[Insight-users] Similarity Transform in 2D

[Maureen Haining]

On an even simpler note ... I have two black and white boundary images, the
second of which is simply translated by (30, 15) pixels. Using the
Translation Transform (ImageRegistration1 example) I would expect to get a
very close match. Instead the images are still misaligned quite a bit. The
results show a translation of (-2, -5) which I believe is in spatial
measurements correct? My images have unit spacing so I would have expected
(30, 15) no?
Again, any help is appreciated.
Maureen

On Fri, Jul 25, 2008 at 12:37 PM, Maureen Haining <mfhaining at gmail.com>
wrote:

> Hello Users,
> I have two images that are essentially aligned that I am using to
> experiment with the Similarity Transform with Regular Step Gradient Descent
> Optimizer (I am using example 7). The fixed image is a traced black and
> white boundary image of moving image. I am having trouble understanding the
> results.
>
> Result =
>
>  Scale         = 4.15432
>
>  Angle (radians) 0.00397598
>
>  Angle (degrees) 0.227807
>
>  Center X      = 374.818
>
>  Center Y      = 361.912
>
>  Translation X = -0.171362
>
>  Translation Y = -0.684359
>
>  Iterations    = 500
>
>  Metric value  = 1337.12
>
>
> I am surprised to see a scaling factor introduced. I have set the
> optimizerScales to 1 (scale)  and 1 (angle). And indeed the moving image is
> scaled down by about a factor of 4. Can anyone shed any light on this for
> me?
>
>
> Also, can anyone recommend a good reference book that is an introduction to
> registration techniques and would explain from an application viewpoint the
> pros and cons of particular transformations. I have Terry Yoo's book but I
> would like something more introductory. I have also read the itk software
> guide but would like to find another reference.
>
>
> Thanks for any help you can give.
>
> Maureen
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.itk.org/pipermail/insight-users/attachments/20080725/fa84b327/attachment.htm>