There are situations in which I use a tilt-shift (TS) lens. TTartisans makes a 100 mm f/2.8 macro TS for Nikon Z mount that I currently am evaluating. I have a wide angle tilt-shift (Nikon F mount) that I have used for specific scenic images that need “large” depth of field or for tall trees or architecture for which I correct converging line perspective rather than doing this entirely in post (with film, a TS was by far the simplest solution). I contacted DxO support and received in the response: we don’t support Tilt shift lenses as they are not compatible with our lens correction models – end excerpt. As one can read in the very introductory overview – link below – from an Adobe Photoshop “article” that PS does work with TS lenses.
I have PL6 Elite Complete that requires the license activation of both DxO ViewPoint 4 and DxO FilmPack 6 Elite in addition to the base PL6Elite . Although ViewPoint (VP) allows for various perspective and “de-fishing” controls, my concern is for lens aberrations and other lens (not perspective) corrections that DxO builds into the lens optics module when the lens/body combination is supported. From what I understand, these lens “errors” are NOT corrected by DxO for TS lenses. Is my understanding correct?
Is there a document that explains how DxO characterises lenses? I have found none other than a marketing/sales type description (but no real engineering nor mathematics – even in the general sense of using a convolution transform without specifying the exact “equation” nor the actual polynomial or other approximation used for a rapid calculation (Approximation Theory).
When a TS is used locked straight (no TS in use), it is a regular lens. I had assumed that at least the basic lens errors that appear in “straight” could be corrected, even if the process of generating corrections for various T, S, or T and S positions would be too time consuming to measure and too much data/size in the module.
As a user of LF film cameras, where we have both front and back movements, I really can’t see how you can “normalise” a lens on a target, because the lens centre will only be central in one displacement or tilt position. As soon as you you de-centre the lens, even a millimetre, all the small imperfections that might have been mapped will no longer be relevant.
When tilting, the centre of the front standard could end up pointing to one of the corners of the image area
The red line is the plane of sharp focus which, as you can see, changes with the front tilt angle. You only really get sharp all over with a combination of that angle and a sufficiently small aperture. This gives you a “wedge” of acceptable focus which passes through the hinge point and progresses out along the line away from the camera, getting wider as it goes. Get the angle of line too steep and you get the miniature effect.
Thank you for presenting that elementary non-mathematical picture of how tilt and shift works, and the difference between a camera with both a tilt and shift lens (lens board) and a similarly adjustable imaging sensor (film in a pre-digital camera); hopefully this will clarify why one would consider a tilt-shift setup for certain imaging situations. All imaging systems have distortions, particular lens systems. This imaging errors include pincushion, coma, and other forms that cause loss of contrast even if the nominal focus is “sharp”. A brief non-mathematical discussion appears in Optical Aberration. In film photography, once the image was exposed but not yet developed, one could “push or pull” to change the ISO, and in printing, one could adjust the exposure within specific portions of the image through “dodging and burning”. Partial correction of all characteristics of an image are possible with digital methods and various mathematical transformations – often based upon the mathematical equivalent of a convolution but upon discrete (digital), not continuous (analog), data. Although as a fixed image plane camera (eg, the typical 35 or 120 format body) tilt-shift lens is both mechanically and optically “complicated”, nonetheless many of the basic distortions do not change “much” as the lens is tilted and/or shifted. It is for these optical aberrations that I would hope DxO could give some automatic lens module corrections as it does for a “fixed” lens – evaluating the lens in the un-tilted and un-shifted mode in which the lens is a “regular” lens (albeit a more complicated design and implementation than many regular lenses of the same focal length and aperture). Is the above more clear? As for the claim that the automagic corrections from DxO applied to a TS lens would eliminate the ability to control depth of field, I think that is misleading if not untrue.
I’d be astonished to see >1% of DxO customers owning/using TS lenses. Considering the price of quality TS lenses in combination with low market share and the unlimited possibilities of tilting and shifting makes such lenses totally uninteresting and challenging to characterise. In general, they don’t report tilts and shifts too, so, even if there were a module, we’d have to set things manually.
Long story short: Get better lenses so that you can ignore the modules
we have to do it manually ( focusing distance) for Sony ( even when it is clearly provided in raw file ) and Fuji …nothing new here in DxO world - so we might as well do it manually for tilt & shift values
I already use “better” lenses. As for the small market share/penetration argument, DxO supports AF-S NIKKOR 600mm f/4E FL ED VR $12,296.95US and Z 800mm F6.3 S $6,496.95US, neither of which lenses (amongst many others I could find on the DxO supported data base) exceed the 1 percent you mention and both of which have excellent optics (and mechanicals).
Does the camera/lens combination actually record tilt and shift metadata?
From what I can find, this doesn’t happen, although Fuji are now announcing a lens that does. Which leads me to believe that, if the metadata were available, it would more than likely be part of the manufacturer’s EXIF, not general metadata.
However, how do you realise a correction table for a lens that can be tilted, shifted and rotated, possibly in increments of less than one degree?
What if I tilt it by 0.76°, raise it by 1.27mm and rotate it by 23.8°? Where, on the image, would the optical centre appear? This would mean an infinitely complicated 3D table with countless measurements, not just a simple lookup table for focal length.