AtrousWT causes stars to blow out?

[Nocturnal]

Hi,

I've been trying AtrousWT, a tool that's been very intimidating to me, to be honest. I'm trying to bring out some detail in a large nebula (Flaming Star). Increasing the bias in the first and second layer actually improves detail nicely but stars become very 'steep'. They used to be pleasantly smooth but now they seem to burn out to max white. How do I prevent this from happening? Do I have to go in and smooth the stars afterward?

http://gallery.tungstentech.com/main.php?g2_itemId=1704 (before)

http://gallery.tungstentech.com/main.php?g2_itemId=1701 (after)

[Simon Hicks]

Hi Sander,

Its certainly sharpened up the nebulosity nicely.....but yes, the stars are very steep. I guess the obvious answer is to use a starmask and then apply the AtrousWavelets. The trick will be to get the starmask just right so that it protects every star but doesn't leave artifacts (halos or the like) after the Atrous.

Also, increasing the DynamicRangeExtension in the Atrous will help reduce the star saturation....but maybe at the expense of the nebulosity brightness.

So as usual...the answer is easier than the execution. And I'm keen to see other replies on this....'cos I have the same problem.

Cheers
         Simon

[Carlos Milovic]

Hi Sander

First thing to try is a good star mask. You may just use the luminance to build it, with a touch of curves. HDRWT also may help to flatten other features, and leave only the stars.

Another thing to try is to expand the dynamic range. The downside is that you loose contrast/brightness, so you must perform new curves or histogram adjustments.

Anyway, indeed the stars blow out, and this happens with almost any high pass filter procedure. To have more control in ATWT with them, just select a peaked scale function.

[Nocturnal]

Thanks for the tips, I'll try them.

[mmirot]

To have more control in ATWT with them, just select a peaked scale function.

What is this?

Max

[Carlos Milovic]

Short answer: It is the kernell filter that is used to build the wavelets decomposition. A narrower filter separates more scales, and a wider one, includes more scales into the same layer. You may see that there are some predefined functions in the current ATWT process: 3x3 linear, 5x5 gaussian, and so on. There are some of them that says "peaked"... this are best suited for small scale only adjustments, and hence they are better suited do deal with stars and edge enhancement. 5x5 gaussian, and others like it, are best suited for large scale only adjustments.

Long answer: This needs a good tutorial about how wavelets work. I made a few years ago a quite long one teaching the basics of convolutions and wavelets, but can't remember in which language. I'll look for it.