POV-Ray : Newsgroups : povray.animations : find the exact camera position (from video/images) : Re: find the exact camera position (from video/images) Server Time
17 May 2024 17:47:39 EDT (-0400)
  Re: find the exact camera position (from video/images)  
From: Chris B
Date: 10 Aug 2007 10:43:54
Message: <46bc79aa$1@news.povray.org>
"Benjamin" <nan### [at] webde> wrote in message 
news:46ab6914$1@news.povray.org...
> Hi,
> I am searching a program that is able to find the exact or even close 
> camera location (and look_at) of a real video I record with my camera. 
> Then I want to use these exported values(txt) in a pov-ray file and render 
> some objects onto the videoimages.
>
> I don't know if there is a program, but I think there could be one, that 
> compares the images of the video and find some points/lines to create a 
> location and look_at for the camrea in 3d space.
>
> Thanks,
> Benjamin
>

Hi Benjamin,
I'm not aware of a program that will do this. There are programs that can 
analyse images, looking for 'common' points. These are usually pretty 
sophisticated programs designed for particular categories of images (e.g. 
face recognition). I think that most movie production of this type uses kit 
to track the position and orientation of the camera when it's filming the 
real world so that the camera position and orientation can be reproduced in 
a virtual environment where the artificial objects can be rendered.

The general problem that you describe has two parts and both are quite 
tricky.

The first you mention is to find common points. It's easy to underestimate 
the difficulty of this because we can spot common points on two similar 
images so easily ourselves, but it's really quite a complex task for an 
image processing system to do unless you're able to introduce a distinctive 
object into the images that can make position fixing easier e.g. brightly 
coloured light sources that can easily be located by scanning the pixels of 
the two images for a colour in a particular range. This problem can be 
exacerbated by poor quality and low resolution images. Also, tiny movements 
in the camera between frames or lighting changes can cause the same point on 
an object to be represented differently on the image.

Even if you can identify common 2D positions on two images, establishing 
where they may be in space relative to the camera depends upon you having 
certain knowledge about the content of the image. Scale is a big problem 
because it's very difficult to tell whether you're seeing a small object up 
close or a large object at a distance. This is why movie makers can film 
something on a model constructed on a table top and make it seem like it's a 
huge real-world sized scene. Once again, one approach is to introduce 
something into the scene that you are filming of a known size and shape. So 
long as you know the camera settings used (ie the zoom) then you should be 
able to get enough information to use trigonometry to calculate the camera 
position relative to the known object.

Regards,
Chris B.


Post a reply to this message

Copyright 2003-2023 Persistence of Vision Raytracer Pty. Ltd.