Multi row panorama: Totally avoiding perspective problems with tilt shift lens?

[sm906]

Hi folks,

I am working on an architecture project to produce large format indoor panoramas (final size >= 3meters x 1 meter). I am using nodal point technology with the Giga Pan Pro head and focus stacking, HDR and stiching afterwards. To avoid as much distortion as possible right in the single frames I use 70 - 200mm whenever possible. This works pretty well except for distortions in the areas of the cealing. This can be corrected in Photoshop, but it's not always very easy and sometimes extremely time consuming for the subjects I am photographing.

I am wondering whether these distortions can be totally avoided when using a tilt-shift lens. My idea is, to newly correct the perspective distortion with the tilt shift lens for every single row of the panorama. Has anybody experience in such a workflow? On the other hand my expection is, that part of the distortion is caused by the rotation around the nodal point and cannot be avoided in camera even when using a tilt shift lens.

Kind regards

Thomas

[Bart_van_der_Wolf]

I am wondering whether these distortions can be totally avoided when using a tilt-shift lens. My idea is, to newly correct the perspective distortion with the tilt shift lens for every single row of the panorama. Has anybody experience in such a workflow?

Hi Thomas,

I'm not sure what kind of distortion you are seeing (an example would help), but I assume that it is caused by the projection method you use (rectilinear?) and a very wide angle of view (>110 degrees?).

If it is indeed caused by the rectilinear projection, which is the almost the only way to have straight lines in the image remain straight, then a T/S lens will give exactly the same effect. When there are no straight line restrictions at all, then a different projection method may help.

Cheers,
Bart

[BernardLanguillier]

As Bart said.

One more point is that avoiding distorsion in each single frame may not be that super critical when using software such as PT gui, they do a great job at correcting that automatically.

Cheers,
Bernard

[allegretto]

would a T/S and a good deal of overlap fix this?

[Bart_van_der_Wolf]

would a T/S and a good deal of overlap fix this?

Depends on the cause of the actual distortion, but it's unlikely. A stitched panorama reconstructs the optical source image, and projects that on a surface. The shape of that surface determines the amount and kind of distortion.

Here are some examples of how that projection distortion looks when displayed on a flat plane. By changing the yaw and/or pitch, and/or a horizontal or vertical offset, the position of the point at which we look in the center of the image, e.g. by raising or lowering the horizon through the image, different distortions can be combined. A Shift lens can be exactly simulated with a rectilinear projection from a rectilinear lens, and some offset added.

Cheers,
Bart

[allegretto]

Depends on the cause of the actual distortion, but it's unlikely. A stitched panorama reconstructs the optical source image, and projects that on a surface. The shape of that surface determines the amount and kind of distortion.

Here are some examples of how that projection distortion looks when displayed on a flat plane. By changing the yaw and/or pitch, and/or a horizontal or vertical offset, the position of the point at which we look in the center of the image, e.g. by raising or lowering the horizon through the image, different distortions can be combined. A Shift lens can be exactly simulated with a rectilinear projection from a rectilinear lens, and some offset added.

Cheers,
Bart

Yep, as Mercator demonstrated a few centuries ago (why Greenland looks bigger than Australia). Just though that if the interval were small enough you could at least squeeze it out of the center, but can't avoid the elongation problem at the edges.