[Bf-vfx] Plane Tracking naming

Sergey Sharybin sergey.vfx at gmail.com
Tue Aug 13 19:04:04 CEST 2013


On Tue, Aug 13, 2013 at 9:34 PM, David Jeske <davidj at gmail.com> wrote:

> On Tue, Aug 13, 2013 at 12:36 AM, Sergey Sharybin <sergey.vfx at gmail.com>wrote:
>
>> I do not happy with calling it "corner pin" because it's not  corner pin
>> at all. You could (and probably even should) use more than 4 point tracks
>> to make plane estimation much more accurate. Also, point tracks have no
>> relation with corners at all: you could track points which are not corners
>> of your plane (we showed this in our video).
>>
>
> The feature you are calling "plane track" appears *exactly* like corner
> pinning in Mocha and After Effects. The "corners" referred to in the name
> "corner track" or "corner perspective pin" are the corners of your "plane"
> not the trackers themselves.
>
> http://www.youtube.com/watch?v=xRpHgVFPqqk&feature=player_detailpage&t=465
> http://www.youtube.com/watch?v=fY4wwZFD5-c
>

So is it a question like "let's call feature XXX exactly the same as it's
called in software YYY"?


> The reason it's not called planar-tracking is that it's not 3d, it's a 2d
> affine image transformation.
>

It's not affine, it's homography if it makes sense.

>
> Why do you call it a "plane"? As far as I can see, it's not a 3d-plane,
> but a 2d-affine compositing surface distorted in camera-space.
>

because it's a plane. You could think of it as a real non-concave plane
being warped by a homography estimated from point tracks. Which is like if
you watch on a viewplane on which that plane is being projected.


>
>
>> Another usage of plane track might be constraining point tracks to belong
>> to this plane while tracking them. Which means we might support
>> other-way-around usecase: you create point tracks, you create plane out of
>> them, and then starts tracking this point tracks taking plane constraint
>> into account.
>>
>
> At that point, the track would be for a 3d-plane, and it would be
> appropriate to call it a planar-track.
>

That wouldn't be 3d, that'd still be based on homography estimation and
making so tracks fits this homography in a best way (well, difficult to
explain this in text, whiteboard would be much easier). At this point i
couldn't see why or how we'll reconstruct 3D from point tacks in that case.



-- 
With best regards, Sergey Sharybin
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.blender.org/pipermail/bf-vfx/attachments/20130813/3e0ad8b9/attachment.htm 


More information about the Bf-vfx mailing list