Deconvolution and RGB

[MikeWiles]

Hello all,

I'm currently processing an image of M101 and I'm trying to add deconvolution to my arsenal of image processing tools.  I've studied the NGC 5189 Processing Example and I feel like I have a decent understanding.  I have deconvolved my Luminance image with good results, but I'm struggling with the RGB image even though I'm using essentially the same process.

At this point in my processing both images are still in a linear state and have not been combined with each other.  The luminance frame was shot at 1x1 binning, but the RGB frame is at 2x2 binning.  I'm using Dynamic PSF, Star Mask and a basic luminance mask in the process of deconvolving.  I have my global dark setting quite low at 0.02.

In the RGB image, I'm getting various results depending on the settings that I use.  In some cases, a lot of color information is lost and the image appears very flat and almost monochrome.  In other cases the background gets very noisy and the deconvolution doesn't seem to improve the details at all.

So my question is....is it necessary to deconvolve the RGB image?  My concern is that with different star sizes between the Luminance and RGB  frames that I'll have trouble with odd looking stars when I LRGB combine.  Is it advisable to deconvolve the individual Red, Green and Blue images prior to RGB combine?

Just looking for a little guidance on how to handle this.

Thanks,

Mike

[pfile]

imho it's not necessary to deconvolve the rgb image, unless i suppose the L decon was so aggressive that the stars are way tighter in the L image than in the RGB image. of course some people mask the stars in the L during deconvolution since you can still get ringing artifacts on bright stars despite using Deringing.

if you do want to deconvolve an RGB image, you may need different PSFs for each channel. however this is more of an issue with an OSC vs. a mono camera.

in the end you replace the implicit L channel in the RGB with the standalone L image, which serves to sharpen the combined image.

rob

[Warhen]

Remember Mike, that Luminance bears the detail/contrast to the eye-brain perception so sharpening Chrominance is both unnecessary and problematic. Problematic because a) it's generally weaker/noisier signal and b) it's 3 channels and prone to artifact. As Rob says, sometimes it is advantageous to perform it on one or more individual color channels, if stars need to be tightened dramatically due to poor focus, dew, or seeing.

[Nocturnal]

I suppose it may be worth repeating that many processes in PI can be applied to just the L of an RGB image. That means you can create your composite RGB image and sharpen L later on. There is hardly ever a reason to extract L, process it and re-apply. You can switch the view of your RGB image to just L or one of the color channels or chrominance and others to help with this processing. Naturally if you have an L from your imaging session you may want to process it before combining it with color channels but you won't see the result till after you combine. The benefit of processing L after it has been applied to your RGB is that you see the effect of your work on the final image and that is after all the goal.