the colour image
Now it's time to get back to the RGB colour image.
Open the RGB raw image. Here the Channel count is 3, and the
Interleaved box should be unchecked (the 3 channels are in separate blocks).
Do the same series of Positive and Negative Curve Stretches (on the combined RGB channel) on this image, but without de-selecting the stars. Eventually you will get a horrible looking
picture like this. Don't panic! Histogram stretching tends to
saturate colours and accentuate any colour casts there may have been in the
original. In the case of a webcam, the colour balance will almost
certainly be way off. In
this case there is too much red and probably not enough green.
image, I did the following Curve adjustments on the individual colour
channels. I normally keep the sky background slightly blue - this is
an aesthetic decision. The background level is kept somewhat high
because in the final image this level will be set by the luminosity
component and the higher levels in the RGB helps to reduce colour noise in
the dark parts of the image.
|Noise reduction of the luminosity image in SGBNR
This part of the process is optional and really depends on how bad the noise
in the luminosity image is. I use SGBNR (Selective Gaussian Blur
Noise Reduction) which, in summary, masks the lighter parts of the image,
and applies a gaussian blur to the dark areas, where noise in astronomical
images is usually most apparent. The Photoshop 16 bit luminosity image
is exported as a 16 bit TIFF file which SGBNR reads. I finish
all the major histogram stretching in Photoshop before I do this because
when the image comes back from SGBNR it is converted to 8 bits.
SGBNR is freeware and can be downloaded from
|The LRGB composite
Photoshop functions that work on 16 bit images are restricted, so before
proceeding I convert both the Y and RGB images to 8 bits (Image - Mode -
8Bits/Channel). Using the RGB image, first do a Gaussian Blur (Filter
- Blur - Gaussian Blur...). Experiment with the radius - 2 is usually
fine. The reason for this is to reduce colour noise, but overdone,
colour definition can be lost.
Now select the whole of the
Y image (Select - All). Copy it and paste onto the RGB image.
Photoshop adds the Y image as a layer on top of the RGB. The default
blending mode is Normal, but select Luminosity instead. This is now an
LRGB image of M83.
Enhancing colour saturation
If necessary, at this
stage colour saturation can be enhanced. I use a technique described
by Rob Gendler, which he calls
Layering. Essentially the colour saturation of the RGB image is
increased in Photoshop, the result is combined with the Luminance image at
50% opacity. This image is then flattened, and the process repeated
until the desired saturation is achieved. In this way colour
saturation can be increased without an unacceptable increase in colour
Once I am happy with the colour saturation, I crop the image and use the
Clone tool in Photoshop to correct trailed stars. The image still
lacked some sharpness, so I applied the Unsharp Mask filter. Unsharp
Masking tends to introduce dark halos around bright stars, so I selected
these stars by clicking on the centre of a star with the Magic Wand
(Tolerance = 60, 'Contiguous' unchecked). Then I de-select the core of
the galaxy and invert the selection. The Unsharp Mask applies then
only to parts of the image other than the bright stars. Some softness
was restored to the stars by re-inverting the selection, enlarging the
selection by 3 pixels (to include whole stars) and applying a slight
Gaussian Blur. Some dark halos remained, so before de-selecting, I
increased the brightness of the selection a bit.
Finally I added a caption and... this is the final image.
(Part 1, Introduction)
(Part 2, Aligning and Stacking in Iris)
(Part 3, Finishing the Image in Photoshop)