Home

Please support mflenses.com if you need any graphic related work order it from us, click on above banner to order!
SearchSearch MemberlistMemberlist RegisterRegister ProfileProfile Log in to check your private messagesLog in to check your private messages Log inLog in

Shift-Lenses vs Software: some examples (TS-E 4/17mm L)
Post new topic   Reply to topic
Home   Manual Focus Lenses
View previous topic :: View next topic  


PostPosted: Fri Feb 17, 2023 8:14 pm    Post subject: Reply with quote
Quick n dirty example I made earlier to show why I often prefer to use a normal lens and stitch than use an UWA.

This is as far away from the building as I can get, shot with a Lumix 1.7/25 on my OM-D E-M5.



Then I quickly shot 8 frames handheld from the same position, same lens.

Stitched them together in photoshop, almost no manual correction needed which is due to the lack of distortion in the frames - the 'normal' length lens is pretty much distortion free:



Rotated and cropped, not sure what the FOV is exactly, but it's pretty wide and undistorted, result is a 39mp image:



Not much PP involved there.


PostPosted: Fri Feb 17, 2023 8:49 pm    Post subject: Reply with quote
That seems to have worked well for that subject.

Nevertheless, for me, that's simply not a way of working I would enjoy. I'm an physicist/engineer and I enjoy more doing that with optics & mechanical solutions rather than with a computer.

I tried stitching a few times; it works, but although you may find this hard to believe, still too much PP time for me. I just REALLY don't enjoy image PP. One single levels adjustment and possibly one profile conversion per image is about my limit of patience there. The rest I much prefer to do in the field with the equipment.

That's just my way of enjoying photography as a hobby.

EDIT: one thing I do need to mention is that the multiple photo & stitching process from a single vantage point as demonstrated above introduces a significant level of apparent barrel distortion across the whole the image that a rectilinear UWA or shift lens would avoid, so would require a bit more work in PP yet.


PostPosted: Fri Feb 17, 2023 8:59 pm    Post subject: Reply with quote
Well, if you refuse to do PP, then this topic really isn't for you I'm afraid.


PostPosted: Fri Feb 17, 2023 9:16 pm    Post subject: Reply with quote
Just sometimes stitching does not bring the expected rectilinear perspective.



PostPosted: Fri Feb 17, 2023 9:18 pm    Post subject: Reply with quote
Well, for that to turn out so bad, you must have done something wrong in how you shot it.


PostPosted: Fri Feb 17, 2023 9:21 pm    Post subject: Reply with quote
Ha-ha, I'd like to be capable to control this, as I liked the result and wouldn't mind to reproduce it when I wish! Unfortunately the process was completely automated, the software just wrongly took repetitive architectural elements for the same stitch point.

As it is capable to make arbitrary errors, the fun is still there.


PostPosted: Fri Feb 17, 2023 10:07 pm    Post subject: Reply with quote
I actually like it, it reminds me of the 80s and 90s Polaroid collages of David Hockney, which is what inspired my interest in digital stitching ofimages - Hockney was one of my tutors for a while.



PostPosted: Fri Feb 17, 2023 10:09 pm    Post subject: Reply with quote
Like 1


PostPosted: Sat Feb 18, 2023 9:46 am    Post subject: Reply with quote
iangreenhalgh1 wrote:
Well, if you refuse to do PP, then this topic really isn't for you I'm afraid.


That's been my point all along; for me, shift lenses are anything but obsolete. Friends


PostPosted: Sat Feb 18, 2023 11:09 am    Post subject: Reply with quote
alex ph wrote:
Just sometimes stitching does not bring the expected rectilinear perspective.



Whilst obviously gone wrong, the top of the building in this image neatly demonstrates a point I've been wanting to make / question I wanted to ask, but not yet done.

