1.7. Contributions by Fraunhofer MEVIS

1.7. Contributions by Fraunhofer MEVIS
	Chapter 1. What's New in MeVisLab 2.3?

Added the registration module MERIT.
The MERIT (MeVis Image Registration Toolkit) module is a software framework for image registration. It is particularly aimed at the rapid prototyping of registration methods often required in everyday work with MeVisLab. Among its core features are:
- 2-D/3-D affine-linear image transformations (translation, rigid, similarity, rigid+scale, affine) with customizable component orders.
- Newton-type optimization (and approximation thereof) with line-searching (linear or Armijo's rule).
- Multi-resolution image pyramids with custom downsampling, stop levels, and number of levels.
- Inherent pre-registration initialization schemes (e.g., initial matching of centers of gravity).
- Five common image similarity measures (SSD, NCC, NMI, LCC, NGF).
- Nearest neighbor, cubic, Lanczos, and linear image interpolation methods.
- Multi-threading support using OpenMP processor threads.
- Robust convergence criteria, e.g. Gill-Murray-Wright, for a unified convergence behavior of all algorithms.
- Entirely graphical usage through a consistent user interface.
- Error curve output, mask image support, plugin methods, and much more.
The python-based 2D plotting library Matplotlib ( http://matplotlib.sourceforge.net/) has been integrated into MeVisLab.
- Qt4 backend has been ported to PythonQt/MeVisLab.
- MatplotlibCanvas MDL control allows to embedd plots in MeVisLab UIs.
- See MatplotlibMDLExample for an example usage.

Figure 1.12. Example graph plotted with matplotlib.


1.6. Other Modules