<br><br>
<div><span class="gmail_quote">
<div style="DIRECTION: ltr">
<div>Siddharth,</div>
<div> </div>
<div>In the past the couple I have tried when I go from level to level is to</div>
<div> - increase the bspline grid (i.e. start with a coarse grid and refine at each level)</div>
<div> - decrease the <font size="2">CostFunctionConvergenceFactor (i.e. make the optimizer convergence more stringent at each level)</font></div>
<div> </div>
<div>HTH,</div></div>
<div style="DIRECTION: ltr"><span class="sg">
<div>Lydia</div>
<div> </div></span></div></span></div>
<div><span class="gmail_quote">On 5/3/06, <b class="gmail_sendername">Siddharth Vikal</b> <<a href="mailto:siddharthvikal@gmail.com">siddharthvikal@gmail.com</a>> wrote:</span></div>
<blockquote class="gmail_quote" style="PADDING-LEFT: 1ex; MARGIN: 0px 0px 0px 0.8ex; BORDER-LEFT: #ccc 1px solid">Hello,<br><br>I'm trying to perform deformable registration of two MRI volume<br>datasets using Bsplines deformation model.
<br><br>I went through really nicely explained examples of registration. There<br>are examples on bspline based registration and multi-res framework<br>based registration independently. In Bspline example, the optimizer<br>
used is LBFGSB.<br><br>So, I was wondering what parameters of this optimizer could be changed<br>at different resolution levels of registration? (considering in<br>MultiRes example, the max and min. step length of<br>RegularStepGradientDescent optimizer are modified analogously).
<br><br>Looking forward for advices.<br><br>regards<br>siddharth<br>_______________________________________________<br>Insight-users mailing list<br><a href="mailto:Insight-users@itk.org">Insight-users@itk.org</a><br><a href="http://www.itk.org/mailman/listinfo/insight-users">
http://www.itk.org/mailman/listinfo/insight-users</a><br></blockquote><br>