ENH: Use non-breaking space before references

cite, ref, pageref, Figure, Table should have the reference
information adjacent to the preceeding text for clarity.

Author: hjmjohnson

SoftwareGuide/Latex/Appendices/CodingStyleGuide.tex

@@ -273,8 +273,8 @@ \subsection{CMake Build Environment}%
 CMake web pages at \href{https://www.cmake.org}{https://www.cmake.org} for more
 information.
 
-See Section~\ref{sec:UsingCMakeForConfiguringAndBuildingITK} on page
-\pageref{sec:UsingCMakeForConfiguringAndBuildingITK} for specifics about the use
+See Section~\ref{sec:UsingCMakeForConfiguringAndBuildingITK}
+on page~\pageref{sec:UsingCMakeForConfiguringAndBuildingITK} for specifics about the use
 of CMake in ITK.
 
 Section~\ref{sec:CMakeStyle} on page~\pageref{sec:CMakeStyle} provides a
@@ -358,7 +358,7 @@ \section{Copyright}%
 \end{minted}
 \normalsize
 
-See Chapter \ref{sec:InsightToolkitLicense} for further details on the ITK
+See Chapter~\ref{sec:InsightToolkitLicense} for further details on the ITK
 license.
 
 
@@ -631,7 +631,7 @@ \subsubsection{Naming Tests}%
 \label{subsubsec:NamingTests}
 
 Following the \code{TestDriver} philosophy, see Section~\ref{sec:Tests} on
-page \pageref{sec:Tests}, test files must be named with the same name used to
+page~\pageref{sec:Tests}, test files must be named with the same name used to
 name the \code{main} method contained in the test file (\code{.cxx}). This
 name should generally be indicative of the class tested, e.g.
 \small
@@ -646,8 +646,8 @@ \subsubsection{Naming Tests}%
 
 Note that all test files should start with the lowercase \code{itk} prefix.
 Hence, the main method name in a test is the sole exception to the method naming
-convention of starting all method names with capitals (see
-\ref{subsec:NamingMethodsAndFunctions}).
+convention of starting all method names with capitals
+(see~\ref{subsec:NamingMethodsAndFunctions}).
 
 A test's input argument number should always be named \code{argc}, and the input
 arguments \code{argv} for the sake of consistency.

SoftwareGuide/Latex/Appendices/GitWorkflow.tex

@@ -415,7 +415,7 @@ \subsubsection{Development}%
 This is where the real work happens. Edit, stage, and commit files
 repeatedly as needed for your work.
 
-During this step, avoid the \ref{par:UrgeToMerge} from an integration branch.
+During this step, avoid the~\ref{par:UrgeToMerge} from an integration branch.
 Keep your commits focused on the topic at hand.
 
 \begin{minted}[baselinestretch=1,fontsize=\footnotesize,linenos=false,bgcolor=ltgray]{bash}

SoftwareGuide/Latex/Architecture/Iterators.tex

@@ -9,7 +9,7 @@ \chapter{Iterators}%
 which are highly specialized to simplify common image processing tasks.
 
 The next section is a brief introduction that defines iterators in the context
-of ITK.  Section \ref{sec:IteratorsInterface} describes the programming
+of ITK.  Section~\ref{sec:IteratorsInterface} describes the programming
 interface common to most ITK image iterators.
 Sections~\ref{sec:ImageIterators}--\ref{sec:NeighborhoodIterators} document
 specific ITK iterator types and provide examples of how they are used.
@@ -807,8 +807,8 @@ \subsection{ShapedNeighborhoodIterator}%
 
 The functionality and interface of the shaped neighborhood iterator is best
 described by example.  We will use the ShapedNeighborhoodIterator to
-implement some binary image morphology algorithms (see \cite{Gonzalez1993},
-\cite{Castleman1996}, et al.).  The examples that follow implement erosion and
+implement some binary image morphology algorithms (see~\cite{Gonzalez1993},~\cite{Castleman1996}, et al.).
+The examples that follow implement erosion and
 dilation.
 
 \subsubsection{Shaped neighborhoods: morphological operations}%

SoftwareGuide/Latex/Architecture/SystemOverview.tex

@@ -98,14 +98,14 @@ \subsection{Generic Programming}%
 \index{template}
 
 Generic programming is a method of organizing libraries consisting of
-generic---or reusable---software components \cite{Musser1996}. The idea is to
+generic---or reusable---software components~\cite{Musser1996}. The idea is to
 make software that is capable of ``plugging together'' in an efficient,
 adaptable manner. The essential ideas of generic programming are
 \emph{containers} to hold data, \emph{iterators} to access the data, and
 \emph{generic algorithms} that use containers and iterators to create
 efficient, fundamental algorithms such as sorting. Generic programming is
 implemented in C++ with the \emph{template} programming mechanism and the
-use of the STL Standard Template Library \cite{Austern1999}.
+use of the STL Standard Template Library~\cite{Austern1999}.
 
