<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">
<HTML><HEAD>
<META http-equiv=Content-Type content="text/html; charset=iso-8859-1">
<META content="MSHTML 6.00.2900.2802" name=GENERATOR>
<STYLE></STYLE>
</HEAD>
<BODY bgColor=#ffffff>
<DIV><FONT face=Arial size=2></FONT> </DIV>
<DIV><FONT face=Arial size=2>I forgot to mention that I always have the -O3
optimization flag ON, while compiling.</FONT></DIV>
<DIV><FONT face=Arial size=2></FONT> </DIV>
<DIV><FONT face=Arial size=2>Regards,</FONT></DIV>
<DIV><FONT face=Arial size=2>Suyash.</FONT></DIV>
<BLOCKQUOTE dir=ltr
style="PADDING-RIGHT: 0px; PADDING-LEFT: 5px; MARGIN-LEFT: 5px; BORDER-LEFT: #000000 2px solid; MARGIN-RIGHT: 0px">
<DIV style="FONT: 10pt arial">----- Original Message ----- </DIV>
<DIV
style="BACKGROUND: #e4e4e4; FONT: 10pt arial; font-color: black"><B>From:</B>
<A title=suyash@cs.utah.edu href="mailto:suyash@cs.utah.edu">Suyash P.
Awate</A> </DIV>
<DIV style="FONT: 10pt arial"><B>To:</B> <A title=millerjv@crd.ge.com
href="mailto:millerjv@crd.ge.com">Miller, James V (Research)</A> ; <A
title=jiafucang@asisz.com href="mailto:jiafucang@asisz.com">jiafucang</A> ; <A
title=jjomier@cs.unc.edu
href="mailto:jjomier@cs.unc.edu">jjomier@cs.unc.edu</A> ; <A
title=insight-users@itk.org
href="mailto:insight-users@itk.org">insight-users@itk.org</A> </DIV>
<DIV style="FONT: 10pt arial"><B>Sent:</B> Monday, December 19, 2005 8:50
PM</DIV>
<DIV style="FONT: 10pt arial"><B>Subject:</B> Re: [Insight-users] Re: On
incorrect matrix inverse</DIV>
<DIV><BR></DIV>
<DIV><FONT face=Arial size=2>Dear ITK Users,</FONT></DIV>
<DIV><FONT face=Arial size=2></FONT> </DIV>
<DIV><FONT face=Arial size=2>I found the cause of the error. It deals with my
build of ITK in the SUSE 10 environment that uses gcc 4.0 by
default. When I build itk and the applications on SUSE 9.3 using gcc 3.4
(definitely not gcc 4.0) or Fedora Core 2 (again not using gcc 4.0), I
did NOT get any vnl error.</FONT></DIV>
<DIV><FONT face=Arial size=2></FONT> </DIV>
<DIV><FONT face=Arial size=2>I confirmed that precision is NOT an issue
because the condition number for the matrix with truncated values is still
very close to 1. I found this via matlab. So numerical "instabilities" in
inverse-finding should not occur.</FONT></DIV>
<DIV><FONT face=Arial size=2></FONT> </DIV>
<DIV><FONT face=Arial size=2>So it seems that there is something strange about
the way the vnl code/libraries interact with gcc 4.0.</FONT></DIV>
<DIV><FONT face=Arial size=2></FONT> </DIV>
<DIV><FONT face=Arial size=2>Thanks again for all your help.</FONT></DIV>
<DIV><FONT face=Arial size=2></FONT> </DIV>
<DIV><FONT face=Arial size=2>Best regards,</FONT></DIV>
<DIV><FONT face=Arial size=2>Suyash.</FONT></DIV>
<BLOCKQUOTE dir=ltr
style="PADDING-RIGHT: 0px; PADDING-LEFT: 5px; MARGIN-LEFT: 5px; BORDER-LEFT: #000000 2px solid; MARGIN-RIGHT: 0px">
<DIV style="FONT: 10pt arial">----- Original Message ----- </DIV>
<DIV
style="BACKGROUND: #e4e4e4; FONT: 10pt arial; font-color: black"><B>From:</B>
