Removing color casts from wide angle lenses

[Paul Ozzello]

Some of my wide angle lenses - Schneider Super Angulon 72mm, Rodenstock Apo grandagon 55 have magenta color casts most noticeable in the corners. Since I scan most of my film what is the best way to remove this color cast in Photoshop or Lightroom? I notice medium format digital users resort to LCC profiles to process their images but this can't be done with film.

 

[MattKing]

I think this is a digital image processing question, so I'm going to move it into a sub-forum appropriate for that.

 

[Alan Edward Klein]

Is this being caused by the falloff on wide angle lenses corrected by center filters?

 

[Paul Ozzello]

Is this being caused by the falloff on wide angle lenses corrected by center filters?

I forgot to mention that I use the appropriate center filters for each lens. The problem isn't exposire falloff but color casts (mostly magenta)

 

[Light Capture]

Latest lightroom has quite easy fix for this. In develop module, under Histogram there is Masking tool (dotted circle on right).
Each corner can be masked separately if color cast is different due to light angle and it probably is. Next step is to select brush and size and feathering can be adjusted.
When selection is made mask is created and adjustments can be performed selectively only on this mask. Color temperature and other color controls are available to neutralize the cast.

Same can be done in Photoshop. Adjustment layer needs to be added and selections added as desired.
These are the easiest fixes. Equivalent to LCC profile could be made but it's only effective on digital since you need to get it for each aperture and light could also change the color cast.

I have some similar lenses but haven't used them extensively. Curious to find out when it happens. I did see something similar in certain situations with some Hasselblad and large format lenses.
Does this happen on all images taken with these or just in certain situations / similar light?
From what I read there is natural polarization of light and it can be seen on images with wide lenses. Polarizers will make it worse.
Also if some other filters are used, for example UV or ND, edges could show any slight color cast in those filters.
Some filters are advertised to minimize color cast. If you used any filters, that might be potential culprit.

 

[Sirius Glass]

What you are seeing on the lens is the coatings which should not change the light color. So that is not the source of the color you are concerned about. However all lenses have a light falloff as it moves the center line of 1/cos4(angle) and it is more pronounced on very wide angle lenses and that may influence the overall color cast whatever the source.

 

[Alan Edward Klein]

I forgot to mention that I use the appropriate center filters for each lens. The problem isn't exposire falloff but color casts (mostly magenta)

Which model CF are you using? Maybe there's a mismatch? Wouldn't you be better off trying to track down the source of the problem rather than trying to correct it in post?

 

[Paul Ozzello]

The route I would explore is to shoot a reference white card with those lenses to isolate the color cast

I think that's similar to what people are doing with medium format digital, creating LLC for extreme wide angle lenses and applying them in Capture 1. although I think instead of a white card they put a piece of translucent frosted material over the lens. I was hoping that one of those sliders in Adobe Raw for lens corrections could do something similar. On my next roll I'll try shooting a frame like you suggested.

What you are seeing on the lens is the coatings which should not change the light color. So that is not the source of the color you are concerned about. However all lenses have a light falloff as it moves the center line of 1/cos4(angle) and it is more pronounced on very wide angle lenses and that may influence the overall color cast whatever the source.

.

Which model CF are you using?

I use a Scheider IVb for the 72 XL - It compensates properly for the light fall off but I think it may contribute to the color cast.

 

[Alan Edward Klein]

Are you using the right center filter?
6x9 lens by Dan Fromm (1)
Schneider center filters

 

[Paul Ozzello]

Are you using the right center filter?
6x9 lens by Dan Fromm (1)
Schneider center filters

I couldn't open your links but yes I am using the right center filters.

For the 72mm Schneider I'm using a IVb (4b). For the 90mm SA XL I'm using a IVa and don't have any color cast problems.

 

[itsdoable]

I'd like to know what is the cause of colour cast in the corners with Film. I understand that in digital, it is due to the cross-talk between the wells and colour filter array. But what is the cause on Film?

I do remember corner vignetting often resulted in a colour saturation change, and that may also cause some colour drift due to the different sensitivity of the 3 layers, but a center filter was used here.