 C++ templating is a programming technique allowing users to write software in
 terms of one or more unknown types \code{T}. To create executable code, the
@@ -330,7 +330,7 @@ \subsection{Event Handling}%
 \index{InvokeEvent()}
 
 Event handling in ITK is implemented using the Subject/Observer design
-pattern \cite{Gamma1995} (sometimes referred to as the Command/Observer
+pattern~\cite{Gamma1995} (sometimes referred to as the Command/Observer
 design pattern). In this approach, objects indicate that they are watching
 for a particular event---invoked by a particular instance---by registering
 with the instance that they are watching. For example, filters in ITK

SoftwareGuide/Latex/DesignAndFunctionality/AnisotropicDiffusionFiltering.tex

@@ -12,10 +12,10 @@
 fine structure of the image and thereby changes subtle aspects of the
 anatomical shapes in question.
 
-Perona and Malik \cite{Perona1990} introduced an alternative to
+Perona and Malik~\cite{Perona1990} introduced an alternative to
 linear-filtering that they called \emph{anisotropic diffusion}.  Anisotropic
-diffusion is closely related to the earlier work of Grossberg
-\cite{Grossberg1984}, who used similar nonlinear diffusion processes to model
+diffusion is closely related to the earlier work of Grossberg~\cite{Grossberg1984},
+who used similar nonlinear diffusion processes to model
 human vision.  The motivation for anisotropic diffusion (also called
 \emph{nonuniform} or \emph{variable conductance} diffusion) is that a Gaussian
 smoothed image is a single time slice of the solution to the heat equation,
@@ -41,7 +41,7 @@
 conductance parameter, $k$, and the time parameter, $t$, that is analogous to
 $\sigma$, the effective width of the filter when using Gaussian kernels.
 
-Equation \ref{eq:aniso} is a nonlinear partial differential equation that can
+Equation~\ref{eq:aniso} is a nonlinear partial differential equation that can
 be solved on a discrete grid using finite forward differences.  Thus, the
 smoothed image is obtained only by an iterative process, not a convolution or
 non-stationary, linear filter.  Typically, the number of iterations required
@@ -50,10 +50,9 @@
 purpose, single-processor computers.  The technique applies readily and
 effectively to 3D images, but requires more processing time.
 
-In the early 1990's several research groups \cite{Gerig1991,Whitaker1993d}
+In the early 1990's several research groups~\cite{Gerig1991,Whitaker1993d}
 demonstrated the effectiveness of anisotropic diffusion on medical images.  In
-a series of papers on the subject
-\cite{Whitaker1993,Whitaker1993b,Whitaker1993c,Whitaker1993d,Whitaker-thesis,Whitaker1994},
+a series of papers on the subject~\cite{Whitaker1993,Whitaker1993b,Whitaker1993c,Whitaker1993d,Whitaker-thesis,Whitaker1994},
 Whitaker described a detailed analytical and empirical analysis, introduced a
 smoothing term in the conductance that made the process more robust, invented a
 numerical scheme that virtually eliminated directional artifacts in the
@@ -78,19 +77,18 @@
 separately, but linked through the conductance term. Vector-valued diffusion
 is also useful for processing registered data from different devices or for
 denoising higher-order geometric or statistical features from scalar-valued
-images \cite{Whitaker1994,Yoo1993}.
+images~\cite{Whitaker1994,Yoo1993}.
 
 The output of anisotropic diffusion is an image or set of images that
 demonstrates reduced noise and texture but preserves, and can also enhance,
 edges.  Such images are useful for a variety of  processes including
 statistical classification, visualization, and geometric feature extraction.
-Previous work has shown \cite{Whitaker-thesis} that anisotropic diffusion, over
+Previous work has shown~\cite{Whitaker-thesis} that anisotropic diffusion, over
 a wide range of conductance parameters, offers quantifiable advantages over
 linear filtering for edge detection in medical images.
 
 Since the effectiveness of nonlinear diffusion was first demonstrated, numerous
-variations of this approach have surfaced in the literature \cite{Romeny1994}.
-These include alternatives for constructing dissimilarity measures
-\cite{Sapiro1996}, directional (i.e., tensor-valued) conductance terms
-\cite{Weickert1996,Alvarez1994} and level set interpretations
-\cite{Whitaker2001}.
+variations of this approach have surfaced in the literature~\cite{Romeny1994}.
+These include alternatives for constructing dissimilarity measures~\cite{Sapiro1996},
+directional (i.e., tensor-valued) conductance terms~\cite{Weickert1996,Alvarez1994}
+and level set interpretations~\cite{Whitaker2001}.

SoftwareGuide/Latex/DesignAndFunctionality/ConfidenceConnectedOnBrainWeb.tex

@@ -6,7 +6,7 @@ \subsubsection{Application of the Confidence Connected filter on the Brain Web D
 The previous code was used in this example replacing the image dimension by 3.
 Gradient Anistropic diffusion was applied to smooth the image. The filter used 2 iterations, a time step of 0.05 and a conductance value of 3. The smoothed volume was then segmented using the Confidence Connected approach. Five seed points were used at coordinate locations (118,85,92), (63,87,94), (63,157,90), (111,188,90), (111,50,88). The ConfidenceConnnected filter used the parameters, a neighborhood radius of 2, 5 iterations and an $f$ of 2.5 (the same as in the previous example). The results were then rendered using VolView.
 
