No subject


Mon Dec 15 09:51:36 EST 2008


compares the true transform parameters not to exceed thresholds --<br>
0.025 for scale and 1.0 for translations.</blockquote><div><br>But as Dan m=
entioned its hard for most transforms. The right way to go about it is gene=
rate the target images manually with a known transform. Then after registra=
tion, you&#39;d multiply the obtained transform with the known transform&#3=
9;s inverse and check the L2 norm / bending energy or its equivalent agains=
t a tolerance.<br>
<br>Given the work it requires, its simpler to just compute the RMSE over s=
ource/target images.<br><br></div><blockquote class=3D"gmail_quote" style=
=3D"border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; p=
adding-left: 1ex;">
<br><div><div class=3D"Wj3C7c">
On Wed, Jan 28, 2009 at 11:30 AM, Bill Lorensen &lt;<a href=3D"mailto:bill.=
lorensen at gmail.com">bill.lorensen at gmail.com</a>&gt; wrote:<br>
&gt; Also, their can be multiple baselines for a given test. If you look at=
:<br>
&gt;<br>
&gt; <a href=3D"http://public.kitware.com/cgi-bin/viewcvs.cgi/Testing/Data/=
Baseline/Registration/?root=3DInsight" target=3D"_blank">http://public.kitw=
are.com/cgi-bin/viewcvs.cgi/Testing/Data/Baseline/Registration/?root=3DInsi=
ght</a><br>