 

[Sirius Glass]

I'd like to know what is the cause of colour cast in the corners with Film. I understand that in digital, it is due to the cross-talk between the wells and colour filter array. But what is the cause on Film?

I do remember corner vignetting often resulted in a colour saturation change, and that may also cause some colour drift due to the different sensitivity of the 3 layers, but a center filter was used here.

See post# 7 about the 1/cos4(angle) falloff which explains the greater density in the corners.

 

[Sirius Glass]

@Sirius Glass the question wasn't about vignetting but about color casts.

It is an overall caste that you have to deal with.

 

[MattKing]

Some of my wide angle lenses ... have magenta color casts most noticeable in the corners

To my mind, this means that either:
1) the transparencies themselves have an overall magenta cast, which is barely visible in the lighter centre parts and most noticeable in the (darker) corners, and which in turn is reflected in the scans; or
2) the transparencies themselves exhibit uneven colours, which in turn is reflected in the scans; or
3) the scans themselves are uneven in their colour.
If #1 is the case, the solution is colour correcting filters on the camera. Alternatively, an overall correction to the colour of the scan may suffice.
If #2 and/or #3 are the case, the OP needs a colour correction that is applied in different amounts at different parts of the image - that is where the LLC tools come into play.

 

[Light Capture]

To my mind, this means that either:
1) the transparencies themselves have an overall magenta cast, which is barely visible in the lighter centre parts and most noticeable in the (darker) corners, and which in turn is reflected in the scans; or
2) the transparencies themselves exhibit uneven colours, which in turn is reflected in the scans; or
3) the scans themselves are uneven in their colour.
If #1 is the case, the solution is colour correcting filters on the camera. Alternatively, an overall correction to the colour of the scan may suffice.
If #2 and/or #3 are the case, the OP needs a colour correction that is applied in different amounts at different parts of the image - that is where the LLC tools come into play.

This is a good starting point.
I still think that suggestion from my previous post about colour cast coming from filter is a further possibility to check.
Even Schneider's center filters could potentially produce colour casts compared to colour reproduction of the lens itself.
Since it's visible in corners only there are 3 possibilities:
1. both center filters for these two lenses are 1.5 stops. So there is differential in F stops and if center filter has a colour cast due to aging or inherent cast, 1.5 stops is just enough to cause visible colour cast.
2. there is still possibility of light polarization across image that manifests differently in far corners

I didn't test it myself but making LCC profiles on film might not be feasible. Phase One is recommending to create LCC profile during every shooting session and for every aperture. Not sure if this is nitpicking or related to properties of their digital sensors in different light but film could certainly respond differently since it's not linear. Correction lens profiles can be created for Lightroom. There was a long procedure on how to do that and how to make calibration targets. Obviously there are quite few things that could go wrong when trying to do this with film. Shutters would have to be linear between f-stops and development would have to be perfect to get consistent results. All this assuming that your center filters don't have colour casts.

Both of these center filters are fixed 1.5 stops and wide open there is probably more than 3 stops of fall-off. That differential gets smaller at smaller apertures and corner density and colour casts will have different strength.

Fastest solution would be testing that filter on digital camera preferably on equivalent focal length for the format and lens with similar filter diameter. It can be just held in front of the lens.

 

[Paul Ozzello]

2) the transparencies themselves exhibit uneven colours, which in turn is reflected in the scans; or

#2. I don't think it's the scanner as it only happens with the super wide lenses. I tried CornerFix but it only works with images created from a Bayer sensor. I'm scanning 617 on a Coolscan 9000 but I have to do it in three sections using offsets (it only scans a maximum of 6x9), but it's the final stitched image that has the color cast in the corners - not the corners of the 3 individual scans. Even when I do it in one pass on an Epson flatbed I have the same problem.

1. both center filters for these two lenses are 1.5 stops. So there is differential in F stops and if center filter has a colour cast due to aging or inherent cast, 1.5 stops is just enough to cause visible colour cast.

.

