using debayered integrated images

[Rowland]

I am processing images integrated in PI. The green caste that remains as a result of debayering is difficult to remove with wipe. I have had to resort to background extraction in PI, and then opening in ST. Also, VNG debayering seems to leave dark pixels scattered across the image addiing to noise.

Which is the best wipe algorithm to use and should I avoid VNG, using bilinear instead.

[admin]

Hi,

Your question seems twofold.

First I'll address the question about debayering.
The best debayering routine is no debayering at all. In DeepSkyStacker such a method is call 'Bayer Drizzling'.
By using a great number of frames which have been recorded with dithering between them, all sub pixels of the bayer matrix are given a chance to record a sample. However, I'm not sure if PixInsight provides such an option. See here for more details; http://deepskystacker.free.fr/english/t ... m#rawdecod
If Bayer Drizzling is not an option, consider using simple bilinear interpolation instead of VNG, as the noise in each individual frame causes the VNG algorithm to introduce artifacts and effectively 'bleeds' the noise to surrounding pixels. VNG attempts to reconstruct gradients and edges, the algorithm for which can quickly become confused by noise and show edges that aren't really there, influenced by a single noisy pixel. Bilinear reconstruction is 'dumb' but effective, not trying to attempt to be clever. This is the best approach in the face of noise.

Regarding your question on how to use the Wipe module to the greatest effect, I can't quite place the green cast you mention. If it is truly a green cast (e.g. a persistent green background color), then it should be removed by Wipe using the default or 'cast' preset.
In order for Wipe to be effective though, it is imperative that non-natural artifacts (such as dust bunnies, CCD artifacts and stacking artifacts) are removed, neutralized or masked out.
A sample image would be greatly beneficial in order to help you further.

Hope this helps!

Cheers,

[Rowland]

VNG is quite harsh. Is there a way of denoising with linear data in that case. Playing with Wipe settings I have fixed the green problem. I will look at DSS more closely.

[admin]

VNG is quite harsh. Is there a way of denoising with linear data in that case.

Absolutely. You can use the De-noise module (ST 1.2+) to remove most types of noise very effectively.

Playing with Wipe settings I have fixed the green problem. I will look at DSS more closely.

Good to hear!

Cheers,

[ghswen]

Hi Ivo,

(Gordon here, just joined!)

I'm using Nebulosity 3 for my pre-processing and it gives me a few options for debayering.

1. Color binning
2. Bilinear
3. VNG
4. PPG
and
5. AHD

Based on the previous post it sounds like 'bilinear' would be what I want to select, correct?

[admin]

Hi Ivo,

(Gordon here, just joined!)

I'm using Nebulosity 3 for my pre-processing and it gives me a few options for debayering.

1. Color binning
2. Bilinear
3. VNG
4. PPG
and
5. AHD

Based on the previous post it sounds like 'bilinear' would be what I want to select, correct?

Hi Gordon,

Welcome to the forums!
I wish there was a quick answer, but, as usual the answer is 'it depends'.
If you have quite a bit of noise in your image, color binning is best.
If you don't have any noise in your image, then AHD will be able to recover the most detail.

If you choose color binning, you'll end up with an image a quarter the resolution. But that makes sense when you think about it; for every 4 pixels on your image sensor, 1 red pixel is allocated, 2 green pixels are allocated and 1 blue pixel is allocated. This means that the red and blue channels are both recording at a quarter resolution anyway, while the green channel records at half resolution.

The debayering process tries to fill in the missing pixels as cleverly as it can, usually by trying to keep the 'full' resolution through 'guessing' the missing pixels (e.g. 3 missing red pixels, 2 missing green pixels, 3 missing blue pixels). Color binning does away with the guessing completely for red and blue, while it actually reduces noise a little for the green channel by averaging the two samples (pixels).

It also makes sense that the quality of the guessing is greatly dependent on how accurate (e.g. noise free) the real pixels were recorded. Better guessing can be performed if the guessing algorithm has more accurate data to work with to fill in the gaps. The trouble is, the more clever the algorithm (AHD is quite clever!) the more it is susceptible to noise and will fill in 'garbage' for the missing pixels.

Hope this helps!

[ghswen]

Thanks for the reply Ivo!

That helps!

I normally image with a Canon 350d (modified) at ISO1600 so I will generally get quite a bit of noise. I have been using the 'default' in Neb 2 which didn't really tell me what method of debayer it was using. I think I will make my next attempt using 'Color binning' (I had been using VNG) since Neb 3 gives me options and see what I get.

Thanks again!
Gordon

[Gedan]

Hi sorry to jump in,
Ivor I will be using an osc with hyperstar and upwards of 40 x25s subs would bayer drizzle in dss be appropriate here?

Gordon

[Gedan]

Actually can that, think I'm way of track.

[admin]

Hi sorry to jump in,
Ivor I will be using an osc with hyperstar and upwards of 40 x25s subs would bayer drizzle in dss be appropriate here?

Gordon

40 subs should work well, if you make sure you dither well between frames (spiraling out for example).