&gt;<br>
&gt; you will see that some tests have multiple baselines because of the<br=
>
&gt; type of variability discussed in this thread.<br>
&gt;<br>
&gt; For example, ImageRegistration13Test has 5 baselines.<br>
&gt;<br>
&gt; Bill<br>
&gt;<br>
&gt; On Wed, Jan 28, 2009 at 11:14 AM, Blezek, Daniel J.<br>
&gt; &lt;<a href=3D"mailto:Blezek.Daniel at mayo.edu">Blezek.Daniel at mayo.edu</=
a>&gt; wrote:<br>
&gt;&gt; Just 2 cents here.<br>
&gt;&gt;<br>
&gt;&gt; 1) For a linear registration, you could expect the registration re=
sults<br>
&gt;&gt; to be ~0.1 degrees and ~0.1 mm cross platform. &nbsp;The major sou=
rce of this<br>
&gt;&gt; problem is differences in floating point representations under dif=
ferent<br>
&gt;&gt; compilers/hardware. &nbsp;I shouldn&#39;t worry about this, it&#39=
;d be difficult for<br>
&gt;&gt; a human to see. &nbsp;For something like BSplines, you might have =
a bit more<br>
&gt;&gt; error.<br>
&gt;&gt;<br>
&gt;&gt; 2) A very simple way to do this is to do a forward transformation =
using<br>
&gt;&gt; one transform (ground truth), then the inverse transform of the ne=
wly<br>
&gt;&gt; calculated transform. &nbsp;I think this is called Target Registra=
tion Error<br>
&gt;&gt; by Fitzpatrick and Co., but you should look it up. &nbsp;This is w=
here you<br>
&gt;&gt; need to decide on your tolerance. &nbsp;As Karthik mentioned, only=
 simple ITK<br>
&gt;&gt; transforms have an inverse, which is really a shame. &nbsp;They ar=
e so<br>
&gt;&gt; useful, even if they are numeric.<br>
&gt;&gt;<br>
&gt;&gt; 2a) I suppose you could forward transform the same point using two=
<br>
&gt;&gt; different transforms and see how far apart they are. &nbsp;This se=
ems<br>
&gt;&gt; reasonable, but you&#39;d have to sample a bunch of points to acco=
unt for<br>
&gt;&gt; rotation, and transform centers, etc. &nbsp;And you&#39;d only get=
 a distance<br>
&gt;&gt; measure, not a rotational measure.<br>
&gt;&gt;<br>
&gt;&gt; For transforms with an inverse, you can do what you are asking, an=
d it<br>
&gt;&gt; would be a valuable contribution to ITK, but it&#39;s not general,=
 as not<br>
&gt;&gt; all transforms support an inverse. &nbsp;And you could always test=
 the<br>
&gt;&gt; transform you care about...<br>
&gt;&gt;<br>
&gt;&gt; Incoherent as usual,<br>
&gt;&gt; -dan<br>
&gt;&gt;<br>
&gt;&gt; -----Original Message-----<br>
&gt;&gt; From: <a href=3D"mailto:insight-users-bounces at itk.org">insight-use=
rs-bounces at itk.org</a><br>
&gt;&gt; [mailto:<a href=3D"mailto:insight-users-bounces at itk.org">insight-u=
sers-bounces at itk.org</a>] On Behalf Of Andriy Fedorov<br>
&gt;&gt; Sent: Tuesday, January 27, 2009 6:18 PM<br>
&gt;&gt; To: ITK Users<br>
&gt;&gt; Cc: Miller, James V (GE, Research)<br>
&gt;&gt; Subject: [Insight-users] Rigid registration testing and reproducib=
ility<br>
&gt;&gt;<br>
&gt;&gt; Hi,<br>
&gt;&gt;<br>
&gt;&gt; I would like to do regression tests of rigid registration with rea=
l<br>
&gt;&gt; images, and compare the result with the baseline transform. Here a=
re two<br>
&gt;&gt; related questions.<br>
&gt;&gt;<br>
&gt;&gt; 1) ITK registration experts, could you speculate on what is the no=
rmal<br>
&gt;&gt; variability in the rigid registration results run with the same<br=
>
&gt;&gt; parameters, metric initialized with the same seed, when run across=
<br>
&gt;&gt; different platforms? What could be the sources of this variability=
,<br>
&gt;&gt; given the same initialization and same parameters?<br>
&gt;&gt;<br>
&gt;&gt; 2) If I understand correctly, the current testing of registration =
that<br>
&gt;&gt; comes with ITK generates a synthetic image with known transform<br=
>
&gt;&gt; parameters, which are compared with the registration-derived trans=
forms.<br>
&gt;&gt;<br>
&gt;&gt; The testing framework implemented in itkTestMain.h allows to compa=
re<br>
&gt;&gt; pixelwise two images, but this does not seem to be practical for<b=
r>
&gt;&gt; registration regression testing, because of the variability in the=
<br>
&gt;&gt; registration results I mentioned earlier. Accounting for this<br>
&gt;&gt; variability using tolerance values of just 1 pixel hugely increase=
s the<br>
&gt;&gt; test runtime, but in my experience, the comparison may still fail.=
<br>
&gt;&gt;<br>
&gt;&gt; Would it not make sense to augment itkTestMain.h with the capabili=
ty to<br>
&gt;&gt; compare not only images, but transforms? Is this a valid feature<b=
r>
&gt;&gt; request, or am I missing something?<br>
&gt;&gt;<br>
&gt;&gt; Thanks<br>
&gt;&gt;<br>
&gt;&gt; Andriy Fedorov<br>
&gt;&gt; _______________________________________________<br>
&gt;&gt; Insight-users mailing list<br>
&gt;&gt; <a href=3D"mailto:Insight-users at itk.org">Insight-users at itk.org</a>=
<br>
&gt;&gt; <a href=3D"http://www.itk.org/mailman/listinfo/insight-users" targ=
et=3D"_blank">http://www.itk.org/mailman/listinfo/insight-users</a><br>
&gt;&gt; _______________________________________________<br>
&gt;&gt; Insight-users mailing list<br>
&gt;&gt; <a href=3D"mailto:Insight-users at itk.org">Insight-users at itk.org</a>=
<br>
&gt;&gt; <a href=3D"http://www.itk.org/mailman/listinfo/insight-users" targ=
et=3D"_blank">http://www.itk.org/mailman/listinfo/insight-users</a><br>
&gt;&gt;<br>
&gt;<br>
_______________________________________________<br>
Insight-users mailing list<br>
<a href=3D"mailto:Insight-users at itk.org">Insight-users at itk.org</a><br>
<a href=3D"http://www.itk.org/mailman/listinfo/insight-users" target=3D"_bl=
ank">http://www.itk.org/mailman/listinfo/insight-users</a><br>
</div></div></blockquote></div><br><br clear=3D"all"><br>-- <br>Karthik Kri=
shnan<br>R&amp;D Engineer,<br>Kitware Inc.<br>Ph: 518 371 3971 x119<br>Fax:=
 518 371 3971<br>

--001485f7c1aa1d3eb204618e9038--


More information about the Insight-users mailing list