Hi Jonathan,
After writing the above, I did manage to get it working for a LensNFunc, but
not for a LensFunc function.
Using the lensNfunction virtual, I can return my own lens function to handle
distortion, and I notice I can test the current projection mode (as set
either in the camera or the scanline render), and return my custom function
only for a certain mode(s).
/*virtual*/ LensNFunc* lensNfunction(int mode) const {
if (mode == LENS_PERSPECTIVE) {
cout << "custom lens func" << endl;
return my_lens_distort;
}
return CameraOp::lensNfunction(mode);
}
But if I try the same with lens_function(), I get no joy. In my case, it
looks like lens_function() is never even called (as in, I don't get its
cout, which is still my lame way of debugging) :)
/*virtual*/ LensFunc* lens_function(int mode) const {
cout << "custom projection func" << endl;
return myLensFunc;
}
In any case, I was only interested in implementing my own LensNFunc, so it's
all good. But I'd be interested in knowing whether lens_function() is meant
to work differently, or if it's just not doing what it should be doing.
Thanks again,
Ivan
On Tue, Jul 5, 2011 at 9:23 AM, Jonathan Egstad <jegs...@earthlink.net>wrote:
> > However, I seem to be stuck at a very basic stage. How do I pass my own
> function to the renderer. Using one of the lens functions you sent as an
> example, I'm trying to get that returned
> > So, taking this as an example:
> >
> > /*! Simple perspective projection lens function. */
> > static void myLensFunc(Scene* scene, CameraOp* cam, MatrixArray*
> transforms,
> > const VArray& in, Vector4& out,
> > void* data) {
> > out = transforms->matrix(LOCAL_TO_SCREEN).transform(in.PL(), 1.0f);
> > }
> >
> >
> > I want lens_function() to return the above for any "mode". Which I'm
> naively trying to do like this:
> >
> > CameraOp::LensFunc* CamTest::lens_function(int mode) const {
> > return &myLensFunc;
> > }
> >
> > However, I'm not getting the result I want. Changing the "projection
> mode" knob in my CamTest Op still behaves as usual, and seems to use the
> right LensFunc for each projection mode, instead of myLensFunc.
> > Am I interpreting this wrong? How should I be returning a pointer to my
> custom function in lens_function()?
>
> The renderer *should* be requesting the lens functions directly from the
> camera pointer that's put into the Scene structure. Don't forget that
> there's a preference in the renderer to override the camera's projection
> setting, however in both cases the lens func should be coming from the
> camera pointer that's stored in the Scene.
>
> Unless something has changed in this logic recently it should work... As I
> said, at DD we had a custom CameraOp that did exactly this - but then again
> the last time I saw it working with my own eyes it was in Nuke version 5.0.
>
> I'll try it out with my custom StereoCam camera when I get a chance in the
> next couple days.
>
> -jonathan
>
>
> > On Mon, Jun 27, 2011 at 1:20 PM, Jonathan Egstad <jegs...@earthlink.net>
> wrote:
> > btw, don't worry about clipping the distorted vertices to the screen
> bbox, the renderer only worries about clipping to the near-Z plane. The
> only time you need to worry about the screen edges is if you're wrapping
> around like the spherical projection does.
> >
> >
> >
> > On Jun 27, 2011, at 11:15 AM, Ivan Busquets wrote:
> >
> >> Wow, thanks for the info, Jonathan. I looked at CameraOp::LensNFunc, but
> didn't know what to make of it, or where to start, really. So that's exactly
> the kind of info I was hoping for.
> >>
> >> At DD there was a custom 'HypeCamera' that supported the 'Hype'
> distortion file format so you could load a Hype file directly into the
> camera and it applied the forward distortion via this method. There was
> also a 'HypeDistort' geometry modifier that warped a card with the reverse
> distortion so you could do a undistort/redistort in a single render/filter
> pass.
> >>
> >> And that's exactly the kind of application I wanted it for. :)
> >>
> >> Thanks,
> >> Ivan
> >>
> >>
> >> On Mon, Jun 27, 2011 at 11:03 AM, Jonathan Egstad <
> jegs...@earthlink.net> wrote:
> >> So the current scanline renderer does handle general camera distortions
> - with poly subdivision memory limitations.
> >>
> >> The CameraOp class supports several built-in linear projection and
> non-linear distortions using its LensFunc/LensNFunc methods that are passed
> back to the renderer via the CameraOp::lens_function()/lensNfunction()
> virtual methods, then on to the Primitives in the Scene structure for use
> during tessellation.
> >>
> >> You can implement your own LensFunc routines and return them via the
> virtual method. The idea is that a primitive can subdivide itself using the
> output of a LensFunc by passing in vertex info which gets modified. The
> 'void *' parameter at the end is a way to get add'l info from the Primitive
> into the lens function.
> >>
> >>
> >> /*! Simple perspective projection lens function. */
> >> static void perspLensFunc(Scene* scene, CameraOp* cam, MatrixArray*
> transforms,
> >> const VArray& in, Vector4& out,
> >> void* data) {
> >> out = transforms->matrix(LOCAL_TO_SCREEN).transform(in.PL(), 1.0f);
> >> }
> >>
> >> /*! Simple UV projection lens function. */
> >> static void perspLensFunc(Scene* scene, CameraOp* cam, MatrixArray*
> transforms,
> >> const VArray& in, Vector4& out,
> >> void* data) {
> >> out = transforms->matrix(CLIP_TO_SCREEN).transform(in.UV()*2.0f -
> Vector4(1,1,1,1));
> >> }
> >>
> >> The LensNFunc method does the same thing, but operates on arrays of vert
> info.
> >>
> >> At DD there was a custom 'HypeCamera' that supported the 'Hype'
> distortion file format so you could load a Hype file directly into the
> camera and it applied the forward distortion via this method. There was
> also a 'HypeDistort' geometry modifier that warped a card with the reverse
> distortion so you could do a undistort/redistort in a single render/filter
> pass.
> >>
> >>
> >> Have fun,
> >>
> >> -jonathan
> >>
> >>
> >> On Jun 26, 2011, at 2:12 PM, Ivan Busquets wrote:
> >>
> >> > Hi,
> >> >
> >> > I've been looking at the lens_distortion parameters in CameraOp, and
> was wondering if there's any way to implement a different lens
> distortion/correction model within a CameraOp, or if one would need to write
> a specific renderer to handle that.
> >> >
> >> > Specifically, I'm looking for a way to handle off-center and
> asymmetric distortions. Assuming I can get a function that maps destination
> to source vectors(uvs), would this be possible within a CameraOp?
> >> >
> >> > Many thanks,
> >> > Ivan
> >> >
> >> >
> >> >
> >> >
> >> >
> >> > _______________________________________________
> >> > Nuke-dev mailing list
> >> > Nuke-dev@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/
> >> > http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-dev
> >>
> >> _______________________________________________
> >> Nuke-dev mailing list
> >> Nuke-dev@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/
> >> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-dev
> >>
> >> _______________________________________________
> >> Nuke-dev mailing list
> >> Nuke-dev@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/
> >> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-dev
> >
> >
> > _______________________________________________
> > Nuke-dev mailing list
> > Nuke-dev@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/
> > http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-dev
> >
> >
> > _______________________________________________
> > Nuke-dev mailing list
> > Nuke-dev@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/
> > http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-dev
>
> _______________________________________________
> Nuke-dev mailing list
> Nuke-dev@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/
> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-dev
>
_______________________________________________
Nuke-dev mailing list
Nuke-dev@support.thefoundry.co.uk, http://forums.thefoundry.co.uk/
http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-dev
