Find Optimum Visualization

Applies to: 3D/Studio. See also View and Compare Multiple 3D Data Sets.

A three-dimensional surface can be drawn in a great many ways, and the choice matters. The same data as a perceptually uniform gradient, as a classic spectrum, as a contour map, as a shaded surface, or seen from a different elevation, can make a feature obvious or hide it completely. Ordinarily you would find the right one by opening the graph settings, choosing, looking, and going back in -- once per candidate.

This procedure does the looking all at once. It renders your surface many ways in one window, you click the rendering you want, and OK keeps it.

Nothing in the data is altered, and no fit or calculation is disturbed. The only thing this procedure changes is the appearance of a graph, and only when you press OK.

Opening It

There are two routes, and they do the same thing to whatever you opened them from.

·    From the Explorer. Right-click a 3D data node -- or a digitized function node -- and choose Find Optimum Visualization.

·    From a 3D graph, inside a procedure. Right-click the surface and choose Find Optimum Visualization. This works in the surface fitting review, the residuals window, and the spline, non-parametric, smoothing and interpolation procedures -- wherever a 3D surface is drawn. When a procedure shows several panes, the one you right-clicked is the one that is sampled and the one that is changed.

Opening it from a graph is the more useful of the two in practice: you are looking at a surface, you can see that it is not reading well, and the remedy is one click away without leaving the procedure.

The Left Column

The column is the whole interface. Choosing any option in it rebuilds the panes in place -- there is no need to close and reopen the window to see a different family.

Sampler -- each of these renders one property across its whole range, everything else held identical. Only the property named varies from pane to pane, which is what makes the panes comparable:

·    TableCurve Gradients (19) -- every colormap TableCurve has carried, including your own two user-defined maps, so you can see yours beside the built-ins.

·    Python Gradients (25) -- a curated selection of the matplotlib colormaps, ordered perceptual first, then the classic rainbows, the thermal ramps, the cool and terrain family, and the diverging maps last.

·    Shaded Renderings -- all eight shaded surface types.

·    TableCurve Contours (19) and Python Contours (25) -- the same two colormap families drawn as contour maps rather than as surfaces, so a map can be compared in both forms.

·    Rotations (XY Angle) -- a full turn in 15 degree steps.

·    Elevations (Z Angle) -- sixteen viewing heights in 5 degree steps.

·    XY / Z Matrix -- five rotations against five elevations, read as a matrix: rows are elevation, columns are rotation.

AI Suggested -- these five do something different. Instead of walking one property, each is a small curated set that crosses the properties, chosen for a kind of question, and one cell of every grid is the rendering this surface is wearing right now, labelled CURRENT. That cell is the point of the grid: it makes the display a comparison rather than a catalogue. You are not being asked which of these is attractive, but whether any of them is better than what you have.

·    Perceptual -- colormaps engineered so that equal steps in z are equal steps in apparent brightness. The eye reads magnitude rather than inventing an edge wherever the ramp happens to change hue. This is the honest choice when the height of the surface is the message.

·    Classic TableCurve -- the spectral and two-hue maps TableCurve users have read for twenty-five years. A spectral ramp is less honest about magnitude and better at resolving a narrow feature, precisely because its hue turns quickly. This set lets you see that trade rather than take it on faith.

·    Residuals - Diverging -- a residual surface has a meaningful zero, and these colormaps place their neutral colour exactly on it, so the surface reads as "where is the fit high, where is it low" rather than "how big". A gray rendering is included as a control: if gray shows no structure, the coloured maps are showing you noise in two colours.

·    Residuals - Signed -- the same job with colour pairs that remain distinguishable to viewers with red-green colour blindness, and that survive printing. Use this when the sign pattern matters more than the size, or when the figure is going into a publication.

·    Threshold Regimes -- for work with limits rather than magnitudes. The zoned maps (RYGC and CGYR four-zone, RYC and CYR three-zone) do not encode a value at all; they encode which band a value falls in, with hard boundaries between bands. A continuous ramp actively hides the boundary that limit work is looking for. Three zones suits the common 90/95/99 case; four suits 90/95/99/99.9. See 3D Colormaps.

Choosing a Rendering

Click a pane to select it. The selected pane is framed, and any previously selected pane clears -- only one can be chosen at a time. Click the selected pane again to clear it.

Press OK to keep the selected rendering. With no pane selected, OK simply closes the window and changes nothing, so you may browse freely and leave without committing. Cancel also changes nothing.

The rendering you keep carries everything the pane was showing: the representation itself (contour, gradient or shaded), the colormap, and the viewing angles.

A note on the mesh, and on the data points. The panes are drawn on a coarser grid than a full-size graph, because a dense mesh inside a twenty-fifth of the screen reads as a dark wash and mutes the very colormap the comparison exists to show. For the same reason, the data points are hidden when there are too many of them to be legible at pane size. Both belong to the display only -- what you keep is drawn at your own mesh density and your own point setti

ng, not the sampler's.

What OK Does Depends on Where You Opened It

From a procedure graph, OK restyles that graph, there and then. Nothing outside the procedure is affected, and the procedure keeps the look the way it keeps any other graph setting you change while you are in it.

From a node that carries its own appearance -- surface fits, estimations, smooths, saved views, digitized function nodes and the rest -- OK gives that node the rendering you chose, and no other node changes. The node's calculations are untouched: a node's look and a node's algorithm are held separately, so restyling a surface fit cannot disturb the fit.

From a raw XYZ data node, it is different. Data nodes do not each carry an appearance; they are all drawn with one shared setting. So on a data node, OK offers to make your chosen rendering the setting for every 3D data node in the notebook. Because that is a change to the whole notebook, and because the previous setting is not kept anywhere, you are asked to confirm, and the message names what you are about to install -- a 2D contour, a 3D gradient, or a 3D shaded surface. This is the same action, and the same caution, as the Use for Tree Data button in the 3D data viewer.

Related Procedures

·    Use for Tree Data, in the 3D data viewer, installs the rendering you have set up by hand rather than one you have chosen from a grid. The two write the same setting.

·    View and Compare Multiple 3D Data Sets tiles different nodes in one window. This procedure tiles one surface rendered different ways.

·    Zooming In on a 3D Surface -- the other thing the graph right-click menu offers, and the answer to a different question: not how should this be drawn but which part of it do I want to see.