ImageJ_javaaddpath

src/main/resources/script_templates/MATLAB/Matlab3DViewerDemo_1.m

@@ -0,0 +1,121 @@
+%% MATLAB 3D viewer demo 1
+%
+% This example demonstrates how to use Miji to render a volume in 3D using
+% accelerated graphics, if your computer is adequately equiped (most
+% computers are, these days). We use the well known 'mri' data distributed
+% with MATLAB, and render it in the ImageJ 3D viewer.
+
+%% Load and prepare data using MATLAB
+% Load an example of 3D data from standard Matlab
+load('mri.mat');
+
+%%
+% Make a simple 8-bit 3D image from it. There is a singleton dimension to
+% remove.
+I = squeeze(D);
+
+%%
+% Ok, now this is an indexed color image (even if it is made a grays). The
+% actual color mapping is contained in the 'map' variable. We will make a
+% RGB image out of it.
+[R G B] = ind2rgb(I, map);
+
+%%
+% Just a short note here: we are going a complicated way to regenerate a
+% color image from an indexed image and its color mapping, because this is
+% the way the data is shipped to us in that MATLAB sample. It is a step
+% proper to MATLAB only, and does not affect nor relate to the Miji
+% framework we demonstrate here.
+% We could skip this step by directly feeding the image 'I' to MIJ, instead
+% of the color image 'J', using another method. The colors would not be
+% right, but would be enough to demonstrate the 3D rendering.
+% However, let us do things properly.
+
+%%
+% An extra step to do: R, G and B now contains doubles (0 to 1), and we
+% would like to make 3 channels of uint8 out of them.
+R  = uint8(255 * R);
+G  = uint8(255 * G);
+B  = uint8(255 * B);
+
+%%
+% We now put them together into one 3D color image (that is, with 4D). To
+% do so, we simply concatenate them along the 3th dimension.
+J = cat(4, R,G,B);
+
+%%
+% A note here: MIJ expects the dimensions of a 3D color image to be the
+% following: [ x y z color ]; this is why we did this 'cat' operation just
+% above. However, if you want to display the data in MATLAB's native
+% implay, they must be in the following order: [ x y color z ]. In the
+% latter case, 'permute' is your friend.
+
+%% Configure data volume using ImageJ objects
+% Now we start the fun and new part.
+% Everything so far has been only about MATLAB operations.
+
+%%
+% First, we launch Miji. Here we use the launcher in non-interactive mode.
+% The only thing that this will do is actually to set the path so that the
+% subsequent commands and classes can be found by Matlab.
+% We launched it with a 'false' in argument, to specify that we do not want
+% to diplay the ImageJ toolbar. Indeed, this example is a command line
+% example, so we choose not to display the GUI. Feel free to experiment.
+Miji(false)
+
+%%
+% The 3D viewer can only display ImagePlus. ImagePlus is the way ImageJ
+% represent images. We can't feed it directly MATLAB data. Fortunately,
+% that is where MIJ comes into handy. It has a function that can create an
+% ImagePlus from a Matlab object.
+% 1. The first argument is the name we will give to the image.
+% 2. The second argument is the Matlab data
+% 3. The last argument is a boolean. If true, the ImagePlus will be
+% displayed as an image sequence. You might find this useful as well.
+imp = MIJ.createColor('MRI data', J, false);
+
+%%
+% Since we had a color volume (4D data), we used the createColor method. If
+% we had only a grayscale volume (3D data), we could have used the
+% createImage method instead, which works the same.
+
+%%
+% Now comes a little non-mandatory tricky bit.
+% By default, the 3D viewer will assume that the image voxel is square,
+% that is, every voxel has a size of 1 in the X, Y and Z direction.
+% However, for the MRI data we are playing with, this is incorrect, as a
+% voxel is 2.5 times larger in the Z direction that in the X and Y
+% direction.
+% If we do not correct that, the head we are trying to display will look
+% flat.
+% A way to tell this to the 3D viewer is to create a Calibration object and
+% set its public field pixelDepth to 2.5. Then we set this object to be the
+% calibration of the ImagePlus, and the 3D viewer will be able to deal with
+% it.
+calibration = ij.measure.Calibration();
+calibration.pixelDepth = 2.5;
+imp.setCalibration(calibration);
+
+%% Display the data in ImageJ 3D viewer
+% Now for the display itself.
+%
+% We create an empty 3D viewer to start with. We do not show it yet.
+universe = ij3d.Image3DUniverse();
+
+%%
+% Now we show the 3D viewer window.
+universe.show();
+
+%%
+% Then we send it the data, and ask it to be displayed as a volumetric
+% rendering.
+c = universe.addVoltex(imp);
+%%
+% Et voilà!
+%%
+%
+% <<MRI-rendering.png>>
+%
+%%
+%
+% _Jean-Yves Tinevez \<jeanyves.tinevez at gmail.com\> - July 2011_