In recent stitching software, can you tell the software the desired projection intent? (I haven't used recent stitching software, so I don't know)

I can see that Lightroom allows for [Spherical/Cylindrical/Perspective] to be chosen, but that's not very specific (I mean is it spherical stereographic, spherical equidistant, spherical equisolid, spherical orthographic etc. They all look quite different)

If no precise projection intent can be specified, then surely by virtue of all the images having been taken from the same centre of perspective, the software will attempt to merge together a load of linear planar projections into what will be most likely an overall equidistant planar projection (either spherical or cylindrical). This will cause curvilinear distortion of any straight lines not going through the centre, which the software will have to introduce by distorting the individual component images until a best fit is obtained. If more images are taken with with a smaller FOV lens, the accuracy of the overall equidistant projection will become progressively better.

Similar (possibly localised) distortions can be seen in Ian's final image, where e.g. the central and corner gutters in the first image are perfectly aligned, whereas in the final image the are at different angles. Likewise, the facia boards on the gable end are perfectly straight in the first image, but noticeably curved in the final image, as well as the roof pitch having been reduced significantly. The overall height of the building has been squashed down as well compared to what a shift-lens would image.

Now, Ian, I'm not at all saying that the final image you produced is not a successful stitch, but noticeable curvilinear, angular and foreshortening distortions away from a pure rectilinear planar projection have been introduced that a shift lens would not.

If the final image is rendered as an equidistant spherical projection, then the conversion to to a shift-lens-equivalent or linear UWA-equivalent image would require a projection transformation, not a perspective transformation. That would be using e.g. de-fishing software, not keystone correction software.

I do know that clients in architecture in particular, generally would be looking for a linear planar projection without any vertical keystoning/foreshortening effects. In order to keep the image natural looking this would usually mean a modest wide angle lens with a modest amount of shift, nothing too extreme. They would not be very tolerant of any curvilinear distortion or uneven foreshortening resulting from use of a PP technique ill-chosen from the available different PP techniques for perspective correction.


PostPosted: Sat Feb 18, 2023 2:11 pm    Post subject: Reply with quote
I didn't bother doing any corrections of distortion or perspective, I could have, but as I said, it was a quick n dirty demonstration.

If a lack of distortion is desired, that can very easily be achieved.









Of course, this is just the simplest type of stitching using Photoshop, I usually use more sophisticated stitching software such as Autopano Giga which has a large variety of different projections available. Tweaks can then be made using Photoshop.

In terms of how sophisticated these tools are, a t/s lens is a sharp stick, but Autopano Giga is an intercontinental ballistic misisle - the discrepancy in capability is that huge.


PostPosted: Sat Feb 18, 2023 3:34 pm    Post subject: Reply with quote
iangreenhalgh1 wrote:


Of course, this is just the simplest type of stitching using Photoshop, I usually use more sophisticated stitching software such as Autopano Giga which has a large variety of different projections available. Tweaks can then be made using Photoshop.

In terms of how sophisticated these tools are, a t/s lens is a sharp stick, but Autopano Giga is an intercontinental ballistic misisle - the discrepancy in capability is that huge.


I agree with Ian, Auto Pano Giga is the best tool for the job with tweaking in Photoshop.

For amateur, ie my own use, I don't use perspective control at all but its necessary when making huge panos

In my darkroom days perspective could be controlled in printing by tilting the baseboard, using f22 the depth of focus was deep enough.

Some interesting 'perspectives' here in this post and I've learned a lot. Thanks


PostPosted: Sat Feb 18, 2023 4:19 pm    Post subject: Reply with quote
iangreenhalgh1 wrote:
I didn't bother doing any corrections of distortion or perspective, I could have, but as I said, it was a quick n dirty demonstration.

If a lack of distortion is desired, that can very easily be achieved.



Of course, this is just the simplest type of stitching using Photoshop, I usually use more sophisticated stitching software such as Autopano Giga which has a large variety of different projections available. Tweaks can then be made using Photoshop.

In terms of how sophisticated these tools are, a t/s lens is a sharp stick, but Autopano Giga is an intercontinental ballistic misisle - the discrepancy in capability is that huge.


I have to take your word for Autopano Giga doing a better job; philslizzy seems to agree with you there. The image above doesn't look right at all to my eyes. Maybe I'm more sensitive to distortions away from the perfectly square and level patterns across a frame.

