Lightroom Lens Corrections

[tashley]

Advertisement (gone after registration)

On 2/13/2022 at 10:17 AM, Adam Bonn said:

AFAIK

Leica doesn't correct for distortion in jpegs only 'Italian flag' issues.. 

C1 doesn't automatically apply lens correction profiles for Leica

Even if you don't shoot jpeg, they can be extracted from the DNG using exiftool/etc

The DNG spec might suggest where to store lens info/corrections but Leica either might not include this (proprietary) info or they put it in maker notes. Either way adobe should be able to read it (adobe update needed methinks) 

 

Nope - my understanding is that Leica has yet to provide the correction profiles and that when they do they will go into the ‘waiting room’ until various Adobe product updates happen. So for now some lenses are of limited serious use. 

[Advertisement]

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

[scott kirkpatrick]

3 hours ago, tashley said:

Nope - my understanding is that Leica has yet to provide the correction profiles and that when they do they will go into the ‘waiting room’ until various Adobe product updates happen. So for now some lenses are of limited serious use. 

Well, I'm using C One 22 (Release 15.1) and it has had a pretty solid M11 color profile for the past two releases.  It also can use metadata provided by Leica in all DNGs for lens-specific distortion corrections in each color (to permit LCA corrections as well as reducing distortion).  Distortions are modeled as a polynomial expansion up to sixth order in the radius.  It is possible to capture mustache distortion, in which at the edges the distortion is barrel (+ive) and at the edges becomes more pincushion (-ve).  C One  gives a slider to modify the strength of this correction.  It can be zeroed out, following the metadata, left uncorrected, or interpolated in between.

[Adam Bonn]

9 hours ago, tashley said:

Nope - my understanding is that Leica has yet to provide the correction profiles and that when they do they will go into the ‘waiting room’ until various Adobe product updates happen. So for now some lenses are of limited serious use. 

Nope what?

Leica are going to give adobe lens info?

and what info, what lenses?

adobe already has the info (that it almost certainly had to figure out for itself) for the lenses it supports, the lenses are not new.

The adobe lens profile files are (eg) Summilux 50 asph.

Not Summilux 50 asph + m9, Summilux 50 asph + m240, Summilux 50 asph + m10 (the differences between the camera sensors would be part of that cameras raw support, so that the same lens profile works on all the cameras)

adobe won’t make new lens profiles for the m11, they’ll get their existing ones to work with m11 raw files

AFAIK Leica M lens native profiles don’t correct for distortion, just corner shading

 

[Adam Bonn]

5 hours ago, scott kirkpatrick said:

Well, I'm using C One 22 (Release 15.1) and it has had a pretty solid M11 color profile for the past two releases.  It also can use metadata provided by Leica in all DNGs for lens-specific distortion corrections in each color (to permit LCA corrections as well as reducing distortion).  Distortions are modeled as a polynomial expansion up to sixth order in the radius.  It is possible to capture mustache distortion, in which at the edges the distortion is barrel (+ive) and at the edges becomes more pincushion (-ve).  C One  gives a slider to modify the strength of this correction.  It can be zeroed out, following the metadata, left uncorrected, or interpolated in between.

That’s interesting, one of the things I dislike about C1 is that if I shoot with my Summarit 35 (which has fairly significant distortion) and open the DNG in adobe, LR detects the lens and automatically applies a profile that fixes it.

In C1 however, there is a profile for the ‘rit 35 but c1 seems incapable of automatically applying it based on exif data and when I apply it manually it does nothing to correct the distortion.

Of course I can (and have) made my own tweaks to that profile to correct the distortion, then saved it as a preset, but IMHO this is very poor support from C1 (= lens profile doesn’t auto apply and also doesn’t work)

C1 does enable the user to correct for barrel and pincushion distortion, but not both at the same time… (at least I don’t think you can stack lens profiles) this makes it impossible to create profiles that resolve any form of complex distortion.

Perhaps c1 plays nicer with the newer Leicas… (my latest M is an M10), they do seem to have taken Leica more seriously after the M10R release.

