Jump to content
Sign in to follow this  

You always have a tilt/shift lens with you


Alan7140
  • https://www.fotozones.com/live/uploads/

It occurred to me after a post I made the other day that a very simple technique to get the effect of a tilt (and shift, if desired) lens may not be widely known. I must admit I had never heard of it until I accidentally stumbled upon it a few years ago while cropping a panorama, but if your subject is stationary and you have a good panoramic stitching program you can use any lens in this capacity.

 

The technique is simply to shoot two or more rows of photographs covering the subject equally above and below the desired end centre point for a tilt lens effect (or left and right in the case of shift).

 

Even something as basic as a single story building can benefit and a wide-angle "falling over" effect can be avoided if one cannot get back far enough (or high enough) to keep verticals parallel. The following is a front on shot, full frame (as in uncropped, not as in 135 frame) taken on my X-T1 with the 10-24mm lens set at 24mm. I was stopped from getting further back or higher by my front fence, so to get the whole house in this was the least wide angle I could use.

 

LeySzae.jpg

 

Shooting one photograph with a lot more foreground in it (and so cropping the roof substantially) and one with more sky (so the building was "leaning back noticeably) I then stitched the two in PTGUI, in this case using the cylindrical projection, but this will vary on the number of shots taken in which case Mercator or spherical may work better. Then I cropped the extraneous foreground back to roughly 3:2 for a very standard-looking shot, but with straight verticals. The benefit of doing things this way rather than using a perspective tool in an image editor is that you don't lose image on the sides, nor does the software have to interpolate extra pixels or discard others at the same rate, therefore image detail is better maintained. Flick between the two in a viewer and the difference becomes very obvious indeed.

 

dFEVSmE.jpg

 

Using a longer lens and do several rows of shots will actually substantially improve resolution, but in that case not only will all the subject have to be still for a longer period, but you will have to use a proper 3-axis panorama head and have the nodal point of the lens accurately worked out and set as the rotation point.

 

If you don't take architectural shots enough to warrant the large cost of a true tilt/shift lens, then this is a handy technique to bear in mind.

Sign in to follow this  


Comments

Recommended Comments

The eagle- eyed will notice some slight mismatches in joins of the verandah posts in these hand-held shots, but this won't happen with a proper pano head and accurate setup. Even so it's still handy in unexpected or tight situations, or when travelling and not carting a case-load of lenses.

Share this comment


Link to comment
Share on other sites

So to provide a better example, the following shot of my other building which stands almost twice the height and has less than half the back-up space as the house was taken hand-held at 10mm, comprising four shots stitched in AutoPano Giga using Planar projection:

HLGm9VX.jpg

 

Here's one taken from the same spot showing just how severe the distortion in a single shot would be:

eTXhlhW.jpg

Share this comment


Link to comment
Share on other sites

Alan, thanks, this is very interesting to me because I often find myself in a city with not enough room to avoid tilted building images.

 

Which of the two programs you mention would you recommend for this purpose?

Share this comment


Link to comment
Share on other sites

Alan, thanks, this is very interesting to me because I often find myself in a city with not enough room to avoid tilted building images.

 

Which of the two programs you mention would you recommend for this purpose?

 

Unfortunately there is no either/or - on some stitches PTGUI works better, others prefer AutoPano. Occasionally Photoshop's Photomerge does a better job than either, but that is seldom indeed.

 

PTGUI gives more controls, albeit in a much less user-friendly (or even comprehensible) interface than APG.

APG is much quicker to set up the images and actually start stitching, but PTGUI is much faster in the actual stitching and saving process (although I'm still using v.3.7 - they claim improved speed for V4.3, but I'm not of a mind to spend €79 on a bit more speed and some claimed "fixes" to problems I certainly never noticed).

Photomerge gives virtually no user control and trying to adjust afterwards if layers/mask options are chosen is an absolute nightmare, and it runs much slower than the other two.

Edited by Alan7140

Share this comment


Link to comment
Share on other sites

I tried with PS and got the usual deformed lines, using the adaptative wide angle filter fixed it some but lines were not straight

 

Tried MS ICE and it worked very well as it is possible to select the projection used for the stitching

Share this comment


Link to comment
Share on other sites

"...have a tilt/shift lens..."   Much wisdom latent within...

 

Alan I assist a considerable number of commercial architectural shooters in my market... Being located in my nation's capital does allow significant opportunity for architectural work... While this has not been my forte it is always wise to embrace opportunities to augment my revenue stream... especially during slower seasons... 

 

To this end I've researched the aforementioned at length...

And to be brutally honest with you a software solution is NOT a viable option for commercial work... at least in my market (yours may be entirely different).  On nearly all the high end large budge assignments I've assisted the client and/or their art director was aboard... They would approve final captures in real-time... Having the deliverable in near final form is absolutely paramount!  Many of the creative (art) directors would direct my photographer on subtle nuances they required to meet needs/expectations... Trust me on how many times a particular room/building facade etc would be re-shot in order to garner agreement and acceptance...   

 

To this end my tentative solution to date is likely going to involve the purchase of the LAOWA 15mm f/4 Wide Angle 1:1 Macro Lens available in Nikon F / Canon EF / Sony Alpha / Pentax K / Sony FE mounts. 

 

This glass has +/- 6mm Shift ability... full Frame Sensor Support...