<A title=millerjv@crd.ge.com href="mailto:millerjv@crd.ge.com">Miller, James
V (Research)</A> </DIV>
<DIV style="FONT: 10pt arial"><B>To:</B> <A title=suyash@cs.utah.edu
href="mailto:suyash@cs.utah.edu">Suyash P. Awate</A> ; <A
title=jiafucang@asisz.com href="mailto:jiafucang@asisz.com">jiafucang</A> ;
<A title=jjomier@cs.unc.edu
href="mailto:jjomier@cs.unc.edu">jjomier@cs.unc.edu</A> ; <A
title=insight-users@itk.org
href="mailto:insight-users@itk.org">insight-users@itk.org</A> </DIV>
<DIV style="FONT: 10pt arial"><B>Sent:</B> Friday, November 11, 2005 11:24
AM</DIV>
<DIV style="FONT: 10pt arial"><B>Subject:</B> RE: [Insight-users] Re: On
incorrect matrix inverse</DIV>
<DIV><BR></DIV>
<DIV><SPAN class=675082118-11112005><FONT face=Arial color=#0000ff
size=2>Suyash, </FONT></SPAN></DIV>
<DIV><SPAN class=675082118-11112005><FONT face=Arial color=#0000ff
size=2></FONT></SPAN> </DIV>
<DIV><SPAN class=675082118-11112005><FONT face=Arial color=#0000ff
size=2>You might not have enough precision in the matrix values that you are
reading from the command line. Try generating the values with
more</FONT></SPAN></DIV>
<DIV><SPAN class=675082118-11112005><FONT face=Arial color=#0000ff
size=2>precision from whatever tool is generating them. Try going out
to about 12 decimal places.</FONT></SPAN></DIV>
<DIV><SPAN class=675082118-11112005><FONT face=Arial color=#0000ff
size=2></FONT></SPAN> </DIV>
<DIV><SPAN class=675082118-11112005><FONT face=Arial color=#0000ff size=2>I
have seen this before where one program computes a matrix that is supposed
to be invertible, writes the matrix in ascii, another program reads that
matrix and it is no longer invertible. </FONT></SPAN></DIV>
<DIV><SPAN class=675082118-11112005><FONT face=Arial color=#0000ff
size=2></FONT></SPAN> </DIV>
<DIV><SPAN class=675082118-11112005><FONT face=Arial color=#0000ff
size=2>Jim</FONT></SPAN></DIV>
<DIV><SPAN class=675082118-11112005><FONT face=Arial color=#0000ff
size=2></FONT></SPAN> </DIV>
<DIV><SPAN class=675082118-11112005><FONT face=Arial color=#0000ff
size=2></FONT></SPAN> </DIV>
<BLOCKQUOTE dir=ltr style="MARGIN-RIGHT: 0px">
<DIV class=OutlookMessageHeader dir=ltr align=left><FONT face=Tahoma
size=2>-----Original Message-----<BR><B>From:</B> <A
href="mailto:insight-users-bounces+millerjv=crd.ge.com@itk.org">insight-users-bounces+millerjv=crd.ge.com@itk.org</A>
[mailto:insight-users-bounces+millerjv=crd.ge.com@itk.org]<B>On Behalf Of
</B>Suyash P. Awate<BR><B>Sent:</B> Friday, November 11, 2005 1:06
PM<BR><B>To:</B> jiafucang; jjomier@cs.unc.edu;
insight-users@itk.org<BR><B>Subject:</B> [Insight-users] Re: On incorrect
matrix inverse<BR><BR></FONT></DIV>
<DIV><FONT face=Arial></FONT> </DIV>
<DIV><FONT face=Arial>Hi everybody,</FONT></DIV>
<DIV><FONT face=Arial></FONT> </DIV>
<DIV><FONT face=Arial>Thanks for the generous help.</FONT></DIV>
<DIV><FONT face=Arial></FONT> </DIV>
<DIV><FONT face=Arial>I was also trying to get the inverse concerning
