Page 1 of 1

moving camera, not panning

PostPosted: Wed Nov 29, 2017 5:22 am
by Cornelis
Is it possible to produce a pano by taking images through MOVING the camera L to R in a horizontal line represented in the object, rather than by PANNING the camera on a pan head?

Re: moving camera, not panning

PostPosted: Wed Nov 29, 2017 10:57 am
by Panoram1x
The short answer is yes ...
Then things get complicated.
Only a satisfactory result will be possible if certain conditions are met.

1 # That the photographed subject to be flat, two-dimensional. If they have a third dimension in direction to the camera, parallax errors will be generated and this will be very difficult to solve.

2 # The conditions of the space where the photos are to be taken should allow the camera to point perpendicularly to the photographed surface, without yaw or pitch.

3 # Do not use a fisheye lens, only rectilinear lenses.

This has worked fine for scanned images, images from a microscope, mapping images from a drone, tapestries and paintings...

See: ... or_artwork

Re: moving camera, not panning

PostPosted: Wed Nov 29, 2017 12:02 pm
by Cornelis
Hello Panoram1x
Thank you very much for your rapid response. :) I see you are in Barcelona - I was there 10yrs ago - photographing Gaudi & Jujol art. Superb outputs.

Re the moving camera objective:

1. I have quickly discovered - just like you have written - that having a 3rd dimension in the direction of the camera viewpoint generates useless outputs when stitching is attempted.
2. I can see that moving a prepared slide on a microscope with a camera mounted, would essentially be equivalent to moving the camera. The side prep - being essentially flat - would create no problem with a 3rd dimension, as in 1 above.
3. I had already taken the steps necessary in your # 2 & # 3. I find even a Zeiss 35mm lens is too distorting, never mind fish-eye lenses.
4. A question for you: Suppose you are moving a camera over a painting, the image of which has in it the optical illusion
of depth. Is this a problem?
5. If there is no problem in 4 above, it appears to follow if a photo image of a 3-D object is printed onto photo paper, followed by photographing that 2-D image (and moving the camera), that should yield a successful stitch. What is your opinion in this situation? It seems to replicate the painting situation.

Re: moving camera, not panning

PostPosted: Wed Nov 29, 2017 1:23 pm
by Panoram1x
I'm glad you keep a good memory of your visit to Barcelona.

Your # 4. I do not think there's any problem. The image is just that, an image, and therefore reproducible. The illusion is created in the brain of the observer.
The observation conditions are another thing. Optical illusions depend on the way of looking, the angles of vision, the peripheral vision ...
When reproducing this image it would be necessary, with much probability, to redefine the observation conditions to maintain the illusion of depth.

As for # 5, there is something that I must have misunderstood.
I do not see how it makes sense to make a panoramic from another image (the photo of the 3D object). It would only be a copy of what we already have.
And if you want to assemble several of these images, parallax problems will reappear.
I think there is something that I have not understood correctly.

Re: moving camera, not panning

PostPosted: Wed Nov 29, 2017 11:59 pm
by Cornelis
Again, I thank you for your rapid, & thoughtful response.

I can see I need to explain further my previous #5, in order for you to imagine what I'm thinking:

(a) I understand that for successful stitching, I first I need to get a flat image of the target object.
(b) Suppose the object is a cup of tea & saucer - 3-D
(c) I take a close-up photo of that tea & print it onto paper. I now have a 2-D image - say A2 size. Let's pretend that print is a painting.
(d) I now take multiple (handheld) photos of that print - moving from Left to Right.
(e) I stitch these multiple images into a panorama.
(f) I now have a 2-D pano of a 3-D object.

Now let's get more complex:

(i) I arrange the following objects in a long line on a table: the cup of tea/saucer, a teapot, a milk jug, and a plate with a slice of bread on it.
(ii) I use the same camera/lens as above.
(iii) Because the line of objects is long, I have to take multiple shots - with a small amount of overlap for each shot - moving the camera as I go from L to R. Say 6 shots
(iv) Plainly, there is going to be some perspective differences in the shots for each object. Minimize these by standing well back, using a long focal length lens - say 200mm.
(v) print these shots onto paper.
(vi) cut off the image overlaps in the the prints, and arrange these trimmed prints in a long line of a single 2-D image.
(vii) repeat step (iii).
(viii) repeat step (e).
(ix) I now have (f) ??

In reality, I'm trying to get moving camera panos of some fascinating sandstone cliffs by the ocean. The cliffs are very long, but additionally have a small amount of depth. I will photograph the cliffs from the beach at very low tide (occurs once a month).
See the cliffs I'm writing about on the internet - "Painted Cliffs, Maria Island, Tasmania, Australia". They are amazing.

Given the unfamiliarity to the viewer of these objects, it is possible that small perspective problems in the final pano will not be noticeable.


Re: moving camera, not panning

PostPosted: Fri Dec 01, 2017 4:15 pm
by Panoram1x
The test with the table and the objects can work ... but ...
... only if each object appears whole in at least one photo. In this way the sewing lines can be established in the spaces between the objects (Automatically, manually by means of the mask tool in Autopano or in extreme case it will be necessary to render layered and use masks in Photoshop)
But if to cover any of the objects it is necessary more than one shot parallax problems will reappear.

And this brings us to the cliff ...

I've been seeing images of the place (it's so nice!) and it does not seem that at low tide there is a beach that allows to get away and have a good shooting angle.

The cliff wall is very wavy and the "third dimension" component in the direction of the camera is remarkable ... it would be necessary to be able to take a certain distance to take the photos.

If it is as I suppose and there is no beach, the question comes. Is it possible to take the pictures from a boat?

Could you try to do some tests on a facade or something similar?

Re: moving camera, not panning

PostPosted: Tue Dec 05, 2017 12:07 am
by Cornelis
Hello Panoram1x
Thank you, once again for putting in the time that you have in your response.
I apologize for the delay in my response. I was actually 15 minutes away from boarding the ferry to Maria Island for the cliffs on Sunday. It takes me 7 hrs of driving to get there, so it took 2 days to get there & back.
However, a huge storm had developed by Sunday morning– 6M swells on the ocean, 70km winds, torrential rain, temperature 10 deg C. Very low tide at 3pm.
So, I had to abandon my trip this time. I have rescheduled for Jan 1, 2 2018. The times that I choose for visits are the times when the moon is full. Consequently, the lowest tides = some beach= a short land-based stand-back from the cliffs.
You are quite right - the cliffs have a significant 3rd dimension in the direction of the camera. However, so tantalizingly attractive!! This 3rd dimension appears to be the major problem for the moving-camera technique. I may have to be satisfied with a number of separate normal panning shots. I shall attempt to visit the island in the future.
The water/boat technique does not appear to me to be a viable option. The water is very rough, the boat will be bobbing about. Positioning the boat & moving it L or R, would be very tricky. Besides, the cost would be prohibitive. Never mind the logistical problem of getting late afternoon sunlight from the west! (The cliffs face west).
Thank you for your very useful comments regarding my hypothetical table top with objects on it. I can understand how you are seeking to avoid parallax error, by having the objects appear once only in a camera movement technique.
I’m thinking that for this photo shooting circumstance to be successful, you would have to ensure that the camera was never moved in the vertical direction – only the horizontal axis. This requirement suggests that the horizontal camera movement is achieved by having the camera mounted on rails. Not an in-field option.