Research‎ > ‎

Medical Image Registration Toolbox

MIRT - Medical Image Registration Toolbox  for Matlab

MIRT is a Matlab software package for 2D and 3D non-rigid image registration. MIRT supports the following features:
Similarity Measures:
  • Mutual Information (MI)
  • Residual Complexity (RC)
  • Sum of Squared Differences (SSD)
  • Sum of Absolute Differences (SAD)
  • Correlation Coefficient (CC)
  • CD2, MS (ultrasound specific measures)
Transformation models:
  • Parametric: Free Form Deformation (FFD) - cubic B-splines
  • Non-parametric: Variational (curvature regularization) (coming soon..)
Optimization: Implicit Euler method (gradient-based)
Regularization: Curvature-based (penalized Laplacian of the displacements) both for parametric and non-parametric transformation models .
Multiresolution: yes (hierarchical registration)
Groupwise registration: all-to-the-mean, frame-by-frame sequentially, all-frames-to-the-reference, etc. Check out some examples of video stabilization [ here ]. There is also a GUI (MIRT_GUI) available for the group-wise non-rigid stabilization of 2D videos.

Please fill in the following form in order receive the email with the MIRT download link.


MIRT download form





Example 1. Non-rigid image registration using Residual Complexity (RC) similarity measure.
Reference image Source image Result image

We register the Source image onto the Reference one. Both images were corrupted by the multiplicative intensity fields shown. We use Residual Complexity as a similarity measure (alpha=1). The transformation error is less than 1 pixel on average. The computational time on my laptop (Core2Duo 2x1.5) is 10 seconds.
Multipl.field in refim Multipl.field in im True transform Found transform

Registration progress



Example 2.
Reference image Source image Result image

We register the Source image onto the Reference one.  For the similarity measure we used simply SSD. The transformation error is less than 1 pixel on average. The computational time on my laptop (Core2Duo 2x1.5) is 10 seconds.
Registration progress

Registration progress


Comments