I just find that in terms of lens support, what’s no work at all in LR is best case a big faff in C1 and worst case not even possible.

 

[tashley]

11 hours ago, scott kirkpatrick said:

Well, I'm using C One 22 (Release 15.1) and it has had a pretty solid M11 color profile for the past two releases.  It also can use metadata provided by Leica in all DNGs for lens-specific distortion corrections in each color (to permit LCA corrections as well as reducing distortion).  Distortions are modeled as a polynomial expansion up to sixth order in the radius.  It is possible to capture mustache distortion, in which at the edges the distortion is barrel (+ive) and at the edges becomes more pincushion (-ve).  C One  gives a slider to modify the strength of this correction.  It can be zeroed out, following the metadata, left uncorrected, or interpolated in between.

Scott, I only have the DB version of C1 so I can't check but clearly there are many aspects to lens profiles. I am here talking specifically about the geometric corrections rather than those that deal with colour shifts and aberrations. LR currently cannot correct geometry for the M11 - which it can for the M10-R - and that is my topic in this thread. I do understand that the 'other' kind of profile - the colour profile itself, is available in LR (and I would assume C1) already. But are you saying that C1 can correct the geometry? Because A Leica store guy told me it couldn't, yet.

[tashley]

5 hours ago, Adam Bonn said:

Nope what?

Leica are going to give adobe lens info?

and what info, what lenses?

adobe already has the info (that it almost certainly had to figure out for itself) for the lenses it supports, the lenses are not new.

The adobe lens profile files are (eg) Summilux 50 asph.

Not Summilux 50 asph + m9, Summilux 50 asph + m240, Summilux 50 asph + m10 (the differences between the camera sensors would be part of that cameras raw support, so that the same lens profile works on all the cameras)

adobe won’t make new lens profiles for the m11, they’ll get their existing ones to work with m11 raw files

AFAIK Leica M lens native profiles don’t correct for distortion, just corner shading

 

I'm afraid that's not true: I have DNGs from a 35FLE shot on an M10-R and when I tick the Enable Profile Corrections tab in LR Develop module, the distortion is removed. You can tick and untick and magically lines will warp and unwrap. The 35FLE is only moderately affected so usually it doesn't matter but for use in architectural shots it does matter, and LR corrects it. This exact same process does not work with the same lens but on n M11 because, I am told, Leica have not yet provided Adobe with the profiles.

You say the profiles are the same: they are not. The profiles have data for things other than distortion so, though the distortion part should remain unchanged because the sensor is the same size, other aspects will change. For example, CA. The M10-R on the long side has 82% of the number of pixels of the M11. Therefore a CA instruction that says 'use a defying width of 4 pixels' for the M10-R would need to say 'Use a defying width of 5 pixels' for the M11.

In other words it isn't just a question of recycling the existing lens profiles because some aspects of them need tweaking. Whoever is responsible for doing that hasn't;t finished doing it, testing it and getting it into the relevant Adobe products yet.

[tashley]

Advertisement (gone after registration)

5 hours ago, Adam Bonn said:

 

C1 does enable the user to correct for barrel and pincushion distortion, but not both at the same time… (at least I don’t think you can stack lens profiles) this makes it impossible to create profiles that resolve any form of complex distortion.

I just find that in terms of lens support, what’s no work at all in LR is best case a big faff in C1 and worst case not even possible.

 

This is the nub of the M11 issue: some lenses, like the WATE, have quite a lot of quite complex distortion - with the WATE it's got more moustache than a Mariachi - and at the moment LR cannot correct it.

[scott kirkpatrick]

56 minutes ago, tashley said:

.... But are you saying that C1 can correct the geometry? Because A Leica store guy told me it couldn't, yet.

yes.  That step uses lens metadata which is in every DNG file.  C1 has always offered that.  It's the WarpRectilinear opcode.

