Dave Green's ‘cubehelix’ colour scheme

Contents:

1. 1) Background: The Problem
2. 2) A Solution
3. 3) Implementations
   • Fortran
   • Astronomical Packages (AIPS / casa / DS9 / karma kvis / miriad / CIAO / PSRchive / HEALpix / APT),
   • Other (gnuplot / IDL / matlab / python (matplotlib / seaborn) / Octave / javascript / R / NCL / IGOR / DigitalMicrograph)
4. 4) Artisitic uses.

Example images using the ‘cubehelix’ colour scheme using AIPS. (left) A portion of the Galactic plane from the VGPS survey at 1.4 GHz, which includes the supernova remnant G35.6-0.4 (see: 2009MNRAS.399..177G). (right) The Crab Nebula at 347 GHz observed with SCUBA on the JCMT (see 2004MNRAS.355.1315G).

Other examples elsewhere: (1) the Sun; (2) edge detection; (3) (a) optical images of Comet Lovejoy (C/2014 Q2), (b) Barnard's Loop in Orion, (c) the Horsehead Nebula, and (d) a galaxy with lensed arcs; (4) particulate polution. (5) two images of fractals: (a) Mandelbrot, and (b) generalized Newton; (6) some geographical images (a) various; (b) of elevation of Europe and surroundings, (c) and (d) elevation for smaller regions; (7) the Mona Lisa.

...

1) Background: The Problem

Many colour schemes used to display astronomical intensity images do not have an underlying increase in the perception of the brightness of the colours used (e.g. burning out to red for the high data values, but using yellow/green for intermediate data values, which are perceived as being brighter than the red).

2) A Solution

I have written up the implementation of a colour scheme – called ‘cubehelix’ – which is intended to be perceived as increasing in intensity. This is a family of colour schemes that go from black to white, deviating away from a pure greyscale (i.e. the diagonal from black to white in a colour cube) using a tapered helix in the colour cube, while ensuring a continuous increase in perceived intensity. The deviation from the diagonal takes into account that red, green and blue are not perceived equally in terms of intensity. These colour schemes print as a monotonically increasing greyscale on black and white postscript devices.

This colour scheme is described in more detail in:

Please cite this paper if you use ‘cubehelix’ in any publications.

This colour scheme is now recommended in the Graphics Guide for authors from the American Astronomical Society (although, rather strangely, they decided to call it ‘cube-helix’ with a hyphen).

There is no single ‘cubehelix’, as there are several parameters that control this family of colour schemes:

1. the ‘start’ colour (this is the direction of the predominant colour deviation from black at the start of the colour scheme, with R=1, G=2, B=3 etc.);
2. the ‘number’ of R → G → B rotations that are made from the start (i.e. black) to the end (i.e. white) of the colour scheme;
3. a ‘hue’ parameter, which controls how saturated the colour of all hues are (if this parameter is zero then the colour scheme is purely a greyscale; if the parameter is larger than 1, then some R, G or B values may be out of range near the start or end colour scheme, so will have to be clipped, although if only a few colour levels are clipped, the resulting colour scheme may still be satisfactory);
4. a ‘gamma factor’ can be used to emphasise low or high intensity values.

Whatever the parameters, the colour scheme is from black to white, i.e. it uses a the full range of the intensity available. This is in contrast to both the matplotlib viridis colour scheme (the default since version 2.0), and the Matlab parula colour scheme (the default since version R2014b), which are from dark blue/purple to a yellow, so only use about 70% of the full intensity range available.

2.1) Some Examples

2.1.1) My ‘default’

The image below shows what I consider my ‘default’ scheme.

And here is a 3-D visualisation of how the colour scheme spirals around the diagonal of the colour cube.

2.1.2) With more colour

Note that since this ‘default’ scheme uses a hue parameter of 1, no R, G or B values are clipped, and the perceived intensity is constantly increasing. If the hue parameter is increased, then the scheme becomes more colourful, but some of the R, G or B values are clipped (although the underlying perception of intensity is not exactly constantly increasing, it is still monotonically increasing). Below is a ‘more hue’ scheme, for a hue parameter of 1.5.

2.2) Other Examples

As this colour scheme has several paramaters, there are alternative versions you can use/make if you do not like the default version. Several people have said they do not like the ‘default’ scheme above, with so much pink. Instead, reducing the number of rotations to −1 gives:

or changing the start colour to 1.5 gives:

This page gives several other colour wedges made with different parameters (along with the corresponding colour scheme files for use in DS9 or kvis, see below).

Or you can make your own – using javascript – to visualise what other combinations of parameters produce, and also produce colour scheme files for DS9, or R/G/B values as decimal or hex numbers (I believe this works in Firefox, Safari, Chrome and Edge, but not with IE).

3) Implementations

3.1) Fortran

As in the published paper, here is the Fortran 77 source of a subroutine that implements this scheme, which is free software released into the public domain, under the unlicense license.

