Digital Lens Corrections

[Jeff S]

Advertisement (gone after registration)

26 minutes ago, Bgood said:

After all in the 1960s software corrections weren’t an option. 

Instead we used a host of other methods… optical lens corrections, view cameras, tilt-shift lenses and such to address distortions, myriad darkroom techniques to optimize results, etc. New tools; similar issues. Again, the discerning user adapts as needed.

[Advertisement]

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

[lct]

1 hour ago, Bgood said:

BastianK has shown, that the M10 applies digital corrections when using a Summilux-M 35 ASPH FLE - even to the raw files and even when “lens recognition” was turned off, see here: https://phillipreeve.net/blog/review-leica-35mm-1-4-summilux-m-asph-fle-mk-ii/ and search for "forced vignetting and color shift correction to the raw files".

Strange this forced vignetting correction thing. I have no experience with the Summilux 35/1.4 FLE v2 but i have the v1 of it, and both use the same code if i'm not wrong. Then when lens recognition is off it is off, there is no vignetting correction anymore. At least on the M11. Am i missing something?

Edited March 30 by lct

[UliWer]

vor 22 Minuten schrieb lct:

Am i missing something?

No. 

Without detecting the lens the camera cannot apply any „forced“ vignetting correction. 

And I fear totally different sorts of „corrections“ are mixed up here. Lens design allowing some distortion which is corrected digitaly to achieve better resolution is a completely different matter than digital correction of vignetting. 

[3D-Kraft.com]

vor einer Stunde schrieb lct:

At least on the M11. Am i missing something?

May be, the M11 is "cheating" less here.

vor 46 Minuten schrieb UliWer:

Without detecting the lens the camera cannot apply any „forced“ vignetting correction.

Yes, but the irritating thing was that even with "lens recognition" disabled, the camera still made corrections to the RAW file. Only by thwarting the recognition of the 6-bit code did the camera stop making RAW adjustments. For me, this raises the question, if the "raws" still can be called raws...

vor 46 Minuten schrieb UliWer:

And I fear totally different sorts of „corrections“ are mixed up here. Lens design allowing some distortion which is corrected digitaly to achieve better resolution is a completely different matter than digital correction of vignetting.

Of course, vignetting and color shift corrections are other types of interventions than distortion correction. But all are "digital lens corrections" because they try to improve the interaction of the lens properties when imaging on the sensor as addressed in the thread title. Distortion correction is only one of the many things, which lens profiles try to support in post. We also do not know, if vignetting and color shift corrections are the only corrections, Leica applies (or will apply in the future) in M series.

Edited March 30 by 3D-Kraft.com

[adan]

51 minutes ago, lct said:

Am i missing something?

One minor feature - see bottom.

When Leica introduced the M8 in 2006, they also introduced 6-bit coding. Whose main purpose was to fix tonal (not distortion) errors, such as vignetting and color vignetting, that were imposed by digital sensors. The same lenses worked fine, and as designed, on film.

The difference?
- Sensors have complex 3-D silicon structures that interfered with light striking the surface at a glancing angle - and more so the closer one got to the corners. And would vary both with focal length and the distance from the sensor of the lens (i.e. focal length). This was "worse" in 2006 - the introduction of back-lit sensors improves it.

- film was a smooth, homogenuous structure of silver-impregnated gelatin, that can be exposed equally from any angle. (Including, if one has a leaky camera back, from behind).

Leica had a choice - redesign all their M lenses "for digital", and orphan the millions of M (and even screw-mount) lenses already in users' hands. Or come up with something in their digital M that would correct the erroneous effects of the sensors.

The 6-bit codes allowed the camera to detect which exact focal length and type (specific design) lens was mounted. And then brighten the sensor-induced vignetting, in a specific pattern for each focal length from each era (back to ~1980) so that the result looked more like their behavior on film. Especially with short-focus wide-angle lenses, which includes the 35 Summilux ASPH versions (basic design introduced 1994).

However, (to lct's question) since no one would have the 6-bit coding already on their lenses in 2006, Leica also included a default correction for vignetting in the camera firmware. "NO lens detected? In that case, use this slight generic brightening of the corners."

I expect it is approximately the correction needed for a 50mm lens (well, a 51.6mm lens, on average, for Leica's "50s") - the most commonly-used lens.

[adan]

45 minutes ago, 3D-Kraft.com said:

For me, this raises the question, if the "raws" still can be called raws...

Raw files have not been purely raw-off-the-sensor for decades. They have tone curves applied, been compressed, perhaps gain adjustments made, converted to "reduced raw", and other massaging. Even occasionally some noise-reduction.

Canon was sharpening their raw data before writing it, at least at one point. To overcome deficiencies in early CMOS sensors. 

See also Adobe's headaches with processing Nikon .NEF files around 2010 - exactly because Nikon used a proprietary format with "undisclosed" changes to the data, in between the truly-raw sensor output and the "raw" file itself.

