-B-G - i.e. its focused PSF was so sharp that it was confined to one row of pixels and it missed one Bayer pixel colour entirely. (With information from the star only being recorded on 2 of the 4 Bayer matrix group pixels, this is why I don't regard it as deBayering "error"..."artefact" would be a fairer term). Hence, it shows as a short trail with either a cyan or a yellow colour bias.
But there is also a slight upward drift as the sky turned, which meant that every so often, the star transitioned from being mainly on RG pixels, to split more equally across RGBG pixels, and then mainly on BG pixels. Thus, each trail alternates between yellow, briefly whiteish, cyan, briefly whiteish, yellow again...
Reprocessing this with moire reduction goes a long way to removing the aliasing. Although to be honest I don't really dislike the effect; it adds interest in my view, emphasising the different underlying colours of the stars, and only looking a bit odd with one's nose up pressed against the print.
There is some longitudinal chromatic aberration present too - the blue halos around the very bright, overexposed stars. Note how the chromatic halos are several pixels wide, and they do not change colour with star drift.
Ray
-
That is a convincing argument Ray.
-
Hi Ray,
Thanks for making that clear. I was pretty sure it was colour aliasing, but I am open to other explanations, that is the reason I posted the question.
Best regards
Erik
Hi Erik,
It is simply colour aliasing, due to imaging a <= 1 pixel wide high-contrast object, onto a sensor without an AA filter.
You should be reassured really: seeing this effect confirms that you are in perfect focus and your lens is really sharp at the f-stop used! :D
It is unquestionably NOT due to any sort of aberration - chromatic or spherochromatic. Aberrations throw light out of focus, and spread it over multiple pixels.
I would also hesitate to call it a "debayering error". One cannot expect a deBayering algorithm to perform well on incomplete data, which is what we really have here, as I'll explain.
I get this effect also with my 9-micron pixel DB, in the right circumstances. I am glad to show you a sample:
(http://img854.imageshack.us/img854/8561/lakd.jpg)
This a 100% crop. No sharpening, no denoising.
ISO 400, 80/1.9 at f2.8, 60 seconds. Central area of Orion, for those who are interested.
This shows that my Mamiya 80/1.9 is already a critically sharp lens at f2.8. As each star image drifted over the sensor, virtually all of its light was focussed into a single pixel in the RGBG Bayer matrix. Most of the motion is horizontally left -> right, so each star spent a good while going R-G-R-G-R-G-R-G-R-G, or B-G-B-G-B-G-B-G-B-G - i.e. its focused PSF was so sharp that it was confined to one row of pixels and it missed one Bayer pixel colour entirely. (With information from the star only being recorded on 2 of the 4 Bayer matrix group pixels, this is why I don't regard it as deBayering "error"..."artefact" would be a fairer term). Hence, it shows as a short trail with either a cyan or a yellow colour bias.
But there is also a slight upward drift as the sky turned, which meant that every so often, the star transitioned from being mainly on RG pixels, to split more equally across RGBG pixels, and then mainly on BG pixels. Thus, each trail alternates between yellow, briefly whiteish, cyan, briefly whiteish, yellow again...
Reprocessing this with moire reduction goes a long way to removing the aliasing. Although to be honest I don't really dislike the effect; it adds interest in my view, emphasising the different underlying colours of the stars, and only looking a bit odd with one's nose up pressed against the print.
There is some longitudinal chromatic aberration present too - the blue halos around the very bright, overexposed stars. Note how the chromatic halos are several pixels wide, and they do not change colour with star drift.
Ray