I think you guys are lumping two (or more) elements into the term "profile".  There are metadata opcodes for every lens Leica has ever made, and both C1 and LR  understand how to use them, if you can find the controls in their workflow.  You can't stack Opcodes like WarpRectilinear, and they are one-parameter corrections, but they can contain the full mustache, not just a generic barrel or pincushion correction, since they are constructed for each lens.

Sensor color corrections require a profile for each camera, which takes time and budget for those to be supplied.  C1 seems to have been more aggressive than Adobe in this case.  Also, Adobe makes more things automatic, which means it is a hassle if you don't want them.  C1 tends to leave them up to you.

Edited February 15, 2022 by scott kirkpatrick

[scott kirkpatrick]

The description of WarpRectilinear is pretty opaque.  It's contained in the Tiff spec, but the things to understand is that lenses have cylindrical symmetry, so every thing depends on distance from the center of the image.  Mustache distortion comes from the fact that the top and bottom sides of the image are closer to the center than the corners, and a common trick in older lens designs was to undo barrel distortion at the corners.  The extra trick in WarpRectilinear is that a different distortion correction is presented for each of the three colors.  That corrects LCA. 

[Adam Bonn]

1 hour ago, tashley said:

I'm afraid that's not true: I have DNGs from a 35FLE shot on an M10-R and when I tick the Enable Profile Corrections tab in LR Develop module, the distortion is removed. You can tick and untick and magically lines will warp and unwrap. The 35FLE is only moderately affected so usually it doesn't matter but for use in architectural shots it does matter, and LR corrects it. This exact same process does not work with the same lens but on n M11 because, I am told, Leica have not yet provided Adobe with the profiles.

You say the profiles are the same: they are not. The profiles have data for things other than distortion so, though the distortion part should remain unchanged because the sensor is the same size, other aspects will change. For example, CA. The M10-R on the long side has 82% of the number of pixels of the M11. Therefore a CA instruction that says 'use a defying width of 4 pixels' for the M10-R would need to say 'Use a defying width of 5 pixels' for the M11.

In other words it isn't just a question of recycling the existing lens profiles because some aspects of them need tweaking. Whoever is responsible for doing that hasn't;t finished doing it, testing it and getting it into the relevant Adobe products yet.

You're confusing what adobe does with what Leica does. The lens profiles are adobe's not Leica's. The distortion removal is because of adobe not because of Leica.

When adobe fully supports the m11 they won't write a new lens profile, they'll write code that manipulates the DNG data for the M11 to work with the incumbent profiles

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!

Here are some of the adobe lens profiles for Leica (mac file path at the bottom if you want to go a see on your own system)

Note that there's only one profile per lens...

It does not have a listing for every conceivable M body and M lens combination

For example take the summarit 35 entries. There are two, one for the 2.4 and one for 2.5, why? Because they're different lenses = different profiles

But it doesn't have a .lcp profile for say

Leica Camera AG M8 (Leica SUMMARIT-M 35 mm f2.4 ASPH.) - RAW.lcp

Leica Camera AG M9 (Leica SUMMARIT-M 35 mm f2.4 ASPH.) - RAW.lcp

Leica Camera AG M240 (Leica SUMMARIT-M 35 mm f2.4 ASPH.) - RAW.lcp

Leica Camera AG M10 (Leica SUMMARIT-M 35 mm f2.4 ASPH.) - RAW.lcp

Leica Camera AG M10R (Leica SUMMARIT-M 35 mm f2.4 ASPH.) - RAW.lcp

Leica Camera AG M11 (Leica SUMMARIT-M 35 mm f2.4 ASPH.) - RAW.lcp

It makes do with one .lcp for all of the M cameras

This is because adobe need to code the rendering of the M11 DNG to the existing lens profile, not create a new lens profile every time Leica make a new camera

Or put how you said it, yes actually it's exactly a case of recycling the recycling the existing lens profiles, and you can confirm this by counting how many M bodies they are Vs how many lens profiles there are for each lens (6 vs 1)

Adobe aren't waiting on Leica here... Adobe need to write code that (as you accurately state) resolves CA issues and this code will work as part of the demosaicing of the DNG RAW data.

