SDM on WINE
- --- In StereoDataMaker@yahoogroups.com, Vladimir Nadvornik <nadvornik@...> wrote:
> Maybe it will be more clear from the image here:
> First camera is at [X,Y,Z] = [0,0,0]
> Second camera is at [X,Y,Z] = [stereo_base,0,0]
> The "Projection plane" at (x,y,-1) is the Stereo Window plane.
> This means that the stereo_base is measured relative to Stereo Window
So they get a sort of "Stereo Distance" in pixels (the units the stereo window are), which is worked out from the distortions found in the photo-mosaic. But I think this is not exactly what most stereo photographers understand by "Stereo Distance".
In fact the explanation of the photo-mosaic warns that this paradigm would only work for mosaics of flat objects, like a flat wall mural.
>"Toe-in" is the same as "yaw". It happens, for example, if the camera at (0,0,0) has the optical axis normal to the image plane, but the axis of the camera at (X,0,0) is not parallel, but converges with an angle towards the other camera. When this happens one camera may "see" a rectangular shape, while the other will "see" a trapezoid with the side edges of different height. This is usually bad for stereo pairs because it involves Vertical Disparity if not corrected properly.
> The yaw, pitch, roll is the direction angle of the camera - it means the
> horizontal and vertical displacement and rotation of the image. You have to
> optimize them unless you are sure the cameras are precisely parallel.
> As I understand your description of the "toe" transformation, I think that the
> "toe-in-out" transformation is determined by camera X position (stereo_base)
> and the "toe-up-down" by Y position (zero in my model), so optimizing X
> parameter means detemining the "toe-in-out".
I notice the "Projection Plane" concept of Hugin corrects this distortion (or at least force the same distortion for both cameras:-). It looks like an useful concept, I am probably getting something similar to that, though I tackle the effects of each variable one by one.
>So "Rectilinear" means just the geometrical projection that is shown in the photo-mosaic figure. I recall Hugin has some very weird projection modes which I never really understood. At least now I understand one of them:-)
> I forgot to mention that the output projection is set to "Rectilinear", this
> means that the output image is equivalent to the projection plane.
P.S. I will be out there for one week (must go to a nice place to take stereo photos:-).