A python implementation for affine and vector deformable transformation on 2D and 3D images

This repository provides an example code for 2D and 3D images transformation using different transformation methods, including affine transform and vector field deformation.

The code includes two parts, transformation grid generation and a bilinear resampler. The image matching metrics will also provides based on image histogram.

User may alter the dimension as MODE = '2D' or MODE = '3D'. To choose affine or vector transform, set METHOD='AFFINE' or METHOD='VECTOR' respectively.

The affine transform result will be displayed with translation, rotation and shear. Please note that in 3D mode the image will be rotated with Euler angle in Tait-Bryan form, concatenation order should always be taken in care.

Similarity metric plots based on histograms:

2D affine transformation result:

3D affine transformation result (From up to down rows: original, translation, rotation, shear):

2D vector transformation result (From top left to bottom right: original, deformed, original image grid (grey), deformed image grid (red), grid deformation vectors):

• Deep Learning Paper Implementations: Spatial Transformer Networks - Part I
• Deep Learning Paper Implementations: Spatial Transformer Networks - Part II
• Mutual information as an image matching metric