 

 

 

 

[tashley]

11 minutes ago, Adam Bonn said:

You're confusing what adobe does with what Leica does. The lens profiles are adobe's not Leica's. The distortion removal is because of adobe not because of Leica.

When adobe fully supports the m11 they won't write a new lens profile, they'll write code that manipulates the DNG data for the M11 to work with the incumbent profiles

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!

Here are some of the adobe lens profiles for Leica (mac file path at the bottom if you want to go a see on your own system)

Note that there's only one profile per lens...

It does not have a listing for every conceivable M body and M lens combination

For example take the summarit 35 entries. There are two, one for the 2.4 and one for 2.5, why? Because they're different lenses = different profiles

But it doesn't have a .lcp profile for say

Leica Camera AG M8 (Leica SUMMARIT-M 35 mm f2.4 ASPH.) - RAW.lcp

Leica Camera AG M9 (Leica SUMMARIT-M 35 mm f2.4 ASPH.) - RAW.lcp

Leica Camera AG M240 (Leica SUMMARIT-M 35 mm f2.4 ASPH.) - RAW.lcp

Leica Camera AG M10 (Leica SUMMARIT-M 35 mm f2.4 ASPH.) - RAW.lcp

Leica Camera AG M10R (Leica SUMMARIT-M 35 mm f2.4 ASPH.) - RAW.lcp

Leica Camera AG M11 (Leica SUMMARIT-M 35 mm f2.4 ASPH.) - RAW.lcp

It makes do with one .lcp for all of the M cameras

This is because adobe need to code the rendering of the M11 DNG to the existing lens profile, not create a new lens profile every time Leica make a new camera

Or put how you said it, yes actually it's exactly a case of recycling the recycling the existing lens profiles, and you can confirm this by counting how many M bodies they are Vs how many lens profiles there are for each lens (6 vs 1)

Adobe aren't waiting on Leica here... Adobe need to write code that (as you accurately state) resolves CA issues and this code will work as part of the demosaicing of the DNG RAW data.

 

 

 

 

Thank you - my bad: I was lead to believe by a guy in the Leica store that the geometric corrections were provided by Leica to Adobe and C1 and that the reason they aren't yet available is that Leica hasn't done so. Interesting that the same profiles are fully recycled: I totally get that the geometry doesn't change but the colour issues might well do so when you move from different cover glasses and pixel types and pitches, with their attendant different micro lens requirements. I guess this shows that the profiles are not optimal.

[Adam Bonn]

1 hour ago, scott kirkpatrick said:

yes.  That step uses lens metadata which is in every DNG file.  C1 has always offered that.  It's the WarpRectilinear opcode.

I think you guys are lumping two (or more) elements into the term "profile".  There are metadata opcodes for every lens Leica has ever made, and both C1 and LR  understand how to use them, if you can find the controls in their workflow.  You can't stack Opcodes like WarpRectilinear, and they are one-parameter corrections, but they can contain the full mustache, not just a generic barrel or pincushion correction, since they are constructed for each lens.

Sensor color corrections require a profile for each camera, which takes time and budget for those to be supplied.  C1 seems to have been more aggressive than Adobe in this case.  Also, Adobe makes more things automatic, which means it is a hassle if you don't want them.  C1 tends to leave them up to you.

 

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 LR profile for the 35 summarit corrects the distortion (one has the adobe profile, the other does not)

 

The C1 profile for the summarit 35 does not. (one has the C1 profile, the other does not)

The only way I've found to get the correction in C1 is to drag the distortion slider, C1 doesn't seem able to figure it out itself, perhaps I'm missing something here... (I've blown hot, tepid and finally cold with C1 other the years)

(In LR if one doesn't wish to use their profiles, then untick the 'enable profile corrections' in the Lens Corrections tool)

 

[tashley]

1 hour ago, scott kirkpatrick said:

yes.  That step uses lens metadata which is in every DNG file.  C1 has always offered that.  It's the WarpRectilinear opcode.

