New methods for colourmap optimisation developed by Centre

New methods for colourmap optimisation developed by Centre

A new paper* authored jointly by academics at the e-Research Centre and the International Seismological Centre (ISC) has been accepted at IEEE SciVis 2016 , to be held from 23-28 October in Baltimore, USA. The IEEE Scientific Visualization (SciVis) Conference, part of  IEEE VIS 2016 , solicits novel research ideas and innovative applications in all areas of scientific visualization.

The paper, to be published in a special issue of IEEE Transactions on Visualisation and Computer Graphics (TVCG), describes the team's work on the problem of perceptually optimising colourmaps. Two case studies are given in the form of the London Tube map, and the ISC's own colourmap for representing the depth of seismic activities. To complement the paper, a tool called Colourmap Hospital has been developed for use within the visualization community.

Designing a colourmap is a basic yet important task in data visualization. Colour is arguably one of the most used 'visual channels' in visualization, as it enables visual searching and aids memorisation and differentiation. In fact there is a common consensus that, among the four commonly-used visual channels, colour is more important than size, shape and orientation. Colourmaps designed manually by experienced visualization researchers and visual artists, or based on the recommendation of software such as ColorBrewer , are adequate for many applications. However, such colourmaps can sometimes create misleading visualizations and increase the mental effort required by the user as they attempt to differentiate between similar colours. At the same time however, colours in colourmaps often convey meaning through, for example, metaphor (eg red conveying danger). These colours can be reasonably effective from a metaphoric standpoint, but not from a perceptual standpoint.

The new optimisation methods developed by the authors provide a means of improving the 'perceptual distance' between the colours (that is, how visually different they are), whilst taking into account that users may not want some colours to stray too far from their intended metaphoric meanings.

The project was initially motivated by the need to design a number of colourmaps in a system for visualizing seismological data. The International Seismological Centre has been collecting and verifying earthquake data for more than a century, which is used in preparing and distributing The ISC Bulletin , the definitive summary of world seismicity. Bulletin data produced by the ISC are used by thousands of seismologists worldwide to estimate seismic hazards and in tectonic studies. A team of data analysts use seven ways of viewing data in their daily work, each with different colourmaps, requiring them to remember colour mappings without continually checking the legends. Colour symbolism therefore plays a critical role in supporting familiarization and memorization, and potentially in speeding up analysis.

The research team worked with the ISC to optimize the colours used in two related visualizations. One of these visualisations, for example, employs a specially-designed colourmap linking depth ranges to geographical regions through metaphoric, symbolic and even alliterative meaning. The colours in the colourmap were then optimised whilst restricting any changes of hue. This resulted in a colourmap which allowed differentiation between the colours in the most effective way whilst still maintaining the metaphoric meanings for the end user.

A Knowledge Transfer Partnership (KTP) associate, Dr Hui Fang, was embedded into the team of data analysts at ISC. Weekly meetings involving the lead analyst, the database manager, and visualization researchers enabled major landmarks in design and development to be examined in detail. The team collected feedback from the data analysts through questionnaires and face-to face meetings, which confirmed the merits of using metaphoric colours in visualization. The analysts felt the metaphoric mappings provided significant benefits, including being easy to remember and easy to articulate when training new analysts and communicating with external stakeholders. After applying the colour optimization, the users felt more confident about the quality of the new colourmap and particularly welcomed the colourmaps featuring the same metaphor but different saturation and luminance, enabling them to interpret the visualisations more easily.

Following the work with the ISC, Research Associate Dr Simon Walton produced a web-based tool called Colourmap Hospital, in order to make the technology available to a wider audience and to provide a valuable free tool to the visualization community. The tool enables a user to specify a set of colours, along with any constraints on these colours, and then view the optimised colours.

The web-based tool was also tested separately by 10 participants, all of whom had experience of using colourmaps in visualization. They evaluated the usability of the tool by performing five optimization tasks. They strongly confirmed the tool's usability in terms of distinguishable colours, and moderately in terms of pleasing or satisfactory results and ease of use. Improvements and additions suggested by the participants are now being incorporated into the tool for the next version.

The image below shows screenshots of Colourmap Hospital, which is available at http://bit.ly/cmhosp . The optimization itself is computed in a remote Python server application. The tool is undergoing active development, with a new version expected in October 2016.

* H. Fang (now Edge Hill University), S. Walton (University of Oxford), E. Delahaye, J. Harris, D. A. Storchak (all International Seismological Centre) and M. Chen (University of Oxford), Categorical Colormap Optimization with Visualization Case Studies