To my eyes, there is still significant uneven angular distortion throughout the frame, which is not the building itself as the first shot shows it is built perfectly square. The degree of incremental foreshortening still visible as you move towards the top of the frame simply wouldn't be present in a normal rectilinear planar shot taken with a shift lens. I'm sure that could be corrected on a stitched image in PP if required, but that would take even more time and by using a shift lens you simply don't have to worry about any of that in PP.

The physics/mathematics involved in removing the keystoning (incl. removing the foreshortening) are fairly simple for a single image where the image plane was at a uniform angle relative to the subject plane, i.e. taken using a rectilinear WA lens pointing up at an angle. PP software can do that relatively easily, provided you are using the correct tools, but will be subject to a loss in resolution and magnification of lens aberrations. When multiple images taken from a single viewpoint are stitched together, it becomes a lot more complicated, and I am not yet convinced the image PP software gets that right, at least not without significant extra work.

E.g., the round clock face above exhibits vertical compression (i.e. foreshortening); in a level-camera UWA or shift-lens shot where the image plane is kept parallel to the front of the building, the clock face would still be perfectly round. A perfect sphere on the other hand (so not a flat round clock-face), would be imaged vertically stretched in a shift lens shot, yet still be perfectly round in a stitched shot. If the image above had been taken with a single shot looking up using a WA or UWA lens, then a true perspective shift in PP would automatically adjust for the required removal of that foreshortening when removing the convergence. When also using stitching, it appears to get rather more difficult judging from the results above.

I'm not saying one method is better than the other, but if a rectilinear planar imaging is required (and most architectural clients would insist on that), a shift-lens or technical camera with rise & fall would seem a quicker and easier option to me... (and for a prof. photographer surely time = money, incl. image PP).

Let's not forget this whole debate started with your assertion (in another thread) that shift-lenses are outdated and no longer relevant. I am yet to be convinced; it seems to me that there are still plenty of situations where shift-lenses (and technical cameras with rise & fall) are relevant, esp. for professionals with architectural clients who won't accept any funny image distortions/artefacts, where time = money, and where the final image might not need the high-resolution of stitched imaging (e.g. project & sales brochures etc.) Call that a niche if you wish...


PostPosted: Sat Feb 18, 2023 4:52 pm    Post subject: Reply with quote
RokkorDoctor wrote:
alex ph wrote:
Just sometimes stitching does not bring the expected rectilinear perspective.



Whilst obviously gone wrong, the top of the building in this image neatly demonstrates a point I've been wanting to make / question I wanted to ask, but not yet done.

In recent stitching software, can you tell the software the desired projection intent? (I haven't used recent stitching software, so I don't know)

I can see that Lightroom allows for [Spherical/Cylindrical/Perspective] to be chosen, but that's not very specific (I mean is it spherical stereographic, spherical equidistant, spherical equisolid, spherical orthographic etc. They all look quite different)

If no precise projection intent can be specified, then surely by virtue of all the images having been taken from the same centre of perspective, the software will attempt to merge together a load of linear planar projections into what will be most likely an overall equidistant planar projection (either spherical or cylindrical). This will cause curvilinear distortion of any straight lines not going through the centre, which the software will have to introduce by distorting the individual component images until a best fit is obtained. If more images are taken with with a smaller FOV lens, the accuracy of the overall equidistant projection will become progressively better.

Similar (possibly localised) distortions can be seen in Ian's final image, where e.g. the central and corner gutters in the first image are perfectly aligned, whereas in the final image the are at different angles. Likewise, the facia boards on the gable end are perfectly straight in the first image, but noticeably curved in the final image, as well as the roof pitch having been reduced significantly. The overall height of the building has been squashed down as well compared to what a shift-lens would image.

Now, Ian, I'm not at all saying that the final image you produced is not a successful stitch, but noticeable curvilinear, angular and foreshortening distortions away from a pure rectilinear planar projection have been introduced that a shift lens would not.

If the final image is rendered as an equidistant spherical projection, then the conversion to to a shift-lens-equivalent or linear UWA-equivalent image would require a projection transformation, not a perspective transformation. That would be using e.g. de-fishing software, not keystone correction software.

