parallax error in a rig of cameras

[giuseppe...@gmail.com]

Hi,

I am new to this forum.
I am planning to set a rig of six small cameras to take panos.
The distance between the nodal points of each cam is known,
As far as I know there is no way to solve this parallax error in stitching the images in Hugin, except in some particular cases.
I also read that some manufacterers of 360 degree cameras consisting of a number of sensors and lenses claim their included stitching software can, partially, deal with the parallax error.
My question is:
is there any exhisting algorithms which may (or which is scheduled to) be included in Hugin for the above purpose?

Thank you.

Best regards,

Giuseppe

[Gunter Königsmann]

If you watch the scene from a different position you see different things. No algorithm can undo that.

If there is enough overlap between the sensors what you can try is to find an object in the overlapping section that is sufficiently similar in all photographs and switch from one sensor to the other there. Which might break the consistence of the rest of the image: for objects in a different depth your 360° panorama will have 340 or 380°. Control points don't lie so they tell inconsistent camera angles and positions and near to the object you switch cameras things might disappear, get doubled or get way too wide or too narrow.

My guess the sensor's algorithm will skip trying to use control points in order to determine the relative position of the cameras: it only needs to be done once. In hugin you can stitch one image and then just replace the image files for the next set of images in order to do so.

Switching between images in the Right Place in hugin can be done using masks. It will automatically fine-tune the border within the remaining overlapping section between images.

Am Mi., 14. Nov. 2018, 18:22 hat <giuseppe...@gmail.com> geschrieben:

--
A list of frequently asked questions is available at: http://wiki.panotools.org/Hugin_FAQ
---
You received this message because you are subscribed to the Google Groups "hugin and other free panoramic software" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hugin-ptx+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/hugin-ptx/a4bc8d44-30e9-4146-8069-fbcebaf132e0%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Erik Krause]

Am 14.11.2018 um 17:54 schrieb giuseppe...@gmail.com:
> I also read that some manufacterers of 360 degree cameras consisting of a
> number of sensors and lenses claim their included stitching software can,
> partially, deal with the parallax error.

Seam optimization can hide the errors sometimes.

The only purpose of shooting from different viewpoints is to get stereo
images, but even to stitch those you have to deal with parallax errors.
There are very advanced systems that create a depth map and a 3D
reconstruction of the scene from such images. However, this is
computational photography, not simple image stitching.

--
Erik Krause
http://www.erik-krause.de

[bugbear]

Erik Krause wrote:
> Am 14.11.2018 um 17:54 schrieb giuseppe...@gmail.com:
>> I also read that some manufacterers of 360 degree cameras consisting of a
>> number of sensors and lenses claim their included stitching software can,
>> partially, deal with the parallax error.
>
> Seam optimization can hide the errors sometimes.
>

> The only purpose of shooting from different viewpoints is to get stereo images.

Indeed. But there may be a purpose to having multiple cameras, e.g. to get full
full spherical coverage at one moment. In this case the multiple
viewpoints is a consequence, not a purpose.

BugBear

[giuseppe...@gmail.com]

Thank you all.
I would need a set of multiple cameras because In many panos I would take there are moving objects/subjects.
So I need to take several pictures at the same time.
As far as i know it is impossible to solve the parallax error in stitching. My question was originated by some hints I got in internet. 
I read this: https://www.innocentive.com/ar/challenge/9933759
It presents a "challenge" for a system that (I am quoting) " given the output of a sensor (e.g. camera) at position A, predict what the same sensor at position B would see, given the relative positions of A and B to each other."
If that would be possible and I have, let's say, a camera A pointing to north and a camera B pointing to east, given the distance from A to B, I could have the view of camera B as seen by camera A and stitch the images to north and east from the same point of view.
Presented in this way it seems to me impossible to do (by the way, it seems the "challenge" has been awarded...)
But in a set of cameras we have some more information: in the overlap section we know what camera A and B see from their point of view. Possibly that information could be used to do the rest. I am a prophane and I have absolutely no idea how to do or even if the above is meaningless.

When driving I met by chance the google car. I saw their set of six or seven cameras is as big as a basket ball and the cameras are set not so close as possible, what I am trying to do with my set instead. It seems they are not much worried about parallax. So I guess they have some way to deal, partially, with this problem.
Of course I am talking about google but they cannot do what is impossible to do.

Thank you in advance for any remarks and suggestions.

Giuseppe Porciani