View Full Version here: : Somebody please explain Wavelets to me !!
28-11-2011, 09:40 PM
I'm flying blind when it comes to registax.
Tyring to process last nights Jupiter Avi and I know the data i have is better than the results i'm getting
my final image looks pixilated before i apply wavelets, and by the time i get to them i'm just sliding the bars with no understanding of what i am actually trying to achieve
Using a Philips SP900NC webcam on my dob, shooting with WXAstrocapture at 640x480
28-11-2011, 09:54 PM
Short and Sweet...
What is the process behind it...
28-11-2011, 11:08 PM
I don't know much about Registax, but I do know a little bit about wavelets in general and the wavelet-based tools in PixInsight, so I can perhaps give you a hint or two...
Wavelet analysis splits an image into a collection of different spatial frequency scales from very fine details up to large scale structures and allows you to perform modifications of the image at each scale. At each scale you can typically remove noise (smooth details) and/or sharpen (which can increase noise). These scales are sometimes called layers, and I think that's the term that Registax uses.
You will probably find that the smallest scale details (first layers) are the most noisy. Sometimes it is a big improvement to heavily denoise or even remove completely the finest scale details. Generally, it's the first few scale levels that you'll want to tweak. Modifying the large scale stuff doesn't usually have positive effects on the image. You need to walk that fine line between removing noise and removing detail.
Hope that helps a little. You will still need to play with the controls and experiment. Images will vary depending on the gear, atmospheric conditions and the object you are imaging. You'll always need to tweak things to see what works best. At least now you might have a better idea of what you're looking at.
29-11-2011, 12:53 AM
That makes some sort of sense to me !!!!
I guess I'll still need a bit of plaaying around to get the best out of it...
29-11-2011, 06:12 AM
Think about the six levels as being the "sharpening" of a pixel box ie 6 x 6 pixel box down to a 1 x 1 pixel box.....
I find on Ha solar imaging that I only use the #4-#6 level to "correct" for seeing conditions, then less and less of the others.
I think I use the #1 -#2 about 10% of the time and only then when the image already appears pretty tight and clear...just to get that extra tweak.....
If the image starts to appear "over cooked" - you've definately gone too far - back off.....
29-11-2011, 10:41 AM
On Lunar and Solar...
I use 1 and 2 sparingly
and 5 and 6 the most
It all comes down to the individual image. I have "saved" a few different wavelet settings, and use them to start off with, then adjust as needed.
29-11-2011, 11:20 AM
Any ideas why my first image, prior to using wavelets is pixelated ?
Looks like the graphics on a very old computer game.
When i get home tonight i might see if i can upload my stacked, unprocessed image
29-11-2011, 11:31 AM
I've been having pixelation issues with my DBK21. Had to do with the debayering process from the raw image capture codec to the RGB colour image. The recommendation was to use Y800 for capture (grayscale raw, but faster capture at 60FPS versus 30FPS max for RGB AVI capture). What codecs and methods are you using for capture?
BTW thanks so much for posting the original question here re: wavelets. I've been meaning to ask for ages, and the explanations here have really helped me understand better.
vBulletin® v3.8.7, Copyright ©2000-2013, vBulletin Solutions, Inc.