veil nebula dark patches

[alacant]

Hi everyone. I'm trying to diagnose the irregular darkish patches appearing on this snap. I think it may be produced by one of the StarTools denoise setting -I'll ask specific quetions on their forum- but am posting here just in case anyone has seen the effect before. The only quirk is a, lp filter but I've eliminated that as the cause. I've hit the contrast to bring out the dark areas. Some say they can't see what I'm on about; you can see the effect better by standing back from the screen. Maybe I'm imagining it... TIA for any pointers and clear skies.

[D4N]

This looks like DSLR noise after a denoise function.  Did you dither?

[alacant]

Yes. There's a random dither of up to 12 px. It does indeed occur after the denoise.

**Edit. Not that sure now. Here it is without the denoise. Dunno...

[D4N]

Without the denoise it looks much better.

 

I always found processing DSLR images a pain, I used to split the RGB channels and process separately then re-combine as the noise seems to vary a lot between channels.

[alacant]

Yeah, thanks. I think so too. Here is the blue separated decon denoised, and recombined. It seems to help a bit with the star halos which are worse with the lp filter. It's a 2" glass filter which worsens the halo around the bright stars; maybe it should be placed closer to the camera rather than at the end of the nosepiece(?). Any comments on anything else in the image and improvements I could make most gratefully received. TIA

 

[Astrosurf]

I have this trouble with StarTools. I use some processing in PS as a layer to cover the patches up. Nice image!

Alexxx

[alacant]

2 hours ago, Astrosurf said:

I use some processing in PS

Hi Alexx. That's interesting. If you get a moment it would be great if you could share your layer method. Cheers.

[Astrosurf]

What I do with the layer method is copy the processed PS image, with no patchiness, and paste onto the StarTools one. Click the Mask icon (a square with a circle). You get a white square on the relevant layer on the layer Palette. Make sure this is highlighted by clicking on it. Then I use the Brush tool with black and make the brush large enough for the area. The brush should be the soft type, not hard-edged. I set it to about 20-30% opacity and go over the areas where I want the StarTools image to come through, as it's sharper and with more detail. Keep going over the area for more and more to come through. The brush's soft edges makes sure the effect is graduated. If you make a mistake, just use a white brush and it undoes what you've done.

Let me know if you don't understand anything. This is a rush job!

 

 

[ollypenrice]

I work mostly in Photoshop. When I don't have suffiicient data to support the hard stretch I'm giving an image I can end up with a widespread dusting of very dark pixels, sometimes singles or in small groups. I can select these by using the Colour Select tool. I want just the errant very dark ones. Once selected I hide the selection 'crawling ants' and attack the dark rogues in a number of ways till I find the method which works best. This might be by going into levels and simply moving the grey point slider to the left to brighten them. That's OK on a monochrome channel. Alternatively I can try the Median filter which will give the rogues the average vaue of their neighbours. While experimenting I'll be zooming into pixel scale and right back out to full field to see how it's going at all scales. Sometimes this approach roundly beats the more invasive ready made denoising algorthithms.

I would never, ever, run a denoise routine on an entire image. (Well, as a layer, maybe, but I would only select the darker areas for inclusion in the flattened final version.)

Olly

[Adam J]

3 hours ago, ollypenrice said:

I work mostly in Photoshop. When I don't have suffiicient data to support the hard stretch I'm giving an image I can end up with a widespread dusting of very dark pixels, sometimes singles or in small groups. I can select these by using the Colour Select tool. I want just the errant very dark ones. Once selected I hide the selection 'crawling ants' and attack the dark rogues in a number of ways till I find the method which works best. This might be by going into levels and simply moving the grey point slider to the left to brighten them. That's OK on a monochrome channel. Alternatively I can try the Median filter which will give the rogues the average vaue of their neighbours. While experimenting I'll be zooming into pixel scale and right back out to full field to see how it's going at all scales. Sometimes this approach roundly beats the more invasive ready made denoising algorthithms.

I would never, ever, run a denoise routine on an entire image. (Well, as a layer, maybe, but I would only select the darker areas for inclusion in the flattened final version.)

Olly

Never say never, it can help in certain specific situations.

[ollypenrice]

9 hours ago, Adam J said:

Never say never, it can help in certain specific situations.

