That so smooth background?

[dmcclain]

I have studied a number of the great pics, and apart from their use of specialized line filters and the NASA colormapping, which places H-alpha in the green channel, I have wondered how they manage to get such smooth backgrounds. Sure, integrating for a long time helps a lot. But I did some example regions, like IC1396 (Elephant Trunk) with my OSC and it is mostly Red channel, which I take to mean H-alpha and S-II, for an equal duration exposure to what the great pics have done, and my images still show a grainy appearance. (e.g., 2-4 hours total integration at F/2)

I just stumbled on a technique, which make a lot of sense in retrospect, whereby the MMT is used to slightly enhance the 256 scale, (bias = +0.1 to +0.2), and lo and behold, my images have lost that grainy appearance, and show the same smooth background in the nebular regions as the great pics show. (Do the MMT in the linear image)

Makes sense in hindsight, since adding a very low-pass filtered image simply adds some of the local mean. If you add an amount equal to the local mean, the SNR increases by 2x, since you haven't increased the standard deviation in the local neighborhood. Hence it appears less grainy, and survives the nonlinear stretch to visibility much more robustly.

So I wonder if I have just rediscovered what the authors of those fabulous images have known all along?

[jkmorse]

Interesting.  Need to test that technique.

Jim

[dmcclain]

Doing something like adding in a lowpass filtered version of an image to the image inevitably decreases the contrast of small features of nebulae. That can be seen by considering that the image is the sum of low and high pass filtered components. Increasing the lowpass component is the same as scaling the image, then subtracting a portion of the highpass component - where the information is.

So, while it may be visually pleasing at a distance, it decreases the information value of the image.

Not being one to be put off for long by admonitions against some techniques (i.e., I eventually get around to abusing them anyway), I just tried MureDenoise on some integrated stacks of deBayered images. I do understand why one is cautioned against such stuff, but try it anyway. It works beautifully on the images I have tried so far (about half dozen of them). I see noise diminish by about half, which produces the kind of glassy smooth backgrounds I was seeking with the other abusive technique mentioned above.

I think, in purist terms, it is improper to MureDenoise a deBayered image because of the mixing of image statistics in a complicated way among the 3 color planes. And while that may violate the conditions under which the MURE algorithm was designed, it does not seem to destroy images by doing so. In fact, so far, it has improved them rather nicely.