Using trigonometric functions and polar coordinates, the whole thing could get fairly deterministic. Judging from what I see when using correction modules, I suppose that DxO would need a new algorithm and data in the modules… 2-3 years to implement?
I think that is a fairly good estimate. Especially bearing in mind that each manufacturer has to generate the data from the lens and then get the body to read that into the metadata, which is not yet a standard part of EXIF.
This discussion has gotten quite out of hand. It is evident that the technology to handle TS lenses with the full automation of DxO modules currently is not feasible. Gear• News FUJIFILM Unveils Two Tilt-Shift Lenses for GFX System and the New Fujinon GF55mmF1.7 R WR By FujiLove · 12.September.2023 . Excerpts from the above announcement state:
Its expanded configuration features the use of a large aperture aspherical, front element as well as 3 ED elements, providing advanced optics designed to eliminate the need for electronically correcting distortion and lateral chromatic aberration. Spherical aberration, comatic aberration and field curvature are suppressed to achieve edge-to-edge sharpness even during shifting.
AND
Furthermore, the lens configuration, with the aperture midway, suppresses distortion and lateral chromatic aberration to achieve edge-to-edge sharpness without relying on electronic correction.
AND
GF110mmF5.6 T/S Macro has a built-in sensor that measures changes, in degrees, to shift or rotation. As with GF30mmF5.6 T/S. changes can be monitored in Live View and are recorded to the image file’s metadata for reference during RAW processing.
End excerpts.
Thus, tilt is not encoded, just shift and rotation.
All I was requesting from DxO is NOT a full TS mapping (not trivial) but basic lens corrections for “straight” as a starting point for the photographer followed by manual corrections, with the lens model being selected manually if not present in the EXIF, etc, data. As for the quality of the Fuji optics to avoid improvement in post, I strongly suspect that were DxO to do the full measurement of these lenses (proprietary secret methodology used by DxO, except for some marketing/sales non-engineering material), such measurement would reveal “flaws”. These flaws might not necessarily be visible to the eye of the photographer nor the client.
The problem is that DxO always takes a scientific approach; It’s either perfect or it isn’t.
As far as I know they don’t allow exceptions to this rule.
Pascal
“Perfect” – there is no such thing either in measurement or engineering. In classical measurement theory, there are two non-removable sources of measurement error: “statistical” (sampling) and “systematic” (bias). Thus, “not” perfect, but “good enough”. As the DxO methodology is secret (not public science nor engineering), DxO does not reveal the error tolerances on the data it uses nor the uncertainties on the final result. Nonetheless, I strongly suspect that were DxO to subject the above mentioned Fuji lenses to the DxO methodology, there would be corrections despite the claims (above) that no such corrections are needed.
Fundamentally philosophic
Nothing against he strive for perfection…but how many lenses of any given model are measured in order to evaluate manufacturing tolerances?
Either/or is never the best choice. Much grey (e.g. in real life) between black and white (e.g. in politics).
Please note that I use black, grey and white in symbolic sense here. It’s not about good and evil, but about everything that lies between extremes. Being able to handle no-so-perfect situations is a capability that can advance PhotoLab (which is what we’re talking about after all) and make it a better product than it is today. I think that some added resilience could do DPL some good indeed.
Of course yes
You’re going to explain to DxO the futility of the race.
Pascal
I don’t think it was said “that no such corrections are needed.” (I’m not native english).
The (documented) DxO procedure said (something like) “We measure at all aperture, distances and focal lengths (zoom)”.
This methodology seems to me hardly compatible with this kind of lenses.
Pascal
well, Canon is rumored to make TS AF lenses for RF mount - if ( IF ) that happens then will see if they provide optics correction in their CR3 files as they do for all of their RF lenses … and if they do the ball will be in somebody’s court
and we can assume that they do not test dozens of samples of each lens to account for manufacturing variances and tolerances, so all that precision is … doubtful …
in the spirit of kicking the dead horse - DxO needs to provide options for users to use optics correction data that manufacturers embed in raw files IF SO USERS DECIDE, users are not stupid - if they find DxO optics modules better ( when and if available ) they will use DxO’s optics modules !
There are no such thing as perfekt optics. Especially when you begin moving optics.
Fixed focal length, fixed aperture, fixed subject distance and you might get pretty good.
I actually wrote a longer reply to you but I changed my mind as I get the impression that you are more annoyed just because of some unknown or personal reason.
I’m happy with DxO’s quality work. They push hard for us to enjoy far better images today in post than what we captured many years back.
I will continue buying their - and others - software as long as they offer me a benefit worthy the cost of the software. Heck - I sometimes buy or support software anyway because I want to support great effort and dedication.
I’ve used the word perfect advisedly
Of course DxO knows the limits.
What I mean is that DxO offers the best of its technique or offers nothing. These facts are documented.
I prefer the word confidential.
What I understood was the care taken and the time given. The secret could lie in the coupling between the sensor’s intrinsic characteristics and the lens.
At last, Lens mounting precision today allows you not to work in batches.
Pascal
One can always take a deep dive into DxO’s patents which explains a lot
I never saw such a description of a patent, or any patent. Thanks.
George
… or confuses the heck out of most people