I'd need to be convinced. 'Noise' is a story about 'Signal to noise' so the ratio of signal to noise can never be constant across the range of signal. Signal to noise has to be worse where the signal is low.

Noise reduction involves blurring, simple as that. Clever blurring and in assorted flavours, but blurring. In AP the eye is drawn to the bright parts of an image and these usually hold the detail. Why blur the detail? It should stand a bit of sharpening or larger scale contrast enhancement. Personally I would exclude the stars from this sharpening.

Olly

[wimvb]

On ‎2016‎-‎09‎-‎03 at 14:19, alacant said:

Yes. There's a random dither of up to 12 px. It does indeed occur after the denoise.

**Edit. Not that sure now. Here it is without the denoise. Dunno...

If you randomly dither UP TO 12 pixels, your AVERAGE dither is only 6 pixels. Try doubling that to get an average of 12 pixels. An alternative is to dither with a fixed step size of 12 pixels, but a random step direction.

[Adam J]

1 hour ago, ollypenrice said:

I'd need to be convinced. 'Noise' is a story about 'Signal to noise' so the ratio of signal to noise can never be constant across the range of signal. Signal to noise has to be worse where the signal is low.

Noise reduction involves blurring, simple as that. Clever blurring and in assorted flavours, but blurring. In AP the eye is drawn to the bright parts of an image and these usually hold the detail. Why blur the detail? It should stand a bit of sharpening or larger scale contrast enhancement. Personally I would exclude the stars from this sharpening.

Olly

Yeah I wont argue with that but for example hot pixel removal is a noise reduction measure that I apply to my entire image every time I process. Same for removing vertical or horizontal banding, its all noise reduction and all applied to the entire image when required. Flat frame and dark frames and bias frames are whole image noise reduction when you really think about it.

I know we are not really talking about that kind of noise here, I just thought that you had made an overly sweeping statement.

[ollypenrice]

9 hours ago, Adam J said:

Yeah I wont argue with that but for example hot pixel removal is a noise reduction measure that I apply to my entire image every time I process. Same for removing vertical or horizontal banding, its all noise reduction and all applied to the entire image when required. Flat frame and dark frames and bias frames are whole image noise reduction when you really think about it.

I know we are not really talking about that kind of noise here, I just thought that you had made an overly sweeping statement.

Bias, darks and hot pixel filtration are pixel-selective so we can both claim them! They are indeed applied globally but only after the relevant pixels have been identified. This is playing with words, though, and I take your point, of course.  But even with hot pixel filtration and banding removal, I think it's as well to have a care and apply them as layers so as to remove unwanted side effects. Hot pixel filtration can damage stellar cores, particularly in colour images I find, so it's easy to apply it as a top layer and then select and erase the stars from the hot pixel filtered layer. And while the banding removal can sometimes work without side effects it can also produce bleed-like streaks so, again, a layers-and-eraser approach has its virtues. Obviously if it never affects your images in this way then there's no need to bother. 

(By the way, if anyone has an issue with a dead column as we do on one of our CCDs, Noel's vertical banding removal is brilliant. Apply it to a bottom layer, maybe giving two or even three iterations, and then erase the line from the top layer. Works like a charm, way better than half an hour with the clone stamp.)

Olly

 

[alacant]

18 hours ago, wimvb said:

dither with a fixed step size

Hi. That's an interesting one. The closest I can imagine is phd2's spiral. Not sure if it's a fixed step and certainly not random... Still thinking. Do you think that's why I have dark patches? TIA.

[wimvb]

It doesn't have to be random. If you want random, you can randomly renumber your subs or add them to the stack in a random fashion, bu it won't change the outcome a bit. If you dither enough subs with random steps, you can reorder them to give a nice pattern. That won't alter the outcome of the stacking either.

The result of adding (numbers/images/pixels) is the same regardless of the order in which the adding was done. It's the offset between subs that matters. (I always use a spiral pattern for dithering with a step size of 15 pixels. This works for my setup.)

I can't say if your dithering scheme is the cause of the dark patches, you can only find out by testing. Denoise algorithms can cause them too, so can overcorrecting darks, or in camera noise reduction. Have you checked individual subs and individual calibrated frames?

As always: experiment and analyze your results carefully. Or, to paraphrase (= missquote) Olly from another thread: we're pretty much all communicative autodidacts.

Good luck