I do know that clients in architecture in particular, generally would be looking for a linear planar projection without any vertical keystoning/foreshortening effects. In order to keep the image natural looking this would usually mean a modest wide angle lens with a modest amount of shift, nothing too extreme. They would not be very tolerant of any curvilinear distortion or uneven foreshortening resulting from use of a PP technique ill-chosen from the available different PP techniques for perspective correction.


Have you tried Hugin panorama? You can do what you want to achieve, and much more. Infact, it is daunting how much you can do with that software...

I like browsing through the different types of projections you can pick. My most used case has been panoramic video game screenshots...


PostPosted: Sat Feb 18, 2023 5:45 pm    Post subject: Reply with quote
Hugin is great, I use it too, very powerful and capable.

Quote:
The image above doesn't look right at all to my eyes. Maybe I'm more sensitive to distortions away from the perfectly square and level patterns across a frame.


There is pretty much zero distortion away from the square and level in that frame.



PostPosted: Sat Feb 18, 2023 8:24 pm    Post subject: Reply with quote
Hmm...

I don't know how to bring this to you gently, but the overlayed grid actually proves to me precisely the distortions I was seeing. I'm not talking about barrel/pincushion distortion, but e.g. looking at just the tower, after the distortion correction you did, the bottom (grey) part of the tower facade seems to have a vanishing point very far to the right, the top (brown) part of the facade seems to have a vanishing point very far to the left. To my eyes the distortion correction you did has made the tower look vertically twisted, and the grid confirms that to me.

It may not be much, but as I said, maybe I am very sensitive to those type of distortions.

TBH the varying degree of foreshortening (increasing vertical compression towards the top of the frame) is something that feels the most unnatural to me. It would be natural if it were still accompanied by converging verticals, as in the first shot you showed with regular keystoning (in your post from 2:11pm). But removing the converging verticals without also removing the progressive foreshortening doesn't look right to my eyes.

Again, with a shift lens all of this is a complete non-issue that doesn't need any work in PP, hence still my personal preferred choice of equipment.


PostPosted: Sat Feb 18, 2023 8:38 pm    Post subject: Reply with quote
This is getting us nowhere, you're seeing things that aren't there and keep insisting a t/s lens can do better when, in fact, it can't do very much at all and is extremely limited in what it can actually achieve.

You won't do PP, that is upto you, but it means you're only doing half a job, even in the film days, half the work was done in the darkroom.

So I offered a couple of samples that are as basic as possible, that have had the very barest minimum of PP. Of course, there is a huge variety of distortion and perspective correction that could be done with more complex examples, but when people are so adverse to even the most basic PP....

I've wasted enough time on this topic, it reminds me too much of the old arguments over HDR a decade ago when some insisted all HDR was horribly garish and had no place in 'proper' photography. A decade on, it's become far more widely accepted, even if many of the critics still don't understand how it works or how to properly utilise it.


PostPosted: Sat Feb 18, 2023 9:50 pm    Post subject: Reply with quote
Camera position level, half-building-height away, on tall ladder half-building high -- no tilt/shift, no PP needed(?)

Automate camera mount for perfect stitching -- how fast can it go? Smile (one frame equivalent "shutter speed")

https://en.wikipedia.org/wiki/Scheimpflug_principle



https://en.wikipedia.org/wiki/Lidar



PostPosted: Sat Feb 18, 2023 9:53 pm    Post subject: Reply with quote
iangreenhalgh1 wrote:
This is getting us nowhere, ...


I agree, let's leave it there then.

But in reference to what started this whole debate, going forward, can you please try and stop making absolute statements re. fellow enthusiasts doing things wrong because they they are not using your methods.

This is a hobby for most of us, part technical, part artistic. There are both technical and artistic degrees of freedom; we make our own preferred choices, and that should be 100% OK. Shift lenses have not been made obsolete by computer software, just like musical instruments have not been made obsolete by the computer-synthesizer.


