[ptx] Streetscape-panorama, new scientific paper

[Ed Halley]

Max Lyons' forums have been discussing this technique for a variety of
purposes.  The general term emerging for this is "ortho-mosaic," where
there are a number of camera positions, usually in parallel facing
directions, to develop a large flat image from many components.

    http://www.tawbaware.com/forum2/

The most general case of this is to take LOTS of skinny pictures.  A
flatbed scanner is just a one-pixel-wide camera with some lights on the
moving arm.  If you use more images, you have to 

If you must use PanoramaTools (and any GUI above it) for this sort of
project, minimize the FOV for each image, so that each image takes up a
very small portion of the overall "image sphere" that the tools assume.

On Thu, 2005-02-24 at 18:08 +0800, Sebastian Nowozin wrote:
> 
> Hello everybody,
> 
> I asked on this list a while back how to go about producing a 
> streetscape panorama, where you have one long linear image that shows a 
> streetblock (basically what Google Maps has now for big US cities), by 
> combining many photographs.
> 
> Today I found a recent paper from the IEEE Visualization 2004 
> conference, written by Augusto Román, Gaurav Garg and Marc Levoy from 
> Stanford.  They do just that in a semi-automatic way from a video.
> 
> "Interactive Design of Multi-Perspective Images for Visualizing Urban 
> Landscapes"
> 
> http://graphics.stanford.edu/papers/multi-cross-slits/
> 
> At the above URL there is the original paper, an example image and a 
> video.  I think their technique can be extended to use images instead of 
> a video stream (ie. making it more practical for us panorama geeks) and 
> the technique does not sound too difficult to implement.  Registering 
> the images may be similar to the recent "large map from many scanned 
> images" discussion on this list.
> 
> Any comments on the technique or similar papers known to you? :)
> 
> bye,
> Sebastian
> 
> 
-- 
[ e d @ h a l l e y . c c ]