<div><span class="gmail_quote">Here's a strategy to try:<br><br>1. Make a 2D Image of size { ImageSize1D , 4 }, where ImageSize is the length of your 1D Vector image.<br>2. Copy the 1D vector into the image. You can do this with simple itk:Iterators, because your vector image fits into the first row of of the 2D image.
<br>3. Do the registration in 2D. I haven't tried this but if it works at all, it should generate a registration that is valid for the 1D vector image.<br>4. It's probably best do do all processing in terms of the 2D image, and then copy out the 1D vector image at the end of the pipeline.
<br><br>Steps 1 and 4 can probably be done with itkFilters, like itkTileImageFilter, itkCastImageFilter, and itkExtractImage filter. But using iterators to do the copy is about the same in terms of lines of code, and can be written as a short template function.
<br><br><br>On 2/12/07, <b class="gmail_sendername">Jochen Bertels</b> <<a href="mailto:jochen.bertels@ua.ac.be">jochen.bertels@ua.ac.be</a>> wrote:</span><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
Hi,<br><br>I'm trying to non-rigidly register two one-dimensional signals.<br><br>My first try was to adapt the DeformableRegistration4 example, by just<br>changing the dimension.<br>This wouldn't compile because the GradientRecursiveGaussianImageFilter,
<br>which is used in every<br>imageToImageMetric to calculate the gradient of the metric, can't cope<br>with 1D images.<br><br>This made me try to mimic the 1D case by using a 2D image with the size<br>in one dimension set tot
<br>1. This results in an error saying that the<br>GradientRecursiveGaussianImageFilter requires at least 4<br>pixels in each dimension.<br><br>So the only options left, as far as i can see, are<br>1) use an optimizer that doesn't use gradient information.
<br>2) write my own metric, calculating the gradient another way than with<br>GradientRecursiveGaussianImageFilter<br><br>I have tried the first option by adapting DeformableRegistration2. This<br>works but is rather slow (12 minutes
<br>for a 2638x1 image).<br><br>Before trying the second option, i would like to ask if there are other<br>options that i haven't thought of.<br>I find it rather hard to believe i am the first person facing this<br>problem, although i haven't found any other
<br>postings on this topic.<br><br>Thanks in advance,<br><br>Jochen<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></div><br>