[ptx] Using nona before autopano-sift for ultra wide-angle lenses
Daniel M. German
dmgerman at uvic.ca
Wed Aug 16 19:22:10 BST 2006
Pablo d'Angelo twisted the bytes to say:
Sebastien> On 8/8/06, Daniel M. German <dmgerman at uvic.ca> wrote:
>> >> [clens]
>> >> will remove the distortion (although the edges will still be naturally
>> >> distorted due to the fact it is a rectilinear lens).
>>
Sebastien> The distortion that bothers most autopano seems to be the rectilinear
Sebastien> distortion.
>>
>> What if you generate a fisheye-projection when you correct distortion?
>> Will that make autopano more precise?
Pablo> I haven't tried for images with a medium FOV, but for the images from my
Pablo> peleng fisheye, it autopano doesn't work well.
Pablo> One idea would be to create multiple smaller rectilinear images (lets say 4)
Pablo> from the wide HFOV images. Then these could be stitched together directly.
Pablo> This avoids the "reproject control points into original image" step, for
Pablo> which no software exists today.
I think this is a good idea, but the rectilinear disrtortions will
still be there. It is a matter of giving it a try. Perhaps one
approach is to try 2 images with rectilinear and fisheye projections:
For each projection
* Correct lens distortion, generating image in given projection
* Slice each image in quarters
* Try to find matching points in each corresponding pair of matching images.
This might work better for lanscape oriented images than portrait.
dmg
--
Daniel M. German "No one can be a great thinker
who does not recognize,
that as a thinker it is his
first duty to follow
his intellect to whatever
John Stuart Mill -> conclusions it may lead."
http://turingmachine.org/
http://silvernegative.com/
dmg (at) uvic (dot) ca
replace (at) with @ and (dot) with .
More information about the ptx
mailing list