-Figure \ref{fig:3DregionGrowingScreenshot1} shows the rendered volume. Figure \ref{fig:SlicesBrainWeb} shows an axial, sagittal and a coronal slice of the volume.
+Figure~\ref{fig:3DregionGrowingScreenshot1} shows the rendered volume. Figure~\ref{fig:SlicesBrainWeb} shows an axial, sagittal and a coronal slice of the volume.
 
 \begin{figure}
 \centering

SoftwareGuide/Latex/DesignAndFunctionality/DeformableRegistration.tex

@@ -24,7 +24,7 @@ \subsection{FEM-Based Image Registration}%
 
 \input{DeformableRegistration1.tex}
 
-Figure \ref{fig:DeformableRegistration1Output} presents the results of
+Figure~\ref{fig:DeformableRegistration1Output} presents the results of
 the FEM-based deformable registration applied to two time-separated
 slices of a living rat dataset.  Checkerboard comparisons of the two
 images are shown before registration (left) and after registration

SoftwareGuide/Latex/DesignAndFunctionality/DeformableRegistration4OnBrain.tex

@@ -14,7 +14,7 @@
 \code{ImageDimension} by 3.
 
 The registration method uses B-splines and an LBFGS optimizer. The trace in
-Table. \ref{tab:LBFGStrace} prints the trace of the optimizer through the
+Table~\ref{tab:LBFGStrace} prints the trace of the optimizer through the
 search space.
 
 \begin{table}
@@ -45,9 +45,9 @@
 minimum, $x$. ``Function Value" is the current value of the function, f(x).
 
 The resulting deformation field that maps the moving to the fixed image is
-shown in \ref{fig:DeformationFieldOutput}. A difference image of two slices
-before and after registration is shown in
-\ref{fig:DefRegistrationDiffScreenshot}. As can be seen from the figures, the
+shown in~\ref{fig:DeformationFieldOutput}. A difference image of two slices
+before and after registration is shown in~\ref{fig:DefRegistrationDiffScreenshot}.
+As can be seen from the figures, the
 deformation field is in close agreement to the one generated from the Thin
 plate spline warping.
 

SoftwareGuide/Latex/DesignAndFunctionality/DemonsRegistration.tex

@@ -10,8 +10,7 @@
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 
 For the problem of intra-modality deformable registration, the Insight
-Toolkit provides an implementation of Thirion's ``demons'' algorithm
-\cite{Thirion1995b,Thirion1998}.
+Toolkit provides an implementation of Thirion's ``demons'' algorithm~\cite{Thirion1995b,Thirion1998}.
 In this implementation, each image is viewed as a set of iso-intensity
 contours.  The main idea is that a regular grid of forces deform an image by
 pushing the contours in the normal direction.  The orientation and magnitude
@@ -25,7 +24,7 @@
 In the above equation, $f(\textbf{X})$ is the fixed image, $m(\textbf{X})$
 is the moving image to be registered, and $\textbf{D}(\textbf{X})$ is the displacement
 or optical flow between the images. It is well known in optical flow
-literature that Equation \ref{eqn:OpticalFlow} is insufficient to specify
+literature that Equation~\ref{eqn:OpticalFlow} is insufficient to specify
 $\textbf{D}(\textbf{X})$ locally and is usually determined using some form of
 regularization. For registration, the projection of the vector on the
 direction of the intensity gradient is used:
@@ -53,9 +52,8 @@
 invariant to the pixel scaling of the images.
 
 Starting with an initial deformation field $\textbf{D}^{0}(\textbf{X})$, the demons
-algorithm iteratively updates the field using Equation
-\ref{eqn:DemonsDisplacement} such that the field at the $N$-th iteration is
-given by:
+algorithm iteratively updates the field using Equation~\ref{eqn:DemonsDisplacement} such
+that the field at the $N$-th iteration is given by:
 
 \begin{equation}
 \textbf{D}^{N}(\textbf{X}) = \textbf{D}^{N-1}(\textbf{X}) - \frac

SoftwareGuide/Latex/DesignAndFunctionality/Filtering.tex

@@ -227,8 +227,7 @@ \subsection{Blurring}%
 of the most commonly used kernels is the Gaussian. Two implementations of
 Gaussian smoothing are available in the toolkit. The first one is based on a
 traditional convolution while the other is based on the application of IIR
-filters that approximate the convolution with a Gaussian
-\cite{Deriche1990,Deriche1993}.
+filters that approximate the convolution with a Gaussian~\cite{Deriche1990,Deriche1993}.
 
 \subsubsection{Discrete Gaussian}%
 \label{sec:DiscreteGaussianImageFilter}