Reducing Star Bloat in RGB Images
With Photoshop's Minimum Filter
One of the problems that arises when combining luminance frames imaged with a narrow-band H-alpha filter with ordinary RGB frames is the difference in the sizes of the stars: stars are much larger ("bloated") in the RGB frames than in the H-a images. Dimmer stars do not even register on the H-a image. This is completely predictable, since stars are black body continuum sources, while the H-a filter selectively records the 656.3 emission line of hydrogen. Unfortunately, when the H-a image is superimposed on an RGB image, the result for brighter stars is often a small, bright, central dot (from the luminosity layer), surrounded by a grayish ring (from the RGB layer). The challenge is to reduce the size of these gray rings, while still retaining as much as possible of the sensitivity of the RGB layer to dimmer stars.
After trying several methods on my M16 Ha-R,RGB CCD image, I found that the following procedure reduces the amount of bloat with a minimum of undesirable effect. It assumes that all of the stars to be reduced are greater than 5 or so pixels in diameter. The more pixels, the smoother the reducing effect (another argument for over-sampling).
1. Open the RGB image in Photoshop. If you have a pre-CS or CS-2 version of Photoshop, you may have to change the mode from 16 bit to 8 bit. Do this using Image/Mode. In CS (version 8) and CS-2 (version 9), the minimum filter works on 16 bit as well as 8 bit images.
2. Open Select/Color Range. The Preview box should show the image with only the brightest stars visible. If they are not visible, touch one of the bright stars with the eyedropper. Then adjust the "fuzziness" to include all of the stars you wish to reduce. (The greater the fuzziness, the more stars will be included. Don't include the smaller stars or you will reduce them out of existence!) Selection "tracks" will appear around the stars. If you need to deselect a star--- after you have the stars selected with the Select/Color Range tool (using the eyedropper) --- click on the magic wand and then do an alt--left-click on the star to be deselected. You should see a little minus sign next to the wand when you hold down the alt key. Anything you left-click while the minus sign is showing will be deselected.
3. Using Modify/Expand in the Select menu, enlarge the diameter of the selections so that the tracks fall outside the actual diameters of the stars. I found that a 2 pixel expansion worked well.
4. Now activate the Filter/Other/Minimum filter. the preview box will show the predicted effect. One pixel seems to be the best value to enter in the dialog box. Hit OK and save the modified image under a new name.
The following before and after images show the results.
Before minimum filter After minimum filter
The filter has reduced the size of the brighter stars by 25-30% without suppressing the dim stars or increasing background noise. Note that you can apply the minimum filter without first selecting the stars, but then, in addition to minimizing stars, you will also spuriously enlarge dark areas.
You can also apply this method to the individual color frames. Here is an example of the process applied to the green master image from the above M16 image:
The same reduction in star size is apparent. At the moment I don't see any advantage to applying the filter to three separate frames and then combining, especially since the frames have to be reduced to 8 bits before filtering.
I have not explored this method with a wide variety of images, but I can imagine problems arising with images in which large, bloated stars are not sufficiently saturated to fall within the selection range.
If this technique is useful to you, or if you find some major problems with it, please let me know at: mailto:firstname.lastname@example.org. Thanks!