PostPosted: Sun Feb 19, 2023 10:44 am    Post subject: Reply with quote
RokkorDoctor wrote:

But in reference to what started this whole debate, going forward, can you please try and stop making absolute statements re. fellow enthusiasts doing things wrong because they they are not using your methods.



PMSL You're having a laugh with that statement. I made a very simple statement that was clearly just my opinion, there was nothing absolute or vehement about it. Then along comes the unplesant pedant from the alps and takes the opportunity to attac kme.

Will I let this stop me from expressing my opinion?

Will it hell as like...

Oh, and I suggest you go back and look at the short, simple statement I made, as nowhere in it did I say or imply that people were doing things wrong nor was their any criticism actual or implied.

I shouldn't have to defend myself from false accusations, so please be more careful in what you think I have actually said in future.


PostPosted: Sun Feb 19, 2023 12:22 pm    Post subject: Reply with quote
iangreenhalgh1 wrote:
RokkorDoctor wrote:

But in reference to what started this whole debate, going forward, can you please try and stop making absolute statements re. fellow enthusiasts doing things wrong because they they are not using your methods.



PMSL You're having a laugh with that statement. I made a very simple statement that was clearly just my opinion, there was nothing absolute or vehement about it. ...


Peace, man; I didn't phrase that properly. What I'm suggesting is that it is not what you say literally, it is about how you say it and anticipating how that may be interpreted. Especially by a new member just joining the forum.


We had a new member joining the forum, his first post. He had taken the time to do a review on the pros/cons of using a shift lens, and said it had been fun using it, and concludes with the remark that

"Whilst images rendered are fundamentally flawed, they have a character which I have found to be eminently appealing"


So, the first comment he reads is yours:

"Shift lenses have ceased to have any purpose in the digital world - perspective correction is a matter of a couple of clicks in software."

That is an absolute statement; your opinion obviously, but an absolute statement nevertheless that does not acknowledge the possibility that shift lenses may still have their uses, even in the digital world.

In context of that users' first post, and his efforts having done a review, does that not potentially read like am implied message he has been wasting his time reviewing a shift lens because it now should be done with software? Not even a thank you for the review? A fine welcome. (fortunately rectified immediately by another member on the forum).

It is about anticipating how other people might interpret the comments you make, and qualifying them accordingly, as annoying and unnecessary that may seem to you. And yes, I got that wrong myself in the previous post.


PostPosted: Sun Feb 19, 2023 2:22 pm    Post subject: Reply with quote
visualopsins wrote:
Camera position level, half-building-height away, on tall ladder half-building high -- no tilt/shift, no PP needed(?)


You say that in jest perhaps, but I have actually seen people do that here (low level domestic properties, to be fair Wink )

visualopsins wrote:
Automate camera mount for perfect stitching -- how fast can it go? Smile (one frame equivalent "shutter speed")



That makes a lot of sense. If you could automate the system with highly accurate stepper motors for all movements, after a one-time calibration you could get a computer to accurately render any projection you like. You might still have issues with people & cars milling about in the shots though, so best for quiet & stationary scenes.

visualopsins wrote:
https://en.wikipedia.org/wiki/Lidar


A bit overkill for my objectives Wink


PostPosted: Sun Feb 19, 2023 2:53 pm    Post subject: Reply with quote
visualopsins wrote:

Automate camera mount for perfect stitching -- how fast can it go? Smile (one frame equivalent "shutter speed")

https://en.wikipedia.org/wiki/Scheimpflug_principle



I think a webcam may do the job if the sensor is fixed and the lens is free to move. Wink


PostPosted: Mon Feb 20, 2023 4:40 pm    Post subject: Reply with quote
and despite all the bloviation and bollocks....

I still think t/s lenses are obsolete in the digital world

Smile


PostPosted: Mon Feb 20, 2023 5:42 pm    Post subject: Reply with quote
iangreenhalgh1 wrote:
and despite all the bloviation and bollocks....

I still think t/s lenses are obsolete in the digital world

Smile

Sure Ian.
But they are so cute and I like so much to operate them. Wink
Post new topic   Reply to topic
Home   Manual Focus Lenses
Goto page Previous  1, 2, 3, 4  Next
Page 2 of 4