Definitely need something like this for those odd lens that are not on the DXO to do list. These could be old and / or obscure lens or more modern lens - camera combinations using adapters that few would use and thus have limited market appeal.
2 Likes
IIRC the Lensfun data are crop-related, rather than sensor-specific. But half a loaf is better than no bread, and a crust is too.
An observation on this subject:
- why did Adobe abandon “Adobe lens profile creator” which should - according to some - make DxO (Optics Pro at the time) totally obsolete?
I think things are more complex than what we users think!
Making a preset for an unsupported lens is possible, complicated but possible.
I happened to do some for unsupported lenses. As long as it is a fixed focal length: for a zoom it is practically impossible.
The distortion is relatively easy to correct: it is the same regardless of the diaphragm (f-stop).
For vignetting, each opening must be analyzed and corrected. This means in theory one preset per opening (f-stop). in practice one can be satisfied with two or three corrections since the vignetting generally decreases while passing from f / 2.8 (for example) to f / 4 and then becomes negligible.
Of course, you have to go read the exifs: there is no possibility of automatically choosing your preset accordingly. And if the Exifs are not present (not transmitted or nonexistent: old lens) it gets complicated … and I ignore the fact that some corrections are sensitive to the shooting distance!
That being said, a preset, even a basic one where only distortion (approximately) and vignetting are corrected are a good starting point.
I would have thought the EXIF was always present in a RAW - but what do I know, I only shoot Sony.
And for the rest, that’s what I meant by “half a loaf” and “crust”.
1 Like
I would say a main problem is variation in quality between the same lens. So testing one lens is not enough, you have to test multiple. And most users will only have access to one. So your profile would only work for your own lens, and not necessarily for others. Never mind not a bad idea to support this, but I guess complicated and very niche.
As I understand the Lensfun database, their measurements are derived from more than one copy of the lenses - though I have to say I have never been part of their project. But as I said, a crust may be only a crust, but it is better than nobread at all.
1 Like
You seem to really like bread! 
If this is true, I recommend that you read the books by Steven Lawrence Kaplan. He is an American university professor specializing in the history of bread and baking. He is very famous in France.
His books are a treat for foodies and really exciting.
Kind regards,
I suggested that over a year ago in the section “Which feature do you need?” with zero user interest and no DxO comments:
Je ne connais pas la version francaise. Mais l’on pourrait bien dire “Un bon ‘tiens!’ vaut deux ‘tu l’auras’”. Et encore DxO ne dit pas que nous l’aurons …
Thanks for the link to the tutorial. About your former proposal, the subject is nearly the same, but this one brings a new approach:
• The profile’s editor would be part of ViewPoint. ViewPoint has no longer improvements, and its interest (as an independent application) decreases with the new version of the Nik Collection.
• The profile’s editor would be also used in a creative purpose, not only to fix optical issues. For example, fisheye/de-fisheye, let or increase the distortion only on the edge… It is the reason why this feature is more appropriate to ViewPoint than to DPL.
• The profile’s editor should be able to fix more complex optics defects than those DPL can do with only the manual corrections, like moustache distortion, color shift, left-right inhomogeneity (centering defect), and maybe others.
• As an independent application, it could be applied on various kind of files, included the DNGs that are not recognized by DPL (e.g. DNG from smartphones).
• Into DPL, exifs data could be automatically read to propose the proper set of customized profiles, or to let selecting one in a library.
2 Likes
Just being able to save manual lens corrections as a named profile would be a good start.
PS. Cluttering presets with this for each custom lens would not be an efficient solution.
I dunno Alec. I only have half a dozen unsupported lenses. It’s true that some of them are zooms, so I might need to save more than one preset x lens; but even if that turned out to be the case, it still looks more attractive than having to remember the same number of corrections (not to mention the order in which to apply them).
All the same, in the interests of saving labour, I would like DxO to find a way of making this more automagical.
I wonder if anyone has tested the existing lens profiles in PL3 to make sure they’re actually any good? I photographed my daughter’s socially-distanced, masked wedding Friday night, it was under a gazebo, and in a couple of shots that showed vertical columns, the automatic lens profile (Nikkor 24-120F/4) showed distinct pincushion effect. I had to switch to manual to adjust that out. The same images looked right in Lightroom with the lens profile applied.
Of course, these are things that are only obvious in situations like that.
Can only speak for myself, but “no”. OTOH I haven’t noticed any obvious errors on the few of my long-discontinued lenses for which PL3 has profiles.
The nature of the system, though, is that the quality of one lens profile doesn’t guarantee that of another - the process of making them might, as long as the lens manufacturer’s process for constructing the lens were equally bulletproof.
I think you would buy into my proposal if it meant that an individual user was able to modify the DxO supplied lens profile when it clearly produced imperfect results in his body/lens combo!
DxO has had to periodically fix lens correction errors. Once or twice for my lenses over the years. In my experience, they are very willing to help with lens correction questions, even when it turns out that the profiles don’t need to be fixed. So I suggest submitting a test RAW photo with description of the problem to support.dxo.com.
Not that I expect to get my wish (commercial sensitivity and all) but I would LOVE to know what DxO are doing in the lens modules to pull the sharpness out of my images. Because try as I might, I cannot get any other software to get the same level of sharpness out of the files.
It may simply be that I am unskilled in such things, but it seems like magic to me that pin sharp images come from my camera.
There should be an option to bake in a lens profile using a DxO supplied test target.
- user prints test target on high quality printer
- user scans test target using a calibrated scanner
- user takes several shots of the printed target using a DxO-supplied procedure
- user sends all to a DxO app or cloud service for baking in a custom profile.
Profile would be usable for viewpoint, for vignette correction and for lens sharpness correction.
User-defined profiles would be sharable in DxO cloud library.
Separate suggestion created.
I could live with that, though I confess I don’t understand the reason for your second bullet. I think its attraction to DxO should lie in the fact that it allows multiple users to provide results for the same body/lens calculation, and thus to even out any errors in the third bullet step. As I understand it, that is the process by which the lensfun database is kept sensible.
The reason for the second bullet is to account for errors in the printing process (ink-printer-paper combination).
I am afraid that a full profile will requires several multiple targets: one with a grid for the geometrical correction, another with strips of different sizes and orientations for the definition mapping, maybe again another with spots of color and calibrated grey for the color shift, and so on.
Another difficulty is the relative position of the test target with the camera. It must be in a strictly perpendicular plane, centered on the optical axis, controlled horizontally, and vertically. The quality of the light must also be controlled (power and color spectrum).
In short, a complete test bench is necessary to guarantee the reliability of the measurements.
Calculate a mean of different profiles will not improve the process, because the sources of errors cannot be compared neither evaluated.
It is why beskope profiles never compete with professional ones. But the interest of the feature still remains!
1 Like