Hi Bruno On Jan 20, 12:56 pm, Bruno Postle <br...@postle.net> wrote: > On Wed 20-Jan-2010 at 05:22 -0800, Tom Sharpless wrote: > > > > >The numeric boxes for yaw, pitch, roll on the "Move/Drag" tab of the > >fast preview window do not display the current settings for the > >panorama. That should be easy to fix. Maybe they could be changed to > >spin boxes at the same time. > > Spin controls would be good for all these numeric entry boxes. > > Though Hugin doesn't have a concept of view-direction for the > panorama - All panoramas are effectively at 0,0,0 roll/pitch/yaw. > Instead the numeric transform function rotates all the photos in the > scene by the angles you set, as does the Drag functionality.
Thomas Modes already pointed that out to me, I withdraw the suggestion. > > >In the longer term, I wonder if we could add another tab to this > >window that would let one adjust lens parameters interactively. That > >would be a big help when using hugin as a "de-fish" tool, and possibly > >for manual lens calibration. > > Yes, new tabs are likely (I have a wish-list ;-), though potentially > a panorama can have hundreds of different 'lenses' so I'm not sure > how the GUI would work. You can currently play with lens settings > interactively if you have both the Camera and lens tab and Preview > side by side. This idea came from experience with some 'de-fishing' mathmap scripts based on the general Panini projection. I found that to get straight verticals all across the picture I needed to adjust both FOV and an assumed projection curve. Just adjusting FOV takes care of the curvature that is visible in MathMap's little preview, but at full resolution the projection function is really needed too. So what I'd really like on this tab is sliders for hFOV and Curve, plus a listbox with the available input formats (aka lens types). Curve would only be active when the input format designates a lens type, perhaps only for fisheye lenses. Now I know perfectly well that 'Curve' is not a libpano or hugin parameter. To do its work in the present context, that control would have to set the a,b,c values for the radial correction polynomial. Which is feasible, although a better way of representing the radial correction function would make things a lot easier. Why not have sliders that set a, b, c, directly? Because no human being can find the optimum combination, and it is far too easy to set a wildly wrong one (even the optimizer, using hundreds of control points, has a hard time getting it right). If you doubt me, fire up PTLens and try correcting your favorite fisheye. What is needed is something that lets you get close enough without letting you go too far wrong. In my Mathmap scripts, which are limited to fisheye input, Curve constructs the radial function as a linear combination of three prototype functions, that span the range of existing fisheye designs: the stereographic projection at one end, the equal-solid-angle aka equal-area projection at the other end, and the equal-angle projection in the middle. In fact all the curves are weighted sums of the equal- angle and whichever of the other two functions is nearest the slider position. Simple, robust, and it works. Translating that function to a,b,c is a complicated but doable least squares fitting problem. Translating it to a cubic spline table is dead easy, and has the big advantage that it is just as easy to tabulate the inverse function. And Dersch's radial polynomial. And its inverse. And any other radial correction function you care to use. So I would really want to change the way libpano represents the radial function, before implementing this new preview tab. Best, Tom > > -- > Bruno
-- You received this message because you are subscribed to the Google Groups "hugin and other free panoramic software" group. A list of frequently asked questions is available at: http://wiki.panotools.org/Hugin_FAQ To post to this group, send email to hugin-ptx@googlegroups.com To unsubscribe from this group, send email to hugin-ptx+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/hugin-ptx
