The computation speed for distance transforms becomes important in a wide variety of image processing applications. Current ITK library filters do not see any benefit from a multithreading environment. We introduce a three-dimensional signed parallel implementation of the exact Euclidean distance transform algorithm developed by Maurer et al. with a theoretical complexity of O(n/p) for n voxels and p threads. Through this parallelization and efficient use of data structures we obtain approximately 3
times mean speedup on standard tests on a 4-processor machine compared with the current ITK exact Euclidean distance transform filter.

Parallel N-Dimensional Exact Signed Euclidean Distance Transform

Grayscale dilation and erosion are basic transformations of mathematical morphology. Used together
or with other transformations, they are very useful tools for image analysis. However, they can be very
time consuming, especially with 3D images, and with large structuring elements. Several algorithms
have been created to decrease the computation time, some of them with some limitations of shape of
structuring element. We have implemented several algorithms, studied their performance in different
conditions, and shown that all of them are more efficients than the others in certain conditions.
 We finally introduce a new structuring element class and a some meta filter designed to select the best
algorithm depending on the image and the structuring element, and to smoothely integrate the different
algorithms available in the toolkit.

Consolidated morphology

A common usage when working with Fourier transform is to shift the
the image to put the zero-frequency component in the center of the
image. This contribution comes with a filter to perform this
transform.

FFT Shift

This document describes a filter called itkOrthogonalBisectImageFilter, which bisects an input image and copies each divided region into two separate output images. Pixel values are copied while maintaining their original pixel location. The output images are of the same size, dimension, and pixel type as the input image. Pixels that have not been copied from the input image are set to a user-specified value. This filter is restricted to bisection along a cardinal axis, i.e., in a direction orthogonal to a bisecting slice (line for 2D images or plane for 3D images). Submitted with this document is the source code for the filter and source code for demonstrating filter usage via an image input as an argument. Also included is source code for testing the functionality of the filter.

Orthogonal Bisection of an Image

This document describes a filter called itkMultipleUnlabeledImagesToLabeledImageFilter, which takes multiple images as input and generates a single image of labeled pixels. Pixels to be labeled for each image are determined by an intensity threshold. Submitted with this document is the source code for the filter and source code for demonstrating filter usage via an image input as an argument. Also included is source code for testing the functionality of the filter.

itkMultipleUnlabeledImagesToLabeledImageFilter

The current ITK release (2.8) already contains a filter that enables permuting coordinate axes of an image: PermuteAxesImageFilter. PermuteAxesTransform provides similar functionality in form if ITK transform. It enables permuting axes as a part of image resampling or it may be combined with other transforms. Such techniques save system resources as they does not require an extra filter to be present in a pipeline.

Permute Axes Transform for ITK

Polar transform is a geometric transform, that transforms points form cartesian coordinates to so-called polar coordinates. In case of 2D space, a point in polar coordinate system is addressed by radius and angle. In image processing, polar transform is usually used to convert rotations around the origin of polar coordinate system to translations. 

This submission contains the implementation of forward and inverse polar transforms in two very simple classes for the Insight Toolkit.

Polar Transforms for ITK

When an image is mostly composed of background pixels, most of the automatic
thresholding methods are failing to produce a relevant threshold. This is mainly
caused because one mode is over represented compared to the other in the histogram
of the image. The Kappa-Sigma clipping, a method largely used in astronomy, don't
try to separate 2 modes in the histogram, but rather try to find the properties of
the only usable mode, the one of the background, and compute a threshold to select
the values significantly different of the background.

Kappa Sigma Clipping

While filtering in N dimensions is a main feature of ITK, filtering an image
in N-1 dimensions, slice by slice, can be very useful in many cases. Currently,
this operation require a consequent amount of work to be done with ITK. A new
filter is provided with this article to perform this operation with only a few
lines of code.

Slice by slice filtering with ITK

Robust Automatic Threshold Selection (RATS) is a fast and noise robust
automatic thresholding method based on gradients in the image. The basic idea is to choose
the threshold at the intensity where the gradient are the highest.