[3D-Kraft.com]

Advertisement (gone after registration)

vor 42 Minuten schrieb adan:

Leica had a choice - redesign all their M lenses "for digital", and orphan the millions of M (and even screw-mount) lenses already in users' hands. Or come up with something in their digital M that would correct the erroneous effects of the sensors.

The 6-bit codes allowed the camera to detect which exact focal length and type (specific design) lens was mounted. And then brighten the sensor-induced vignetting, in a specific pattern for each focal length from each era (back to ~1980) so that the result looked more like their behavior on film. Especially with short-focus wide-angle lenses, which includes the 35 Summilux ASPH versions (basic design introduced 1994).

Agreed. But still quite a dilemma...

All 35mm Summiluxes except the "Steel Rim" are coded with the same 6-bit code (not sure, how the Summilux II (1966-1995) will be coded, when sent to Leica). The Summilux 35mm ASPH FLE v2 is already the third iteration since "digital era" began for Leica.

So, when Leica starts applying modifications to the raw files based on this 6-bit code, they have to keep these characteristics of the lens similar at least to the first ASPH from 1994 even though lens design allowed many improvements for digital sensors since then.

vor 26 Minuten schrieb adan:

Raw files have not been purely raw-off-the-sensor for decades. They have tone curves applied, been compressed, perhaps gain adjustments made, converted to "reduced raw", and other massaging. Even occasionally some noise-reduction.

Canon was sharpening their raw data before writing it, at least at one point. To overcome deficiencies in early CMOS sensors. 

See also Adobe's headaches with processing Nikon .NEF files around 2010 - exactly because Nikon used a proprietary format with "undisclosed" changes to the data, in between the truly-raw sensor output and the "raw" file itself.

Yes, but here we talk about signal processing independent of the attached lens. Usually these manufacturers provide lens profiles that allow raw converters to apply the lens corrections after the raw file was recorded.

Edited March 30 by 3D-Kraft.com

[lct]

1 hour ago, adan said:

However, (to lct's question) since no one would have the 6-bit coding already on their lenses in 2006, Leica also included a default correction for vignetting in the camera firmware. "NO lens detected? In that case, use this slight generic brightening of the corners."

Or, by hiding the coding pits, the reviewer tricked the camera into thinking that an uncoded lens was mounted, so the camera applied the last lens profile entered manually ... Reminds me of a music...

 

[pgk]

On 3/29/2025 at 11:09 AM, Bgood said:

I keep reading that there is always a price for digital lens correction but if the result is better with it why worry about since it reduces size and weight?

In brief, and roughly speaking, the thing is that by allowing software to correct for those lens parameters it works very effectively with (vignetting, distortion, some chromatic aberation), this then enables lens designers to design lenses with better correction for other parameters which software cannot deal with (astigmatism, coma, etc.). So the end result is a better final image than a lens might otherwise be capable of providing, without software adjustment, or at least for the price anyway. So I don't see this as 'cheating' at all, software adjustment is effectively just an additional tool in the lens' designers toolbox. That said, in order to maintain good improvements there are no doubt limits to those parameters which software is viable for I am sure - excessive distortion, vignetting, chroma, etc., will potentially show artifacting if requiring excessive correction, so as ever it wll be a balance.

[Bgood]

The responses to this thread have taught me a lot. I thought software corrections amounted to cheating or something only for cheap lenses like super zooms. I now see I was wrong. I guess it’s all a compromise. Even high end M lenses have some compromises. That’s not to say they aren’t good but they are not perfect, even a costly APO. 
 

Thank you. 

[250swb]

13 hours ago, adan said:

 

However, (to lct's question) since no one would have the 6-bit coding already on their lenses in 2006, Leica also included a default correction for vignetting in the camera firmware. "NO lens detected? In that case, use this slight generic brightening of the corners."

I expect it is approximately the correction needed for a 50mm lens (well, a 51.6mm lens, on average, for Leica's "50s") - the most commonly-used lens.

But vignetting usually only appears at wider apertures and goes away by f/8. So is this default correction still applied where no vignetting is likely to occur, say shooting at f/16, in which case corners would become over bright wouldn't they? If the correction is applied based on the camera's guessed aperture that's fine if you're shooting at f/2 and the camera thinks it's f/2 and corrects mild vignetting, but put a polarising or ND filter on at f/2 and the camera now thinks it's f/16 so vignetting won't be corrected. So a default value of correction seems like a strange way to show off the quality of the lenses if it can be fooled so easily and also lead to confusing results for photographers, perhaps blaming other components like the lens or filters for 'faults' in the photograph caused by the firmware itself? 

[pgk]

36 minutes ago, 250swb said:

But vignetting usually only appears at wider apertures and goes away by f/8. So is this default correction still applied where no vignetting is likely to occur, say shooting at f/16, in which case corners would become over bright wouldn't they? If the correction is applied based on the camera's guessed aperture that's fine if you're shooting at f/2 and the camera thinks it's f/2 and corrects mild vignetting, but put a polarising or ND filter on at f/2 and the camera now thinks it's f/16 so vignetting won't be corrected. So a default value of correction seems like a strange way to show off the quality of the lenses if it can be fooled so easily and also lead to confusing results for photographers, perhaps blaming other components like the lens or filters for 'faults' in the photograph caused by the firmware itself? 