with the registration application, like Fucang (see email
attached).</FONT></DIV>
<DIV><FONT face=Arial></FONT> </DIV>
<DIV><FONT face=Arial>Here is my piece of code that gave me the wrong
result:</FONT></DIV>
<DIV><FONT face=Arial></FONT> </DIV>
<DIV><FONT face=Arial>------------</FONT></DIV>
<DIV><FONT face=Arial>const unsigned int Dimension = 3;<BR>typedef double
PixelType;<BR>typedef itk::Matrix <PixelType, Dimension, Dimension>
MatrixType;<BR>MatrixType matrix;<BR> matrix.Fill (0);<BR>
matrix (0,0)= atof (argv[4]);<BR> matrix (0,1)= atof
(argv[5]);<BR> matrix (0,2)= atof (argv[6]);<BR> matrix (1,0)=
atof (argv[7]);<BR> matrix (1,1)= atof (argv[8]);<BR> matrix
(1,2)= atof (argv[9]);<BR> matrix (2,0)= atof (argv[10]);<BR>
matrix (2,1)= atof (argv[11]);<BR> matrix (2,2)= atof
(argv[12]);<BR>vnl_matrix_fixed <PixelType, Dimension, Dimension>
inverse;<BR>inverse= matrix.GetInverse();</DIV></FONT>
<DIV><FONT face=Arial>-----------</FONT></DIV>
<DIV><FONT face="Courier New">matrix<BR> 0.99925
0.0371382 -0.0109982<BR>-0.0373772 0.99905
-0.0223916<BR> 0.0101562 0.0227859
0.999689<BR><BR>warnings:<BR>InsightToolkit-2.2.0/Utilities/vxl/core/vnl/algo/vnl_svd.txx:
suspicious return value (2) from
SVDC<BR>InsightToolkit-2.2.0/Utilities/vxl/core/vnl/algo/vnl_svd.txx: M is
3x3<BR>M = [ ...<BR> 0.9992500000000 0.0371382000000
-0.0109982000000<BR>-0.0373772000000 0.9990500000000
-0.0223916000000<BR> 0.0101562000000 0.0227859000000
0.9996890000000 ]<BR><BR></FONT><TT>matrix inverse using vnl in itk
(incorrect)<BR>-0.982557 -0.0644525 -0.174434<BR>-0.0677511
0.997618 0.0130162<BR>-0.17318 -0.0246073
0.984583<BR><BR></TT><TT>matrix inverse in matlab
(correct)<BR> 0.9992 -0.0374
0.0102<BR> 0.0371 0.9991
0.0228<BR>-0.0110 -0.0224
0.9997<BR>------------</TT></DIV>
<DIV><TT></TT> </DIV>
<DIV><FONT face=Arial>I had also tried using the AffineTransform but for
some reason could not get that to work and so I wrote this simple matrix
code. Perhaps that was due to the same problem that Fucang
reported.</FONT></DIV>
<DIV><FONT face=Arial>Currently, however, my problem is different. As we
can see I am not using any Transform class at all. Moreover I am getting
warnings from the vnl_svd class, so vnl is definitely involved
here.</FONT></DIV>
<DIV><FONT face=Arial></FONT> </DIV>
<DIV><FONT face=Arial>Julien: I am giving 9 numbers (truncated: <FONT
face="Courier New">0.99925 0.0371382 -0.0109982 -0.0373772
0.99905 -0.0223916 0.0101562 0.0227859 0.999689) to the matrix
in itk and also the same numbers to the matrix in matlab. I got these
numbers from the output of a separate program.</FONT><BR></FONT><FONT
face=Arial><FONT face=Arial>Just FYI, I have compiled ITK on Suse 10.0
using GCC 4.0. I wonder if that has anything to do with this
!</FONT></DIV></FONT>
<DIV><FONT face=Arial></FONT> </DIV>
<DIV><FONT face=Arial>Best regards,</FONT></DIV>
<DIV><FONT face=Arial>Suyash.</FONT></DIV>
<DIV><FONT face=Arial></FONT> </DIV>
<BLOCKQUOTE dir=ltr
