Distributed anisotropic diffusion provides a wrapper program around ITK's anisotropic diffusion filters that allows an input image to be spread across the memory of several computers and the processors of all of the computers to work simultaneously on the output. Distributed anisotropic diffusion allows the Visible Woman Head dataset to be smoothed with 100 iterations of the vector gradient magnitude anisotropic diffusion filter in 47 minutes on an 8 node 64 core cluster versus 53 minutes for just 10 iterations of standard vector gradient magnitude anisotropic diffusion on a single node.

Distributed Anisotropic Diffusion

New functions are proposed for itkMinimumMaximumImageCalculator. One function allows for excluding pixels of a specified value from the computations. Another function stores all pixel indices of the maximum/minimum value. This paper is accompanied with source code for the filter and test, test images and parameters, and expected output images. 

New Features for MinimumMaximumImageCalculator

This document describes an ITK-based implementation of intensity-based 2D/3D rigid image registration for patient setup assessment in external beam radiotherapy. The registration framework was designed to simultaneously register two projection images to a 3D image volume. The projection geometry was set up to simulate the x-ray imaging system that attached to a medical linear accelerator for cancer treatment. The normalized correlation was used as the similarity measure and the Powell's optimizer was used as the optimization method. Siddon-Jacobs fast ray-tracing algorithm was implemented to compute projection images from a 3D image volume.

ITK-Based Implementation of Two-Projection 2D/3D Registration Method with an Application in Patient Setup for External Beam Radiotherapy

This paper tests MPI in the Insight Journal testing environment.

MPI Read Write

ITK is becoming increasingly complex as it continues to grow and evolve. The SimpleITK initiative aims to address this concern by providing an easy-access layer around ITK for non-C++ expert developers. This article proposes an alternate (perhaps complementary) approach utilizing the strategy pattern. The software design pattern known as "strategy" allows an algorithm to be selected from a family of algorithms on-the-fly at runtime. Because ITK consists of many such algorithm families, this particular design pattern is quite interesting. This article describes a number of algorithm families which have been identified; so far thirteen families have been implemented, encapsulating nearly 130 filters/objects. Complete source code and examples are provided to demonstrate the concept.

Using the strategy pattern to simplify ITK

This document presents a system to fill a hole in an image by copying patches from elsewhere in the image. These patches should be a good continuation of the hole boundary into the hole. The patch copying is done in an order which attempts to preserve linear structures in the image. This implementation is based on the algorithm described in ``Object Removal by Exemplar-Based Inpainting'' (Criminisi et. al.).

The code is available here:
https://github.com/daviddoria/Inpainting

Criminisi Inpainting

This document describes a set of classes that designs a generic explicit deformable model in VTK. The iterative mechanism is first introduced through an inheritance of the vtkPolyDataAlgorithm class. This vtkIterativePolyDataAlgorithm is then a base for an implementation of the deformation. Two examples of deformation are presented through an inheritance of this base class. The provided source code may be used to build a ParaView plugin that harnesses the animation feature.


Explicit Deformable Model in VTK

This code provides a base implementation of Loopy Belief Propagation on MRFs in ITK. We use binary image denoising as an example problem to demonstrate this code.

This document is intended only to describe the implementation, not the theory. A complete tutorial on loopy belief propagation on MRFs is available here:

https://github.com/daviddoria/Tutorials/blob/master/BeliefPropagation/BeliefPropagation.pdf?raw=true

The code at the time of this submission is here:
https://github.com/daviddoria/BinaryDenoising/commit/e8bd152723584c3af7e30b5e529b8a80f2f2ee5c

And the latest version will be here:
https://github.com/daviddoria/BinaryDenoising



Loopy Belief Propagation on MRFs in ITK

Uniform B-splines are used widely in image processing because they provide maximal smoothness compared to any other piecewise polynomial of the same degree and support. This paper describes VTK classes for performing two functions: image interpolation via B-splines, and non-rigid coordinate transformation via B-splines. Special attention is paid to different boundary conditions for the ends of the spline: image interpolation supports clamped, mirrored, and repeated boundary conditions while B- spline grid transformations support clamped and zero-past-boundary conditions. The use of these classes for image deformation is demonstrated.

Uniform B-Splines for the VTK Imaging Pipeline

This code provides an implementation of two techniques from ``Poisson Image Editing'' on ITK images. First, we fill a hole in an image given only the pixel values on the boundary. Second, we copy a patch of an image into another image and make the result as smooth as possible. We also explain how these techniques can be used to reconstruct an image after perform editing operations in the gradient domain.

The code is available here: https://github.com/daviddoria/PoissonEditing