I have noticed that if I fail to reset my camera (M9) to read the 6-bit code, after having used an uncoded lens, that occasionally one or two corners show a highlight warning if exposure is marginally high, especially on the 75mm Summarit. I've never bothered to figure this out, but it makes sense that if a very small amount of vignetting correction is applied to all images then this is what might happen. Jury's still out but its a possibility and if it is applied it is only a marginal correction.

[adan]

1 hour ago, 250swb said:

But vignetting usually only appears at wider apertures and goes away by f/8.

Natural lens vignetting, yes - but not necessarily the sensor-induced vignetting and color shifts (which are the main point of the exercise). The light rays will still be hitting the sensor at the same glancing angle, whether at f/2.0 or f/16. Although see below, regarding focus settings.

Really, you'd have to ask Leica's firmware engineers about the fine details of their algorithms.

Conversely - focusing really close (0.7m, or now even less) is going to move lenses further from the image plane, and slightly reduce the shallowness of the angle of the light rays reaching the sensor.

Do the algorithms track the movement of the RF mechanism? And reduce the vignetting corrections accordingly, by estimating the lens position/focus?

And given that the M Digitals' "guesstimate" of the aperture in use can be off by up to 2 stops (the common Leica digital M newbie complaint - "I know my lens was set to f/4 - but the EXIF says it was at f6.8!!?")- are we putting too much brain-grease into trying to expect "precision" in a system based on such loosely-approximate (a.k.a. "ratty") inputs in the first place? 🤪

[250swb]

1 minute ago, adan said:

are we putting too much brain-grease into trying to expect "precision" in a system based on such loosely-approximate (a.k.a. "ratty") inputs in the first place? 🤪

That is the point isn't it, if you want precision (or best approximation) use a Canon or Nikon which properly communicate between lens and camera. And don't try and make a silk purse out of a sows ear. The M system is very compromised if looking for perfection, and anyway the photograph is made by having an opinion during post processing (or in the darkroom). I'm just outraged Leica think they can secretly stop me from enjoying a bit of vignetting! 😁

[UliWer]

vor 15 Stunden schrieb adan:

Leica also included a default correction for vignetting in the camera firmware. "NO lens detected? In that case, use this slight generic brightening of the corners."

I could reproduce the differences in vignetting correction with a Summilux 1:1.4/35mm asphalt FLE (1. Version) and an M10:

With lens detection off there is less vignetting than with lens detection off and the area for lens detection covered by a piece of paper. Though lens detection off and the area for lens detection uncovered shows more of a slight magenta tint. If you press the lens release button and overturn the lens a little bit so that 6-bit detection cannot work properly you again get different results: more vignetting and a slight cyan cast. In the "overturned" position the frames for 28/90mm are triggered so it looks as if there was a different default correction for 28, 35 and 50mm (the longer focal lengthes for which the frames are also triggered seem to be irrelevant).

[jaapv]

Maybe the thing here is that we are expecting the camera to resolve everything. However in digital photography hardware and software are 50-50. What is wrong with Flat Field Correction in post? 

[UliWer]

vor 8 Minuten schrieb jaapv:

What is wrong with Flat Field Correction in post? 

I don't see anything wrong with it. 

Though I think there is a difference between arbitrary correction to make a photo look as you like it and lenses which are dependant on correction to be usable.

The 21mm Super-Angulon is not really usable with digital color cameras because of the strong magenta cast. With "Adobe Flat Field correction" applied I got good results. Same with a 1:3.5/19mm Canon Lens.

Other lenses I know (from Leitz/Leica or others) which can be used with digital M cameras don't really depend on any correction - even though the results may seem a bit different if you use lens correction or not. The lens of the Q (and I think most SL-lenses) seem to depend on digital correction. 

[lct]

13 minutes ago, UliWer said:

[...] The 21mm Super-Angulon is not really usable with digital color cameras because of the strong magenta cast.

No color cast on the M11, thanks to its BSI sensor. Ditto on my Sony a7r2 mod that has a BSI sensor as well. I mean with the S-A 21/3.4 as i have no experience with the S-A 21/4 but i suspect it is the same for it.

[Advertisement]

Hallo,

Look here: Leica M

[jaapv]

37 minutes ago, UliWer said:

I don't see anything wrong with it. 

Though I think there is a difference between arbitrary correction to make a photo look as you like it and lenses which are dependant on correction to be usable.

The 21mm Super-Angulon is not really usable with digital color cameras because of the strong magenta cast. With "Adobe Flat Field correction" applied I got good results. Same with a 1:3.5/19mm Canon Lens.

Other lenses I know (from Leitz/Leica or others) which can be used with digital M cameras don't really depend on any correction - even though the results may seem a bit different if you use lens correction or not. The lens of the Q (and I think most SL-lenses) seem to depend on digital correction. 

Flat field is not arbitrary. It uses reference images to neutralise the photograph.