Re: [PanoToolsNG] Re: How constant are a,b,c,d,e for a lens?
- John Houghton wrote:
> --- In PanoToolsNG@yahoogroups.com, paul womack <pwomack@...> wrote:Interesting. I have done this once.
>> Is my camera a bit loose, or am I doing something wrong,
>> or are my expectations wrong?
> You cannot expect uniform optimization results using handheld shots.
> There is always the likelyhood of parallax problems, unless
> everything in view is very far away. You end up with the optimizer
> trying to fix parallax errors by distorting the images with
> inappropriate lens parameters. For your calibration test, you would
> do better to take a full 360 degree panorama, having the camera on
> the head with the entrance pupil accurately positioned at the no
> parallax point. Include some 50% overlaps. Careful control point
> placement and optimization should ensure a good result.
If/when I do another, I will compare the a,b,c with interest.
> However, when using the calibrated lens parameters in a heandheld
> camera project, don't expect an excellent optimization report. That
> would be unrealistic. Place control points on distant features only,
> and the optimizer should align those very well. But expect to see
> some stitching errors on nearer features - they are inevitable.