# Revisiting Rigid Mesh Registration

**URL:** https://discourse.slicer.org/t/revisiting-rigid-mesh-registration/9424
**Category:** Support
**Created:** [December 7, 2019, 1:43pm UTC](https://discourse.slicer.org/t/revisiting-rigid-mesh-registration/9424 "2019-12-07T13:43:30Z")
**Posts on this page:** 1
**Showing post:** 5

<div class="post-metadata">

### Author: ![manjula](https://sea2.discourse-cdn.com/flex002/user_avatar/discourse.slicer.org/manjula/32/80981_2.png) [@manjula](https://discourse.slicer.org/u/manjula)
#### Post date: [December 8, 2019, 3:46pm UTC](https://discourse.slicer.org/t/revisiting-rigid-mesh-registration/9424/5 "2019-12-08T15:46:11Z")

</div>

> [@lassoan](#):
>
> application log?

@lassoan

These are the results I get with the cloud compare fine registration with RMS. As you can see when i look at the model to model distance and view it in shape population viewer with the same setting… our land marks show almost exact matches ( Green ) in the viewer.

 ![Screenshot from 2019-12-08 13-16-08](https://us1.discourse-cdn.com/flex002/uploads/slicer/original/3X/6/0/6032afd17d99626c43152c7376cc69a11673b5e9.png) ![Screenshot from 2019-12-08 13-15-52](https://us1.discourse-cdn.com/flex002/uploads/slicer/original/3X/4/2/42bdc7101eb0a609528bd60a6c9461dcc1841c9e.png)  
 ![Screenshot from 2019-12-08 13-17-27](https://us1.discourse-cdn.com/flex002/uploads/slicer/original/3X/a/e/ae13521e98b27443c7f8f6d26d1d439d2ebe8bc9.png) ![Screenshot from 2019-12-08 13-22-16](https://us1.discourse-cdn.com/flex002/uploads/slicer/original/3X/4/b/4ba7e585f5ef56784afc255624a62e0a86899099.jpeg)

I get almost same resutls with the cloud compare iteration method also. Again very good match on the landmarks we put in surgically.

 ![Screenshot from 2019-12-08 13-28-20](https://us1.discourse-cdn.com/flex002/uploads/slicer/original/3X/7/c/7cb0b07f8d90c83cc49ef26d419cf9081d2a4290.png)  
 ![Screenshot from 2019-12-08 13-25-09](https://us1.discourse-cdn.com/flex002/uploads/slicer/original/3X/6/f/6f8e1c646f18e7bb0a9ce14627bf0a9b1e1fd3fa.png)  
 ![Screenshot from 2019-12-08 13-27-29](https://us1.discourse-cdn.com/flex002/uploads/slicer/original/3X/e/d/edfbb8109f67a3436ac8406929ea77daf2f38389.jpeg)

However when i used the CMF surface registration results are like this, with not very good registration.

 ![Screenshot from 2019-12-08 13-37-12](https://us1.discourse-cdn.com/flex002/uploads/slicer/original/3X/1/8/18888b8ec90f26173464b5c042a9e3b57e06f6cf.jpeg)

I managed to get good results with IGT fiducial registration wizard but it was after lot of playing around with fiducial markers and changing ti manually until i could get very good alignment. This took lot of time to chaging the location of fiducials until it gave this result.

 ![Screenshot from 2019-12-08 16-16-46](https://us1.discourse-cdn.com/flex002/uploads/slicer/original/3X/1/6/166452ded916843979a1c21b17833ef6c42b2ffc.jpeg)

\*\*So my major Problem is \*\* why is cloud compare fine alignment gives better results in comparison to CMF surface registration ? Because as i read on both the module they use the ICP algorithms and why should it then give different results ?

Today i installed nightly version because i wanted to try it with CMFreg ROI method since i did not knew how to do only to select the region of interest as you suggested in a stl file and as for some reason in the stable version in my computer ROI registration was not marking the regions of interest for some reason. So i tried with nightly version and it worked and gave excellent results with ease compared to IGT registration. But still cloud compares registration is fully automated as i do not need to select any points.

 ![Screenshot from 2019-12-08 16-40-13](https://us1.discourse-cdn.com/flex002/uploads/slicer/original/3X/c/6/c66a44c196eaffdafe1bf370a2e16442937fca5a.png) ![Screenshot from 2019-12-08 16-40-06](https://us1.discourse-cdn.com/flex002/uploads/slicer/original/3X/d/9/d98c7787af8ff4db3f195090acea0ec8de6c2bac.jpeg)

So i am out of interest and also because almost all the research that we have planned or are currently been conducted involves registration i am very much interested in learning about this and for publication purposes i think it is much convenient to stick to one software (3D slicer) to do the registration and the model to model distance and colour mapping.

---

_[View the full topic](https://discourse.slicer.org/t/revisiting-rigid-mesh-registration/9424)._