style="PADDING-RIGHT: 0px; PADDING-LEFT: 5px; MARGIN-LEFT: 5px; BORDER-LEFT: #000000 2px solid; MARGIN-RIGHT: 0px">
<DIV style="FONT: 10pt arial"><FONT size=3></FONT> </DIV>
<DIV style="FONT: 10pt arial"><FONT size=3>----- Original Message -----
</FONT></DIV>
<DIV
style="BACKGROUND: #e4e4e4; FONT: 10pt arial; font-color: black"><FONT
size=3><B>From:</B> </FONT><A title=jiafucang@asisz.com
href="mailto:jiafucang@asisz.com"><FONT size=3>jiafucang</FONT></A><FONT
size=3> </FONT></DIV>
<DIV style="FONT: 10pt arial"><FONT size=3><B>To:</B> </FONT><A
title=suyash@cs.utah.edu href="mailto:suyash@cs.utah.edu"><FONT
size=3>suyash@cs.utah.edu</FONT></A><FONT size=3> ; </FONT><A
title=jjomier@cs.unc.edu href="mailto:jjomier@cs.unc.edu"><FONT
size=3>jjomier@cs.unc.edu</FONT></A><FONT size=3> </FONT></DIV>
<DIV style="FONT: 10pt arial"><FONT size=3><B>Sent:</B> Friday, November
11, 2005 7:10 AM</FONT></DIV>
<DIV style="FONT: 10pt arial"><FONT size=3><B>Subject:</B> On incorrect
matrix inverse</FONT></DIV>
<DIV><FONT face=Arial></FONT><BR></DIV>
<DIV
style="FONT-SIZE: 10pt; FONT-FAMILY: tahoma; BACKGROUND-COLOR: white"><FONT
size=3>Hello, Suyash, Jomier,</FONT></DIV>
<DIV
style="FONT-SIZE: 10pt; FONT-FAMILY: tahoma; BACKGROUND-COLOR: white"><FONT
size=3></FONT> </DIV>
<DIV
style="FONT-SIZE: 10pt; FONT-FAMILY: tahoma; BACKGROUND-COLOR: white"><FONT
size=3>I have reported similar problem on</FONT></DIV>
<DIV
style="FONT-SIZE: 10pt; FONT-FAMILY: tahoma; BACKGROUND-COLOR: white"><A
href="http://public.kitware.com/pipermail/insight-users/2005-October/015441.html"><FONT
size=3>http://public.kitware.com/pipermail/insight-users/2005-October/015441.html</FONT></A><FONT
size=3> </FONT></DIV>
<DIV
style="FONT-SIZE: 10pt; FONT-FAMILY: tahoma; BACKGROUND-COLOR: white"><FONT
size=3></FONT> </DIV>
<DIV
style="FONT-SIZE: 10pt; FONT-FAMILY: tahoma; BACKGROUND-COLOR: white"><FONT
size=3>and added a bug 2450</FONT></DIV>
<DIV
style="FONT-SIZE: 10pt; FONT-FAMILY: tahoma; BACKGROUND-COLOR: white"><FONT
size=3></FONT> </DIV>
<DIV
style="FONT-SIZE: 10pt; FONT-FAMILY: tahoma; BACKGROUND-COLOR: white"><A
href="http://www.itk.org/Bug/bug.php?op=show&bugid=2450&pos=24"><FONT
size=3>http://www.itk.org/Bug/bug.php?op=show&bugid=2450&pos=24</FONT></A></DIV>
<DIV
style="FONT-SIZE: 10pt; FONT-FAMILY: tahoma; BACKGROUND-COLOR: white"><FONT
size=3></FONT> </DIV>
<DIV
style="FONT-SIZE: 10pt; FONT-FAMILY: tahoma; BACKGROUND-COLOR: white"><FONT
size=3>I think it is because ITK does not computer inverse when
setparameters used, and not VNL problem.</FONT></DIV>
<DIV
style="FONT-SIZE: 10pt; FONT-FAMILY: tahoma; BACKGROUND-COLOR: white"><FONT
size=3></FONT> </DIV>
<DIV
style="FONT-SIZE: 10pt; FONT-FAMILY: tahoma; BACKGROUND-COLOR: white"><FONT
size=3>HTH</FONT></DIV>
<DIV
style="FONT-SIZE: 10pt; FONT-FAMILY: tahoma; BACKGROUND-COLOR: white"><FONT
size=3></FONT> </DIV>
<DIV
style="FONT-SIZE: 10pt; FONT-FAMILY: tahoma; BACKGROUND-COLOR: white"><FONT
size=3>Fucang</FONT></DIV></BLOCKQUOTE></BLOCKQUOTE></BLOCKQUOTE></BLOCKQUOTE></BODY></HTML>