3.2) Astronomical packages (AIPS / casa / DS9 / karma kvis / miriad / CIAO / PSRchive / HEALpix / APT)

This scheme is also implemented in:

• AIPS: from the 31DEC10 release as TVHELIX – for examples, see the plots at the top of this page (note: you may have to RUN NEWPARMS so that TVHELIX is declared as verb within AIPS; and if you run a midnight job, for 31DEC11 you should have a DOPRINT option for TVHELIX to print out the four parameters used to control the colour scheme, which is probably useful).
• casa: from version 3.1.0, as a colo(u)r scheme called Cube Helix (with a space). You can make this the default for imview by adding a line "display.colormaps.defaultcolormap: Cube Helix" to your ~/.casarc file.
• Carta: (Cube Analysis and Rendering Tool for Astronomy) has cubehelix as one of the ‘colormap’ options.
• DS9: recent versions of DS9 have several cubehelix colour schemes built in (use the top menu ‘Color’ option). It is also possible to read into DS9 specific colour maps from text files, using
  
      prompt> ds9 -cmap file <cmap-filename>
  
  and several version of the scheme are available, or you can make others here.
• karma/kvis: it is possible to read in specific colour maps from .kf binary files, using the
  
      Intensity → PseudoColour (emulated) → Load
  
  menus. Appropriate .kf files for several version of the scheme are available (these work fine for me, under Linux and Solaris, but may not work with all implementations, as the .kf files are, I believe, memory dumps, so for example might need byte-swapping), or you can make others here.
• miriad: Cubehelix has been implemented in a miriad (for releases from r1.24), in cgdisp and cgspec.
• Chandra Interactive Analysis of Observations (CIAO): Cubehelix is included, as ‘contributed’ script chips_contrib.helix, from Version 4.6.1.
• Stefan Osłowski has implemented cubehelix in the development branch of the PSRchive radio pulsar processing package, and also has produced ls_image.h in C++ for use with HEALpix (use the Get_Colour_Cubehelix function instead of Get_Colour).
• Russ Laher, of IPAC, has implemented cubehelix in Aperture Photometry Tool (APT) and Thoth (a data visualisation tool).

3.3) Other implementations

Other implementations I know of are:

• gnuplot: Ingo Thies (of Universität Bonn) has added this to his other useful colortools for gnuplot (also there is a ‘cubehelix’ palette option is recent development versions of gnuplot, as shown in this demo).
• IDL:
  • Aaron Barth (of University of California, Irvine) has added this colo(u)r scheme to his ATV data display and analysis package for IDL.
  • James Davenport has written an implementation for IDL – cubehelix.pro.
• matlab:
  • Phillip Graff (a former colleague of mine here at the University of Cambridge, now at NASA) has written function implementing this for matlab – CubeHelix.m.
  • Stephen Cobeldick has written a another version, which extends cubehelix is a variety of ways, including a interactive demonstration (note: the ‘hue’ parameter is renamed ‘sat’ for saturation).
• python/matplotlib:
  • The cubehelix scheme is available from the 1.1.0 release of matplotlib.
  • Ian Heywood (of the University of Oxford) has implemented this for python (with matplotlib) – cubehelix.py.
  • Oleg Smirnov (of ASTRON) has incorporated cubehelix into his .fits viewer Tigger (see screenshot and Colormaps.py, specifically see CubeHelixColormap.colorize()), which was added to Version 1.2 (of 17th September 2011) of MeqTrees.
  • James Davenport (of the University of Washington) has written a full implementation for python, available on github.
  • Micheal Waskom has included cubehelix in his seaborn statistical data visualization, based on matplotlib.
• cpt-city:Several cubehelix color schemes are in the cpt-city archive of colour gradients, in a variety of formats (cpt: colour palette tables for use with the Generic Mapping Tools; c3g: CSS3 gradients; ggr: gradients for the GNU image manipulation program, GIMP; gpf: Gnuplot palette files; grd: (version 3) for Paint Shop Pro and Photoshop; inc: POV-Ray colour map headers; sao: the SAO format used by the astronomical image viewer, DS9; svg: scalar vector graphics gradients).
• Octave: since version 4.0, a cubehelix function is provided, for a single version of this colour scheme (i.e. it is not possible to adjust the parameters).
• javascript: Mike Bostock has implemented and extended cubehelix in JavaScript as a plugin for the D3.js visualisation library (see examples); Bass Jobsen has written a custom function to use this in Less (a preprocessor for CSS); Menelaos Perdikeas has written a standalone javascript library.
• R: Robin Evans has an implementation for R (from version 1.6 of his useful functions).
• NCL: A cubehelix colour scheme is included in the NCAR Command Language (NCL) plotting package, from version 6.2.0, as MPL_cubehelix.
• IGOR: cubehelix has been implemented for IGOR Pro (a scientific data analysis package), for version 6.30.x and above.
• DigitalMicrograph: A script is available (see 2014 Dec 16), by Bernhard Schaffer, for the DigitalMicrograph software package used in electron microscopy.

If you have another implementation, please let me know if you want it listed here.

4) Artistic uses

• Cube Helix 2022 by Otilia Sabina.
• Tangible Variations, by Laura Splan and Adam Lamson.

home | addresses | teaching | research