search

dlutgy / odl-LDDMM

2 stars 0 forks source link

About the code and the paper #1

Open cs123951 opened 4 years ago

cs123951 commented 4 years ago

Hi, thank you very much for sharing such useful code! But when I read Beg's paper, I found some of your codes in LDDMM_Beg.py are incorporate with psudo codes in Beg's paper.

  1. Beg did not mention kernel while you used a kernel in your code. I found some papers also talk about kernel. But It is hard for me to understand. Could you give me some advice?
  2. The divergence operator also confused me for "Computing Large Deformation Metric Mappings via Geodesic Flows of Diffeomorphisms" did not have such things.

Thank you again if you could give me some help.

dlutgy commented 4 years ago

I am sorry for checking my email so late, I don’t know if it will help you.

  1. For your first problem, in fact, My code is based on the other paper[1] and the Python package ’odl’ (https://github.com/odlgroup/odl) which used the LDDMM frame. The kernel is from the reproducing kernel Hilbert space theory. It can make up the solution, because the solution is not unique in LDDMM frame, this is a way to get one of the solutions. All these relationships can see in the paper[1]. In Beg’s paper, he used a differential operator L. Moreover, you can find this way and other ways in the book[2] Chapter 9.
  2. Well, this is also appear in the paper[1]. The divergence operator in the code is used to compute divergence of some vector.

I’m glad to answer your questions. And if you have some other question, please contact me. Thank you, and I hope you safe and happy.

[1] Chen C, Oktem O. Indirect image registration with large diffeomorphic deformations[J]. SIAM Journal on Imaging Sciences, 2018, 11(1):575–617. [2] Younes L. Shapes and diffeomorphisms[M]. Vol. 171: Springer Science; Business Media, 2010.