src/main/resources/script_templates/MATLAB/Matlab3DViewerDemo_2.m

@@ -0,0 +1,134 @@
+%% MATLAB 3D viewer demo 2
+%
+% In this example, we will demonstrate the surface rendering capabilities
+% of the 3D viewer.
+%
+% Rather than using an existing dataset, we will create a dummy one, in the
+% shape of multiple balls or octahedrons. The balls will be created one by
+% one, from a volume where all pixels are 0, but within the object volume,
+% whose location and size is set randomly. From this volume, we ask the 3D
+% viewer to create an iso-surface (one for each object), and to display it
+% in the 3D universe, with a random color.
+
+%%
+% Let us define the size of our target image volume. Not too big to save
+% memory.
+width = 128;
+height = 128;
+depth = 64;
+
+%%
+% Here we determine how many objects we want, and their mean size. We will
+% generate them randomly.
+nballs = 10;
+radius = 20;
+
+%%
+% We pre-allocate the image volume.
+I = zeros(width, height, depth, 'uint8');
+
+%%
+% To simplify the object creation step, we create the grid containing each
+% pixel coordinates.
+[X, Y, Z] = meshgrid(1:width, 1:height, 1:depth);
+
+%%
+% Since we want to add the objects in live mode, we need to create and show
+% the universe before creating and adding the balls. So we call Miji first,
+Miji(false)
+
+%%
+% Then deal with the 3D universe.
+universe = ij3d.Image3DUniverse();
+universe.show();
+
+%%
+% Now add the objects one by one, in a loop.
+for i = 1 : nballs
+
+    %%
+    % Determine the ball center and radius randomly.
+    x = round( width * rand );
+    y = round( height * rand );
+    z = round( depth * rand );
+    r = round( radius * (1 + randn/2) );
+    intensity = 255;
+
+    %%
+    % Logical indexing: we find the pixel indices that are within the
+    % object's volume.
+
+    % This gives you a ball
+    % index = (X-x).^2 + (Y-y).^2 + (Z-z).^2 < r^2;
+
+    % This gives you an octaedron
+    index = abs(X-x) + abs(Y-y) + abs(Z-z) < r;
+
+    %%
+    % Then we set the intensity of these pixels to 'intensity' (here, 255)
+    % and the value of pixels outside the object volume to 0.
+    I(~index) = 0;
+    I(index) = intensity;
+
+    %%
+    % We now export this MATLAB volume to a new ImagePlus image, without
+    % displaying it.
+    imp = MIJ.createImage(['Ball ' int2str(i)], I, false);
+
+    %%
+    % It is possible to specify the color of the iso-surface at creation,
+    % but the 3D viewer expects a |Color3f| object, which is part of the
+    % |org.scijava.vecmath package|. We determine its actual color
+    % randomly again.
+    color = org.scijava.vecmath.Color3f(rand, rand, rand);
+
+    %%
+    % Finally, we add the object to the 3D universe, in the shape of an
+    % iso-surface. Arguments meaning is detailed below.
+    c = universe.addMesh(imp, ...          - this is the source ImagePlus
+        color, ...                         - this is the destination color
+        ['Ball ' int2str(i)], ...          - the name of the iso-surface
+        1, ...                             - the threshold, see below
+        [true true true], ...              - what channels to display
+        1 ...                              - the resampling factor
+        );
+
+    %%
+    % Ok, so the meanings of the |imp|, |color| and |name| arguments are
+    % trivial. The |threshold| value, here set to 1, is important for this
+    % display mode.
+    %
+    % Isosurface rendering uses the fast marching cube algorithm, which
+    % assumes that there is a threshold that separates the background from
+    % the foreground. In our case, it is easy to determine, since pixels
+    % out of the ball have a 0 intensity, and inside the ball they have an
+    % intensity of 255. So any threshold value between these two numbers
+    % would have worked.
+    %
+    % The array containing the 3 boolean values is used to specify what
+    % channels will be included in the surface rendering, if you feed it a
+    % color image. In our case, we simply set them all to true and do not
+    % care much. Finally, the resampling factor is used to win some time
+    % and memory at the expense of precision. If set to more than 1, the
+    % image is downsampled before the marching cube algorithm is applied,
+    % resulting in less meshes obtained in smaller time.
+
+    %%
+    % Now, this individual step is going to take some time, so it is very
+    % likely that you see each ball appearing one by one. That is because
+    % computing the surface mesh is expensive, so the rendering lags a bit
+    % behind the command.
+
+end
+
+%%
+% That's it. Note that some of our objects might be clipped because they
+% cross the boundaries of the volume we defined in the variable |I|.
+%
+%
+%
+% <<octahedrons-rendering.png>>
+%%
+%
+% _Jean-Yves Tinevez \<jeanyves.tinevez at gmail.com\> - July 2011_
+