I think you guys are lumping two (or more) elements into the term "profile".  There are metadata opcodes for every lens Leica has ever made, and both C1 and LR  understand how to use them, if you can find the controls in their workflow.  You can't stack Opcodes like WarpRectilinear, and they are one-parameter corrections, but they can contain the full mustache, not just a generic barrel or pincushion correction, since they are constructed for each lens.

Sensor color corrections require a profile for each camera, which takes time and budget for those to be supplied.  C1 seems to have been more aggressive than Adobe in this case.  Also, Adobe makes more things automatic, which means it is a hassle if you don't want them.  C1 tends to leave them up to you.

I think there are actually several issues here:

1) The colour profile - as in how the RAW converter deals with the colours off the sensor so as to be as natural as possible. In LR you get to choose between M11, Adobe Landscape etc etc.

2) The Lens Corrections profile, which splits into geometry, CA and colour shift.

At the moment, LR has 1) but it has none of those I mention in 2) so you cannot correct the geometry, for now.

[tashley]

2 minutes ago, Adam Bonn said:

 

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 LR profile for the 35 summarit corrects the distortion (one has the adobe profile, the other does not)

 

The C1 profile for the summarit 35 does not. (one has the C1 profile, the other does not)

The only way I've found to get the correction in C1 is to drag the distortion slider, C1 doesn't seem able to figure it out itself, perhaps I'm missing something here... (I've blown hot, tepid and finally cold with C1 other the years)

(In LR if one doesn't wish to use their profiles, then untick the 'enable profile corrections' in the Lens Corrections tool)

 

But that's with an M10, right? Because it absolutely does not do this with M11 files at the moment.

[Adam Bonn]

2 minutes ago, tashley said:

Thank you - my bad: I was lead to believe by a guy in the Leica store that the geometric corrections were provided by Leica to Adobe and C1 and that the reason they aren't yet available is that Leica hasn't done so. Interesting that the same profiles are fully recycled: I totally get that the geometry doesn't change but the colour issues might well do so when you move from different cover glasses and pixel types and pitches, with their attendant different micro lens requirements. I guess this shows that the profiles are not optimal.

No problem, thank you.

I'd actually be highly surprised if Leica ever said "hey adobe and C1 and DxO and On One and Silkypix, etc here's all the data for our optical formulas, now please promise not to give it to anyone else, ok?"

A bit hunch-y on my part here... but their code will work like how sharpening and NR works (sharpens/blurs the image), it'll tweak the demosaicing of the RAW data to resolve CA, shading etc so the profile will say this lens needs xyz with CA and the look up table for camera will say ah this is an m11 we're working with here, so that means make abc adjustments at a pixel level for this particular camera

Something like that, but way more complicated (as all this under the hood RAW stuff generally is)

 

[Adam Bonn]

34 minutes ago, tashley said:

But that's with an M10, right? Because it absolutely does not do this with M11 files at the moment.

yeah I don't have an M11, so my lens corrections work fine (except in C1 where they do nothing, but that's just C1)

Actually that's an M9! (I also have an M10)

[scott kirkpatrick]

Of course Leica doesn't pass an archive of lens design data to anyone outside their collaborations, just standard metadata .  Even the L-mount consortium shares very restricted information. This has led to miscues such as the unexpected power drain that some L lenses impose, which trigger warnings and shutdowns from the SLs but not from Panasonic and Sigma cameras -- somebody didn't put the right constraints in a spec.  And C1, Adobe and others seem to do corrections differently, perhaps at different stages of the rendering, although the separation of lens-specific corrections and camera-specific corrections is clear, or should be.  I can't imagine why Adobe would make distortion correction a step that depends on the camera as well as the lens.

What matters for getting corrections out early for a new camera is early production cameras in the hands of the software folks.  This was possible for the M11, since the beta testing took place last summer, and the design was probably stable soon after that. (But not the firmware.)