Comments on: Homography

By: Mark Lundin: Musings and things I've stolen

Mark Lundin: Musings and things I've stolen — Thu, 02 Dec 2010 11:47:37 +0000

[...] homography if you already know a few properties about the projections before hand. As noted by both nicoptere and makc, there’s a special relationship between the 4 corners of a rectangle, that mean if [...]

By: Alan Shaw

Alan Shaw — Sun, 24 May 2009 16:40:25 +0000

All I'm saying is, is this not the same general 2D warping problem I addressed in AstroMorpher? The Distort class was created to get around the pre-Flash 10 lack of built-in perspective transformation (using the Matrix class you are limited to affine transformations only). So Distort cuts up the polygon into small triangles and thus minimizes the mapping error. But in Flash 10 there is drawTriangles() which can do the mapping even on non-affine transformations.

By: nicoptere

nicoptere — Sat, 16 May 2009 18:53:08 +0000

hi all, thanks for the comments.

C4RLO5 > thanks :)
makc3d > not quite half the job :)
recognizing the fudicial pattern is important yet the trickiest part is to find the invert projection matrix for the given points.

Alan > indeed the FP10 drawing methods would be both cleaner and faster.
I haven’t started working with it yet … shame on me.
but maybe you were talking about th algorithm itself in which case I have to admit that I didn’t understand what you were saying.

By: Alan Shaw

Alan Shaw — Sat, 16 May 2009 15:28:08 +0000

Great but would it not be cleaner using flash 10 warping? I believe that the Distort class (essential up to now) is superseded by flash 10.

By: makc3d

makc3d — Tue, 21 Apr 2009 05:54:10 +0000

indeed this is half of what FLARToolkit does

By: C4RL05

C4RL05 — Tue, 14 Apr 2009 10:51:22 +0000

Thanks so much, this can be very useful for pattern recognition. Great blog, lots of interesting stuff.