The IVb filter is a whole 2 stops, probably why it's worst

Correction lens profiles can be created for Lightroom

I checked out Lightroom and there is a flat-field correction option that works with any file type. I don't have a correction frame yet but I played around by creating one in photoshop using a circular magenta gradient and it looks like it should work. I'll try to shoot some LCC over the weekend and try it out .

 

[Light Capture]

#2. I don't think it's the scanner as it only happens with the super wide lenses. I tried CornerFix but it only works with images created from a Bayer sensor. I'm scanning 617 on a Coolscan 9000 but I have to do it in three sections using offsets (it only scans a maximum of 6x9), but it's the final stitched image that has the color cast in the corners - not the corners of the 3 individual scans. Even when I do it in one pass on an Epson flatbed I have the same problem.

.

The IVb filter is a whole 2 stops, probably why it's worst



I checked out Lightroom and there is a flat-field correction option that works with any file type. I don't have a correction frame yet but I played around by creating one in photoshop using a circular magenta gradient and it looks like it should work. I'll try to shoot some LCC over the weekend and try it out .

If you prefer to use gradient. Masking option in Lightroom can be also set to gradient. I find Lightroom way faster to work with images and only get them to Photoshop if I really have to.
Masking is basically separate layer just like in Photoshop and isn't destructive.
Quickest workaround here is probably:
1. Acquire LCC image of white background (at multiple f stops if needed).
2. Import in lightroom and add gradient mask
3. Settings or mask itself can be copied to any images you have. If you find any images that need further correction, every single one can be fine tuned for that masked area only.

To adjust gradient to your liking when mouse is hoovered over white area in the center, RGB values will show bellow histogram in develop mode.
They you can adjust gradient to get the same RGB values in corners.
If you're not using RAW dng scanning through Vuescan, I would recommend trying out Vuescan for this. It can be trialed with watermarked output.
dng output makes big difference for editing in Lightroom. Size is also smaller than TIFF.

 

[Alan Edward Klein]

If you prefer to use gradient. Masking option in Lightroom can be also set to gradient. I find Lightroom way faster to work with images and only get them to Photoshop if I really have to.
Masking is basically separate layer just like in Photoshop and isn't destructive.
Quickest workaround here is probably:
1. Acquire LCC image of white background (at multiple f stops if needed).
2. Import in lightroom and add gradient mask
3. Settings or mask itself can be copied to any images you have. If you find any images that need further correction, every single one can be fine tuned for that masked area only.

To adjust gradient to your liking when mouse is hoovered over white area in the center, RGB values will show bellow histogram in develop mode.
They you can adjust gradient to get the same RGB values in corners.
If you're not using RAW dng scanning through Vuescan, I would recommend trying out Vuescan for this. It can be trialed with watermarked output.
dng output makes big difference for editing in Lightroom. Size is also smaller than TIFF.

The falloff in wide angle lenses is based on some calculation using the cosine, which frankly I don't understand. However, wouldn't the gradient of the correction have to be able to match the cosine falloff exactly?> Otherwise, you'll never be able to match it and it will be obvious.

 

[Light Capture]

The falloff in wide angle lenses is based on some calculation using the cosine, which frankly I don't understand. However, wouldn't the gradient of the correction have to be able to match the cosine falloff exactly?> Otherwise, you'll never be able to match it and it will be obvious.

Looking at various lens datasheets fall off seem to be quite predictable. Didn't really look at any calculations.
Gradients can be adjusted in different ways. Lightroom is simpler. Photoshop has more ways to adjust it including scripts and batches that can be created for the task.
However, it looks good to my eye when adjusted like this. Lightroom lens correction tool has two sliders for intensity of correction and gradient linearity. This resolved all issues I had with fall off.

Lightroom masking is really powerful and there's little need for me to get outside of Lightroom. Masking can be applied in different ways and it's plenty for almost any adjustments.

