Human Viewable (color composite) Image creation.

Creates a composite colour image from sets of input FITS files, following the Lupton et al (2004) composition algorithm (with extensions by Hogg & Wherry.)

• Phil Marshall (KIPAC) [email protected]
• Cato Sandford (NYU)
• Anupreeta More (IPMU)
• Hugo Buddelmeijer (Kapteyn)

The main executable script is compose.py. It takes in 3 FITS files as input, and returns a color composite, color-saturated png image with an arcsinh stretch. Make sure this script is on your PATH, and than the humvi directory is on your PYTHONPATH.

Example: I have an image in each of three bandpasses, called i.fits, r.fits and g.fits, in the current directory. I wish to combine the three new images into an RGB color composite.

compose.py  -s 0.4,1.0,1.7  -p 1.0,0.02  -o gri.png  i.fits r.fits g.fits

Run compose.py -h to read about these options, but basically you can choose the contrast via the parameters -p Q,alpha and then the R,G,B color balance with the scales -s R,G,B. Good strategy is first to set alpha with Q=1 so you can just see the noise, then adjust Q if necessary to brighten up the features, and finally choose the scales so that the noise looks like an equal mixture of red, green and blue. This latter step may not be possible if the different channel images have very different sensitivity.

As well as Lupton's arcsinh stretch, a major advantage of the HumVI algorithm is that you can set the same scales and parameters for all your images, so that you can compare them across your survey. HumVI reads the zero points out of the FITS headers and uses them to put all the images on the same flux scale (ie so that all the pixel values have the same units), so do make sure your images are well calibrated and have informative, accurate headers.

Notes: In the attic there is an attempt (deconvolve.py) at a reworked version of the Magain, Courbin & Sohy (1998) deconvolution algorithm, that is non-operational. The problem of how to bring images from 3 different filters to a common resolution remains open. For now, don't go in the attic!

The composition script requires:

• the Python Image Library (PIL) available from http://www.pythonware.com/products/pil/
• numpy