NA-MIC Project Weeks
Back to Projects List
Point set registration
Key Investigators
- María Armas López-de-Vergara (MACbioIDi)
- Abián Hernández-Guedes (ULPGC - GTMA-IUIBS - MACbioIDi)
- Juan Ruiz-Alzola (ULPGC - GTMA-IUIBS - MACbioIDi)
Project Description
This project focuses on applying a point set registration in 2D multichannel images and integrating it in a new 3D Slicer module. In order to use this kind of registration, it is necessary to obtain spatial features from the image and represent them as a point cloud.
Objective
- Explore strategies to extract a set of features from a 2D multichannel image and convert it to point cloud
- Select a registration algorithm for two point clouds and validate it with test cases
- Adapt the whole flow (features extraction and point cloud registration) to the original image
- Design and implement the 3D Slicer user interface for the proposed registration module
Approach and Plan
- Extract features based on contour information
- Select, integrate and validate a point set registration algorithm
- Apply registration result in the whole data imag
Progress and Next Steps
Progress:
- Points clouds have been created from the contour information of the images that we wanted to register.
- As point set registration algorithm the Coherent Point Drift (CPD) one has been chosen.
- A simple interpolation has been applied in order to make the deformation of the points set affects all the pixels of the image.
- The visualization of the obtained results has been performed through ParaView.
- Some trials have been carried out with both simple and real images.
Next steps:
- Improve the points cloud sampling strategy and the interpolation applied.
- Integrate the whole workflow in a 3D Slicer module.
- Optimize the developed code.
Illustrations
Point set registration example from a blue point set M to the red point set S:
First example
Input images:
Result:
Hand example
Input images:
Result:
