Transparency

How to superpose plots and how to visualize point densities by translucent markers

Transparency is a simple but powerful tool whenever a plot contains more information than can be displayed by opaque colors, e.g., when markers cover each other, or when two maps should be displayed on top of each other. In MTEX transparency is controlled by the options

All of them take values between 0 (completely transparent) and 1 (completely opaque). Transparency requires the renderer of the figure to be set to 'opengl', which is the Matlab default.

Transparent Markers

Let us start with a set of orientations that is strongly concentrated around the identical orientation

cs = crystalSymmetry('m-3m');
odf = unimodalODF(orientation.id(cs),'halfwidth',10*degree);
ori = odf.discreteSample(2000);

h = Miller({1,0,0},{1,1,0},{1,1,1},cs);

Plotting these orientations in a pole figure the markers cover each other and the preferred orientations show up as solid, uniformly colored blobs. Neither the number of orientations nor the shape of the maxima is visible

plotPDF(ori,h,'MarkerSize',5,'all')

Making the markers almost transparent by the option 'MarkerAlpha' the scatter plot becomes a density like plot. Positions where many markers overlap remain dark, while isolated orientations are barely visible

plotPDF(ori,h,'MarkerAlpha',0.05,'MarkerSize',5,'all')

Face and edge of the markers may be made transparent independently by the options 'MarkerFaceAlpha' and 'MarkerEdgeAlpha'. Since the edges of overlapping markers accumulate much faster than their faces it is often useful to keep the edges slightly more opaque than the faces

plotPDF(ori,h,'MarkerFaceAlpha',0.01,'MarkerEdgeAlpha',0.05,...
  'MarkerSize',10,'all')

It should be stressed that transparency is only a visual approximation to the point density. Whenever the density itself is of interest it should be computed by kernel density estimation as explained in Density Estimation

plotPDF(ori,h,'contourf')
mtexColorbar

Superposing EBSD Maps

The most common application of transparency are superposed EBSD maps. Here the band contrast is plotted as a gray scale background and the orientation map is plotted half transparent on top of it. This way the texture and the image quality of the measurement are visible at the same time

mtexdata forsterite silent

plot(ebsd,ebsd.bc)
mtexColorMap black2white

hold on
plot(ebsd('Forsterite'),ebsd('Forsterite').orientations,'faceAlpha',0.5)
hold off

Transparency Depending on a Property

Instead of a single value the option 'faceAlpha' also accepts a list of values - one for each pixel. This allows to fade out unreliable measurements, e.g., all pixels with a low band contrast. Since the transparency values have to be within the interval \([0,1]\) we normalize the band contrast by its mean value and cut off everything above 1

ebsdF = ebsd('Forsterite');

alpha = min(ebsdF.bc ./ mean(ebsdF.bc), 1);

plot(ebsdF,ebsdF.orientations,'faceAlpha',alpha,'figSize','large')

In the resulting map the pixels along the grain boundaries, where the Kikuchi patterns of two grains overlap and the band contrast is low, fade into the background. A second common choice for the transparency value is the local misorientation, see Grain Reference Orientation Deviation.

Transparent Grains

Grain maps are made transparent by the very same option 'faceAlpha'. Note that for grains the transparency value is additionally weighted by the color of the grain, i.e., light colored grains become more transparent than dark colored ones. The option 'translucent' is a synonym for 'faceAlpha'.

grains = calcGrains(ebsd('indexed'),'angle',10*degree);
grains = smoothBoundary(grains,5);

plot(ebsd,ebsd.bc)
mtexColorMap black2white

hold on
plot(grains('Forsterite'),grains('Forsterite').meanOrientation,'faceAlpha',0.5)
hold off

Transparent Grain Boundaries

Line plots as they are used for grain boundaries are made transparent by the option 'edgeAlpha'. Similarly as 'faceAlpha' it takes either a single value or one value for each boundary segment. In the following example the transparency is used to fade out low angle boundaries

gB = grains.boundary('Forsterite','Forsterite');

plot(grains,'translucent',0.5,'micronbar','off')
legend off

hold on
plot(gB,'edgeAlpha',gB.misorientation.angle ./ (30*degree),'lineWidth',3)
hold off

Transparent Surfaces

Transparency is also the method of choice to look inside of three dimensional objects. Plotting a crystal shape with the option 'faceAlpha' makes the back faces of the crystal visible

cS = crystalShape.olivine;

plot(cS,'faceAlpha',0.2)

This becomes even more important when additional objects are plotted inside the crystal, e.g., the slip systems

sS = slipSystem.fcc(crystalSymmetry('432'));
cSfcc = crystalShape.cube(crystalSymmetry('432'));

plot(cSfcc,'faceAlpha',0.2)
hold on
plot(cSfcc,sS(1),'faceColor','blue','faceAlpha',0.5)
hold off

Finally, three dimensional plots of orientation distribution functions make automatic use of transparency - the transparency of a contour level is proportional to its value such that the maxima of the ODF remain visible from outside. See Visualizing ODFs for more details.

plot3d(SantaFe)

Transparency and Export

Transparency is a feature of the 'opengl' renderer. Accordingly, figures containing transparent objects can not be exported as true vector graphics - when exporting to pdf or eps Matlab either rasterizes the figure or drops the transparency. It is therefore recommended to export such figures as bitmaps, e.g., by

saveFigure('transparency.png')

See Exporting Figures for more details on exporting.