# GPU Registration - Mutual Information

**URL:** <https://discourse.itk.org/t/gpu-registration-mutual-information/1779>\
**Category:** Uncategorized\
**Created:** [April 11, 2019, 7:26am UTC](https://discourse.itk.org/t/gpu-registration-mutual-information/1779 "2019-04-11T07:26:34Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![Knut](https://discourse.itk.org/letter_avatar_proxy/v4/letter/k/bbe5ce/32.png) [@Knut](https://discourse.itk.org/u/Knut)\
**Post date:** [April 11, 2019, 7:26am UTC](https://discourse.itk.org/t/gpu-registration-mutual-information/1779/1 "2019-04-11T07:26:34Z")

</div>

Hello,

by GPUDemonsRegistrationFilter (itk 4.9.1) I was able to set up image registration.Registration results are really nice after fine tuning of parameters: SetMaximumError(0.08), SetMaximumKernelWidth(10) and SetIntensityDifferenceThreshold(0.001) in my case. Although using full image resolution of 3D MRI images, it is working fast. However, I have two questions:

- Is there a possibility of multi resolution approach to further increase speed?
- Is it possible to use mutual information metric to register differents types of images?

Thank you!

---

<div class="post-metadata">

**Author:** ![matt.mccormick](https://discourse.itk.org/user_avatar/discourse.itk.org/matt.mccormick/32/7_2.png) [@matt.mccormick](https://discourse.itk.org/u/matt.mccormick)\
**Post date:** [May 1, 2019, 4:35pm UTC](https://discourse.itk.org/t/gpu-registration-mutual-information/1779/2 "2019-05-01T16:35:08Z")

</div>

Hi @Knut,

Welcome to the ITK community! ☀

Good questions…

> [@Knut](#):
>
> Is there a possibility of multi resolution approach to further increase speed?

It may be possible to use `GPUDemonsRegistrationFilter` with the [`itk::MultiResolutionPDEDeformableRegistration`](https://itk.org/Doxygen/html/classitk_1_1MultiResolutionPDEDeformableRegistration.html) class via `SetRegistrationFilter`, but this is untested.

> [@Knut](#):
>
> Is it possible to use mutual information metric to register differents types of images?

The PDE / Demons registration is based on the “flow” of image intensities based on the assumption that intensities are similar. To use a mutual information metric to register different types of images, a different registration method is required, like the ITK registration framework.