Source-linked AI summary
A colour scheme for the display of astronomical intensity images
D. A. Green
TL;DR
Astronomical intensity images need colour schemes whose perceived brightness increases monotonically, because unequal red, green, and blue brightness can hinder interpretation. The paper describes cubehelix, which varies colour around a black-to-white path while controlling perceived intensity, hue, rotations, and gamma. The scheme is designed for screen display, should produce monotonic greyscale density on black-and-white PostScript devices, and has been incorporated into CASA and AIPS.
Problem
Astronomical colour schemes may not produce monotonically increasing perceived brightness because red, green, and blue are perceived unequally, hindering intensity interpretation.
Method
The paper constructs cubehelix by adding colour variation in a constant-perceived-intensity plane around the black-to-white greyscale diagonal, with hue, rotation, start-colour, and gamma controls.
Results
The scheme is designed for monotonically increasing perceived intensity, should print as monotonic greyscale density on black-and-white PostScript devices, and was incorporated into CASA and AIPS.
Takeaways & Limitations
Cubehelix provides a colour scheme for intensity images that preserves a monotonic perceived-brightness progression across screen display and supported black-and-white PostScript output.
Takeaways & Limitations
Colour printers may not preserve monotonically increasing perceived brightness because their representable colour gamut is poorer than that of computer displays.
Abstract
from arXiv · showhide
I describe a colour scheme that is appropriate for the screen display of intensity images. This -- unlike many currently available schemes -- is designed to be monotonically increasing in terms of its perceived brightness. Also, when printed on a black and white postscript printer, the scheme results in a greyscale with monotonically increasing brightness. This scheme has recently been incorporated into the radio astronomical analysis packages CASA and AIPS.
1. The problem
Astronomical intensity images are difficult to interpret when colour schemes do not produce monotonically increasing perceived brightness. The paper addresses this by designing schemes that account for unequal perceptions of red, green, and blue.
- Non-monotonic perceived brightness can make image intensities appear ordered incorrectly.For example, yellow at an intermediate intensity may appear brighter than blue or red assigned to higher intensities.
- The proposed colour schemes account for differing red, green, and blue brightness perceptions to maintain monotonically increasing perceived intensity.
2. Background
Colour-to-greyscale conversion uses a weighted luma value because green, red, and blue contribute unequally to perceived brightness. These coefficients underpin display and printing conversions in relevant systems.
- NTSC maps red, green, and blue to a Y luma signal representing black-and-white brightness.
- The NTSC background example concerns adding colour while preserving perceived brightness in black-and-white television.
- The NTSC luma coefficients reflect the ordering green, then red, then blue for perceived brightness.PAL uses similar coefficients, while HDTV uses a different but similarly ordered weighting.
- The NTSC coefficients are used for DeviceRGB-to-DeviceGray conversion in PostScript and colour-to-greyscale conversion in PGPLOT.
3. A Solution
The paper adapts a colour-cube helix so perceived brightness increases monotonically from black to white while colour varies in a constant-perceived-intensity plane. The scheme exposes controls for hue, rotations, start colour, and intensity mapping, and has been added to CASA and AIPS, with printing caveats.
- A Solution: The adapted scheme spirals around the greyscale diagonal while correcting the earlier equal-treatment assumption for red, green, and blue.
- A Solution: A 256-level example uses start colour 0.5, −1.5 rotations, hue 1.2, and gamma 1.0; the scheme was added to CASA and AIPS.
- A Solution: Cubehelix starts at black and ends at white, adding colour elements within a plane of constant perceived intensity.The colour element uses two unit colour vectors and is controlled by three parameters.
- A Solution: The start-colour parameter sets the predominant colour deviation from black, while the rotation parameter controls R, G, B cycling from black to white.
- A Solution: The hue parameter controls saturation, with colour deviation zero at both endpoints and largest in the middle; excessive hue can clip R, G, or B.
- A Solution: The gamma factor maps λ to λ^γ to emphasize low intensities when γ < 1 or high intensities when γ > 1.
- A Solution: Colour printing may not preserve monotonic perceived brightness because printer gamut is poorer, but black-and-white PostScript devices should print monotonic greyscale density.
4. Discussion and Conclusion
Cubehelix is one suitable scheme for intensity images, while other colour mappings may be needed for polarisation and spectral-line applications. Added colours should not distort an underlying intensity perception, and mappings must account for unequal discrimination between similar colours.
- The described scheme is only one suitable example for displaying intensity images.
- For polarisation and spectral-line images, colour may encode angle or velocity over an underlying greyscale intensity representation.
- Colour additions should not distort the intended perception of intensity in these applications.
- Mappings between colour and polarisation angle or velocity should account for unequal perception of differences between similar colours.
A. Fortran 77 code
The Fortran 77 implementation constructs the cube-helix colour table described in Section 3. It computes RGB values across NLEV levels using a gamma-adjusted position, rotating angle, and tapered amplitude.
- The code implements the colour scheme as a Fortran 77 subroutine that calculates a cube-helix colour table.The helix runs around the RGB cube diagonal from black to white.
- The colour table is generated for NLEV levels and stores separate RED, GRN, and BLU arrays.
- Each level uses a gamma-adjusted FRACT and a tapered AMP proportional to HUE*FRACT*(1-FRACT)/2.0.The RGB channels combine FRACT with angle-dependent trigonometric terms.