All metal construction... 14 blade aperture... this list goes around the block...

 

And amazingly it is only $500 US dollars...  considering alternatives? There really isn't anything else at the price point of professional build/quality... save previously owned PC lenses...

 

The down side?

 

Manual focus only... but seriously a 15mm is set it and forget it... with a hyperfocal distance of 2.48 ft game's over... just do whatever it takes to frame effectively...

 

Yes it only corrects on one axis but it's your choice... and this is nearly always going to be vertical...  

 

I found this via architectural shooter forums... The results they posted are nothing short of amazing...

 

Important Note: This is not an effective macro lens... at 1:1 the subject is only a few millimeters from the front glass element... virtually useless for macro work...

 

Hope this helps or is at least food for thought...

Share this comment


Link to comment
Share on other sites

I only meant to suggest this as a solution for infrequent casual use, Thomas, not as a serious replacement for the real thing in commercial use.

 

My further playing with tilt/shift concepts has also led me down the path of adapting a PB-4 bellows for use with enlarger lenses, the shift and swing becomes shift and tilt if the bellows ids mounted on its side via adapted L bracket. Unfortunately the length of the bellows means that a lens of greater that 100mm needs to be used if infinity focus is required which limits architectural applications, but is great for focus plane shifts and perspective control in macro or product work.

 

I had never heard of the lens you mentioned, though - I'll investigate that as well.

Share this comment


Link to comment
Share on other sites

This certainly works and proper stitching programs work better than the features in general photo editors.  I fancy Autopano gives best correction of parallax errors, though there is an element of luck in it.

 

I rarely shoot panoramas anything other than hand held and AP stitches them well, though I have one or two projects in mind where a panorama head will help.

 

Incidentally it is the entrance pupil, not nodal point.  It is nodal point for swinging lens cameras (like the ??Horizon), but not when you swing the whole setup.

Share this comment


Link to comment
Share on other sites

Just to clarify in general terms:

 

There is a proper method for working out the swing point of axis for a particular lens which involves using a proper panorama head, having a distant vertical object edge, and a close vertical object like a light stand, then lining the two up on one side of the frame, and swinging the camera to see the point of registration of the two on the opposite edge. Moving the camera rail forward or back over the head's pivot axis will result at some point where the registration at one edge is the same as the other. That is the axis which that lens has to be swung on for perfect stitching - it will rarely be the same for each lens, and equally, if the front of the lens is deemed to be the entrance pupil, it will also almost never be that point, either.

 

This point should be worked out for each lens likely to be used, and at several focus distances, and the rail relative distance from 0 recorded in some sort of a table for reference in the field.

 

"Swinging the whole setup" will indeed be incorrect if pivoting on your own axis as is the natural way to pan, and for sure guestimating the front of the lens pivot point and walking around that will achieve a far better result, but true accuracy is only guaranteed with a proper three-way panorama head and the lens distance over the pivot point calculated as above. I use the term "nodal point" more as an indication that this is something that the user will have to calculate, as virtually no lens manufacturers actually publish the front and rear nodal points of their lenses, whether by measurement or diagram (Zeiss is the only one I've actually seen who do that, and then it's not always), and that point is where the lens must pivot for accurate stitching. Perhaps I should have termed it "Pivot Point", but then questions as to what that means would also certainly arise.

 

For single row panos with prime detail in roughly the same plane you can get away with not being so precise much of the time, but once you enter the realm of multi-row panos, and worse, stacked focus multi-row panos, getting the pivot point and the camera's vertical axis over that pivot point becomes critical, as you then have all four edges of the frame which need to be in perfect parallax alignment, or the whole house of cards tumbles. If a sapling in the immediate foreground appears on the other side of a distant tree on one side of the frame, and on the other side of that distant tree on the other side of the frame in the preceding shot to be stitched, a double image will probably result. While dedicated pano stitching programs do a passable job with removing such "ghosting" in general, this can really become a very real problem with highly detailed sharp foregrounds and equally sharp backgrounds.

 

While it won't look much heavily to this small forum size, the following shot taken on 10th December 2012 is a multi-row, focus stacked pano shot into a rising sun covering about 210° horizontally and about 100° vertically, 25,127px x 11910px (300MP near enough), well over 300 shots altogether from memory (including manual focus stacking), shot with an X-Pro1 and 35/1.4 lens @ f/8, on a very still summer morning around 5:00 a.m. with the pivot point using a Nodal Ninja Ultimate head having been worked out as above beforehand. There was no double imaging and it makes a beautiful and highly detailed print 48"x23" print, and is exactly the sort of thing that will show the slightest mismatch through an even slightly incorrect pivot point being used.

 

Dt49g8H.jpg

 

On the other hand, if there is no foreground detail or stacking involved, and just a single row pano, you can literally get away with murder, hand-held, particularly with a long lens where the pivot point almost becomes irrelevant - X-T1, 100-400mm @ 196mm):

 

oyIZkxs.jpg

Share this comment


Link to comment
Share on other sites


Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

  • Join Our Small Community

    Like what you see on Fotozones? Why not join up and make friends with like-minded photography enthusiasts from all across the planet. We are limiting our membership to no more than 2000 individuals, so if you are seeing this message there is still space available for you to join. We'd love to have you along. :)  

×

Important Information

By visiting this website you are agreeing to our Terms of Use, Privacy Policy & Guidelines.