On Jun 12, 8:54 am, iiuu <duangn...@[EMAIL PROTECTED]
> wrote:
> Dear all,
>
> I am wondering that how come there are many solutions to
> the Cornerness measurement on Harris corner dectection.
>
> The novel one is:
> C= det(M)-k*trace(M).^2
>
> right?
>
> but there are those in:
> 1) C= trace(M)./det(M)
> 2) C= det(M)./(trace(M)+eps)
> 3) C= (k+1/k)*det(M)-(trace(M).^2-2det(M))
Your forgot min(l1, l2).
> Does anyone have any theories sup****t these latter three
> equations? It seems that some of them also concentrate on
> edges plus corners but I really can't find any do***ents
> sup****ted these equations.
>
> Please help me out. Thank you so much.
Your forgot min(l1, l2). Harris noted that this one is what he eanted,
but was inefficient on the hardware of the day, because of the sqrt()
needed at every pixel. People have done various chunks of theory on
these, involving things like localization, what it means in terms of
surface curvature, and etc. I have a paper with information about some
of this here:
http://mi.eng.cam.ac.uk/~er258/work/rosten_2006_machine.pdf
Look in the "previous work" section. The results section also shows
little practical difference between Harris' approximation and the
complete computation (the ****-Tomasi detector), in my tests.
-Ed
--
(You can't go wrong with psycho-rats.)(http://mi.eng.cam.ac.uk/~er258)
/d{def}def/f{/Times s selectfont}d/s{11}d/r{roll}d f 2/m{moveto}d -1
r 230 350 m 0 1 179{ 1 index show 88 rotate 4 mul 0 rmoveto}for/s 12
d f pop 235 420 translate 0 0 moveto 1 2 scale show showpage
|
