About Greedy

What is Greedy?

Greedy is a tool for fast medical image registration. It was developed by Paul Yushkevich at the Penn Image Computing and Science Lab at the University of Pennsylvania. The motivation for developing greedy was to have a really fast CPU-based deformable image registration tool that could be used in applications where many images have to be registered in parallel - like multi-atlas image segmentation.

Greedy borrows many concepts (and some implementation strategies) from the Symmetric Normalization (SyN) in ANTS. But greedy is non-symmetric, which makes it faster (in applications like multi-atlas segmentation, symmetric property is not required). Greedy also uses highly optimized code for image metric computation that adds extra speed.

This work is funded by the NIH/NIBIB under grants R01 EB-017255 and R01 EB-014146

Main Features

• Greedy diffeomorphic image registration
• Affine and rigid image registration (also, matching by moments of inertia)
• Support for normalized cross-correlation, mutual information, and sum of squared differences metrics
• Multiple images and multi-component images can be registered
• Fast CPU-based implementation (takes advantage of multi-threading and SIMD)
• Supports 2D and 3D registration
• Masks can be supplied for restricting registration to a region
• Supports NIFTI image format
• Single executable for registration and image/mesh re-slicing
• High-level API that can be called from other software (e.g., ITK-SNAP)
• Free open-source software (licensed under GPL3)
• It works!