Registration, Comparison, and Transformation of 3D images

The programs in this section relate to the alignment, transformation and comparison of pairs of 3D images.


This program evauates the cost function as given by the free parameters on the command line.


Evaluate the cost gradient between two images and evaluate the transformation related gradient for it based on the given transformation model.


Transform a 3D image by applying a given 3D transformation that is defined by a 3D vector field v according to x:=x-v(x)


This program is used for non-rigid registration based on fluid dynamics as described in: Wollny, G. and Kruggel, F., 'Computational cost of non-rigid registration algorithms based on fluid dynamics', IEEE Transactions on Medical Imaging, 11(8), pp. 946-952, 2002. It uses SSD as the sole registration criterion.


This program implements the registration of two gray scale 3D images. The transformation applied is a symmeric diffeomorpic fluid dynamic registration. To work most efficiently, this program makes a few assumptions about the imput data, i.e. the image must be of the same size, have the same voxel spacing, and any intensity range normalization or equalization should also be done before calling this program.


This program is used to create an image comprising the pixel-wise norm of the ggradient of a given cost function.


Evaluate Euclidian distances between the corresponding landmarks in two landmark sets. The programs prints out only values for landmarks that are available and have location values in both sets


Transform the locations of the landmarks by means of a given 3D transformation. Note, landmark transformations do the inverse of an image transformation (in a manner of speaking), i.e. given a transformation V(x) in an image transformation the pixel intensity at x is set to the original intensity at V(x), while a landmark at x is moved to V(x).


This program implements the registration of two gray scale 3D images.


This program runs a non-rigid registration based on the given cost criteria and a given transformation model. Other than mia-3dnonrigidreg it doesn't support specific command line parameters to provide the images. Instead the images are specified dirctly when defining the cost function. Hence, image registrations can be executed that optimize the aligmnet of more than one image pair at the same time. Note, however, that all input images must be of the same dimension (in pixels)


This program implements the registration of two gray scale 3D images. The transformation is not penalized, therefore, one should only use translation, rigid, or affine transformations as target and run mia-3dnonrigidreg of nonrigid registration is to be achieved.


Transform a 3D image by applying a given 3D transformation.


Create a 3D vector field from a given transformation. The output vector field will have the dimesions as given in the transformation description.


Creates a 3D transformation from a vector field. The input vector field is simply encapsulated into the transformation file format. The boundary conditions and the image interpolator kernel can be set at the command line.


Compare two vector fields and print the out the difference norm per pixel to cout it it is larger than delta.