One of the difficulties in preparing colour pictures is to compress the huge dynamic range of a 32-bit FITS images into an 8-bit colour picture that can be displayed on a screen. For me, this is a 2-step process. In a first step I clip the dynamic range of the FITS images and create 16-bit TIFFs, which are then further processed in Photoshop.
The clipping is as follows:
Dynamic range selection
My coadded images are sky-subtracted, normalised to 1 second exposure time, and then rescaled by the colour calibration factors. Sky subtraction means that the statistical mode of the background is zero. Due to noise, pixels can have negative values in these images. I determine the maximum brightness in the image which I do not want to become saturated in the colour image. This can be the core of a large galaxy or a bright filament in some nebulosity or else. For example, in an LRGB data set of some galaxy I have
| Filter | R | G | B | L |
| smallest and largest desired value | -0.023, 5.31 | -0.021, 4.38 | -0.026, 3.38 | -0.011, 8.3 |
| range chosen for the 16bit TIFF | -0.028, 5.5 | -0.028, 5.5 | -0.028, 5.5 | -0.014, 8.6 |
Note that the range chosen for the three TIFFs in filters R,G and B has to be identical, so that no colour distortions are introduced. The lower threshold of -0.028 is chosen such that the dynamic range is certainly not clipped at the lower end, and that not too much dynamic range is wasted. The upper threshold of 5.5 ADU ensures that the brightest feature of interest will not saturate in any of the RGB channels. One can safely ignore stars at this point, as their centres do not carry any interesting information and thus can burn out (appear white). The fainter stellar haloes will still show the star colours nicely.
The dynamic range of the luminance channel can be set independently from RGB, using the same criteria though (no saturation of regions of interest, no clipping at the lower end). The numbers for L will vary as individual exposure times are different, the throughput is higher, etc...
Before the 16-bit TIFF is created, I set all values higher than the upper threshold of 5.5 to 5.5, and I add 0.028 since the TIFF cannot hold negative values. The FITS images are then multiplied by 65536 / (5.5 + 0.028) to exploit the full range a 16-bit TIFF file offers, then converted to TIFF format. Further processing takes place in Photoshop.
FITS Liberator
One way to avoid cumbersome conversions from FITS to TIFF is the ESA / ESO / NASA FITS Liberator. With this tool one can load FITS images directly into Photoshop. Still, one of the first steps is to select the dynamic range. Extensive processing examples are available from their webpage.