src/main/resources/script_templates/MATLAB/Matlab3DViewerDemo_3.m

@@ -0,0 +1,88 @@
+%% MATLAB 3D viewer demo 3
+%
+% In this demo file, we will use the surface plot mode of the 3D viewer. It
+% is a mode of rendering where a 3D surface mesh is created from a 2D
+% image. Here, the pixel intensity of the 2D image is interpreted as the
+% elevation (Z coordinates), which is enough to generate the 3D surface.
+% 
+% It is a mode which is particularly useful for images that have indeed a
+% pixel intensity that can be interpreted as elevation, such as maps. Here,
+% we try to use the surface plot mode of the 3D viewer with a surface well
+% known to MATLAB users.
+
+%% Generate elevation data in MATLAB
+% This is the well known *membrane* dataset, that serves as a generator for
+% the MATLAB logo.
+
+%%
+% This will generate a 51*51 image, for which intensity should be
+% interpreted as height. It is not a 3D data per se, but its rendering will
+% be. 
+Z = membrane(1,25);
+
+%%
+% The trouble is that Z is made of doubles, the most common MATLAB type,
+% whether the 3D viewer only accepts 8-bit images. So we have do some
+% conversion before rendering it. The following commands will stretch the
+% range of Z to [0; 255] and cast to |uint8|.
+Imin = min(Z(:));
+Imax = max(Z(:));
+I = uint8( 200 * (Z-Imin) / (Imax-Imin) );
+
+%% Send data to the 3D viewer, through miji
+%  Launch Miji
+Miji(false);
+
+%%
+% We create an ImagePlus from the 2D image.
+imp = MIJ.createImage('MATLAB peaks', I, false);
+
+%%
+% Create and display a new 3D universe.
+universe = ij3d.Image3DUniverse();
+universe.show();
+
+%%
+% Feed it the previous ImagePlus, but render it as a surface plot, where
+% the intensity is encoded as height in a 3D space.
+color = org.scijava.vecmath.Color3f(240 / 255, 120 / 255, 20 / 255);
+c = universe.addSurfacePlot(imp, ...
+        org.scijava.vecmath.Color3f(), ...
+        'Matlab Peak in 3D', ...
+        1, ...
+        [true true true], ...
+        1);
+
+%%
+% Rotate it a little, so that it shows the same orientation that of the
+% actual MATLAB logo.
+universe.resetView();
+c.setColor(color);
+c.setTransform([1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1]);
+universe.fireContentChanged(c);
+universe.centerSelected(c);
+universe.rotateUniverse(org.scijava.vecmath.Vector3d(-1, -0.5, +0.2), +120 * pi / 180);
+
+%%
+% Et voilà! A beautiful monochrome MATLAB logo, rendered in an accelerated
+% 3D viewer. You can try the _Fullscreen_ option in the _View_ menu, to
+% maximize your experience.
+%
+%%
+% Note that it is monochrome: the MATLAB logo (type |logo| in the command
+% window) has two colors: the close side is yellow-orange-ish and the back
+% face is blueish. If you look at the |logo.m| code, you will see that
+% MATLAB guys generated these colors using 2 different light source of 2
+% different colors, which you cannot do in the 3D viewer.
+%%
+% 
+% <<MatlabLogoInFiji.png>>
+%
+%%
+%
+% _Jean-Yves Tinevez \<jeanyves.tinevez at gmail.com\>_
+%
+% _Johannes Schindelin \<johannes.schindelin@gmx.de\>_
+%
+% _- August 2011_
+