And then when gradient (or any other one of these masks) is selected, it can be adjusted with different controls:

Contrast, Texture, Clarity and Dehaze can even repair flare if it's not to far gone. Selective sharpening or blurring, selective Moire reduction etc.
Only if I run out of options, or it doesn't look right, Photoshop is there to help since it works significantly better but requires much more involved editing.

 

[Sirius Glass]

What you are seeing on the lens is the coatings which should not change the light color. So that is not the source of the color you are concerned about. However all lenses have a light falloff as it moves the center line of 1/cos4(angle) and it is more pronounced on very wide angle lenses and that may influence the overall color cast whatever the source.

See post# 7 about the 1/cos4(angle) falloff which explains the greater density in the corners.

#2. I don't think it's the scanner as it only happens with the super wide lenses. I tried CornerFix but it only works with images created from a Bayer sensor. I'm scanning 617 on a Coolscan 9000 but I have to do it in three sections using offsets (it only scans a maximum of 6x9), but it's the final stitched image that has the color cast in the corners - not the corners of the 3 individual scans. Even when I do it in one pass on an Epson flatbed I have the same problem.

.

The IVb filter is a whole 2 stops, probably why it's worst



I checked out Lightroom and there is a flat-field correction option that works with any file type. I don't have a correction frame yet but I played around by creating one in photoshop using a circular magenta gradient and it looks like it should work. I'll try to shoot some LCC over the weekend and try it out .

The falloff in wide angle lenses is based on some calculation using the cosine, which frankly I don't understand. However, wouldn't the gradient of the correction have to be able to match the cosine falloff exactly?> Otherwise, you'll never be able to match it and it will be obvious.

The problem is with the filter on a very wide lens and the 1/cos4 is killing you. Be careful with dark filter or especially a polarizer filter on a very wide angle lens.

 

[Sirius Glass]

Here's my "reverse-engineering" of Sirius formula:

Light decreases with distance (inversely to its square i.e. 1/distance^2). The same distant object becomes brighter as you bring it closer. Now visualize a triangle with the lens at one tip, "looking" at the opposite side which is the flat focus plane. Obviously the distance to the center of the plane (adjacent) is shorter than to the edge (hypotenuse). Hopefully this brings back middle school trigonometry lessons on how to express the relationship between the two. Then you combine the trigonometry with 1/distance^2. The wider the lens, the greater the distance to edge objects, the less light you get at the edges.

BTW I suspect that in a real world lens the fall off is mostly caused by the engineering trade-offs (image circle vs weight?) by lens designers. The new Voigtlander Classic line shows a lot of vignetting and at the same time they're suspiciously compact, so I am quite unsure if the formula cited by Sirius is actually "visible" in our photos.

It is noticeable in photographs as I learned from using a polarizing filter on a 21mm lens.

 

[MattKing]

It is noticeable in photographs as I learned from using a polarizing filter on a 21mm lens.

Polarizers also add to the problem, because their effect varies according to angle.

 

[Sirius Glass]

Polarizers also add to the problem, because their effect varies according to angle.

As I found out the hard way.

 

[Alan Edward Klein]

Polarizers also add to the problem, because their effect varies according to angle.

Which lenses would you avoid using polarizers in 4x5? Does that apply to both BW and chromes?

 

[itsdoable]

See post# 7 about the 1/cos4(angle) falloff which explains the greater density in the corners.

It is an overall caste that you have to deal with.

1/cosine4 is a geometric effect, the circular aperture appears smaller (oval) when you view it from an angle. You see this particularly on pinhole images. It does not affect colour cast.

To my mind, this means that either:
1) the transparencies themselves have an overall magenta cast, which is barely visible in the lighter centre parts and most noticeable in the (darker) corners, and which in turn is reflected in the scans; or
...

...I still think that suggestion from my previous post about colour cast coming from filter is a further possibility to check...

These make sense to me, vignetting increases the saturation in the corners, which may accentuate an over all magenta cast.

Also, most neutral density filters I've used have a slight cast, if the center filter has a slight green cast, the corners (which have no density attenuation) would not have that cast, so when the image is colour corrected, the corner will go magenta.


PS: took me a while to re-find this thread after the move...!