Hello All,<br><br>I read the extensive explanation about how to select suitable optimizer for different parameter space of the Transform.<br><a href="http://www.itk.org/pipermail/insight-users/2006-September/019324.html">http://www.itk.org/pipermail/insight-users/2006-September/019324.html</a><br>
<br>It is pointed about that the scale parameter in the Similarity Transform *does not* form a vector space.<br>Scales shouldn't be added and subtracted. Their operation should be done in a logarithmic space. <br><br>
The Similarity Transform derived from VersorTransform : ScaleVersor3DTransform is the object transform used in my experiment.<br>In my situation, I customize the three scale factors in ScaleVersorTransform to be ( x_scale = y_scale = s, z_scale = 1.0 ). <br>
Thus, s is the only parameter in scale space I need to optimize.<br><br>For the optimization of this customized transform, I can follow the Versor3DRigidTransformOptimizer to do with the rotation + translation parameters.<br>
But for the scale factor, I don't know how to add it into the Versor3DRigidTransformOptimizer.<br>I am thinking whether I can just use the log of s to be the input variable for the optimizer.<br><br>Hope someone can give me some help.<br>
<br>Thanks a lot.<br><br>Best,<br>Chunfang<br><br><br>