DSLR noise reduction no darks no flats!

[Bassnut]

I tripped on this randomly.
https://www.youtube.com/watch?v=PZoCJBLAYEs
Its a talk by Tony Hallas on how he produces noise free DSLR images. Its right out of left field, im not sure what to make of it. I havent heard of this method before.

He doesnt bother with darks or flats.
He sees DSLR noise as "colour Mottle", constant large diameter random colour dots 5-10 pixels in size that are the same in each exposure. Ive seen this effect as everyone has, but I didnt know it was constant!.

He dithers each sub by at least 5-10 pixels (during capture), aligns the subs, combines them, and bingo, no noise!, no other noise reduction used!.

Have I missed something?.

[Redshift13]

Might give it a try - if it works it could be a handy processing method.
Wonder how this method would work for nightscapes if you had, say, some trees, a lake or a rock formation in the foreground...

[Shiraz]

that's what Mike S has been saying for years - no calibration, just dither. Dither de-correlates any fixed pattern noise and then it is subject to reduction by stacking. Without dither, fixed pattern noise is not reduced by stacking.

[AstroJason]

Cheers for sharing Fred, very informative. Looking forward to giving this a try. I wonder how much benefit there would be to cooling the camera in addition to using his dithering technique?

Rohan, you could definitely use this technique with nightscapes. You would take all your frames of the sky and landscape foreground. If you're using a static tripod, just pick one of those frames to mask the landscape back in over the top of the stacked image within Photoshop so the landscape looks sharp.

[Redshift13]

After doing some more reading on the subject this morning, it seems to be the case that to achieve dithering, your camera needs to be attached to a telescope mount, with the dithering controlled by software. My nightscape setup is much more simplistic - camera on a tripod.
So I'm guessing that this makes 'manual' dithering much more difficult, if not impossible?

[AstroJason]

Yes, I think it would make it more difficult to slightly pan the camera on just a simple tripod. Although, if you can keep track of where and how far you have panned I think this technique it would still be possible. Worth a try anyway.

[Camelopardalis]

With a manual non-tracking mount surely the rotation of the Earth is dithering for us automatically? :D

[AstroJason]

Very good point Dunk!

[strongmanmike]

What Ray said :lol:

Actually, I just last night dithered..?..without dithering.

Like most, without guiding I get slow drift due to polar missalignment and tube/mount flexure etc so because I was doing lots of very short subs I didn't bother getting the software to move the scope between subs ala my normal dithering (this would have added too much time as I was trying to image through breaks in the clouds :doh:) I just let the dozens of quick subs come down and over time the slow drift created enough shift between subs (with out elongating stars) so that when I median combined - bingo noise gone :D

Got THIS (http://www.pbase.com/strongmanmike2002/image/159445895/original) image which proves you can image when it is cloudy :lol:

Mike

[DJT]

Great share, Fred. I have a 60d and a Teegul Patrol I havnt used yet so will have a play with this approach over the next couple of weeks.

[Poita]

It would be interesting if you could use a Pentax or similar camera that moves its sensor (normally for Opitcal image stabilistation) and program it to automatically dither. The amount would be exactly known, so in theory, if the manufacturer wanted to, this could be implemented in the camera itself.

With the nightscape photography, perhaps mounting the camera on a micrometer stage or macro set might work?

[astrodavid]

an interesting read - thanks for posting the link.

[rustigsmed]

scroll down to Ivo's comments.

http://www.cloudynights.com/topic/482133-tony-hallas-on-using-a-dslr/

Cheers

Rusty

[Bassnut]

Thanks Russell, Ivos post in cloudynights is very interesting, WOW !!!!.

edit: its worth reading the comments after Ivos post too.

I thought something wasnt right, it all seemed so out of left field.
Aint astrophotography and what ppl do with it interesting!.

[multiweb]

Yes, Ivo knows his stuff and is on the money. A lot of quackery in that video. Pretty misleading actually. :shrug: There's no freebie in data acquisition and calibration.

[Geoff45]

I find myself agreeing with Ivo.
Geoff

[Poita]

Still, a good reminder of the advantages of dithering.

What would be great is if someone had some really nice data from a DSLR and ran it through the traditional method and the Hallas method to show just what is lost doing it the Hallas way.

[Shiraz]

Agree, lets not throw the baby out with the bathwater. Looks like some of the processes that Hallas used are not very sensible for high resolution imaging, but the basic idea of using dither instead of calibration to remove fixed pattern noise is very powerful. It would be great if someone has a set of dithered DSLR subs that they could stack using either full calibration or dither+stack without cal to see what happens - hope this is not too far from your original post Fred?

For comparison, I just did a test to see how effective the two stacking methods are on 30 subs of dithered mono, temp controlled data. With dither stacking I did a bad pixel removal followed by a simple average stack and no calibration at all. With standard stacking I used an average stack with bias (100 bias subs), darks at the same temperature (70 darks) and flats (100 flats). The full calibration stack image showed evidence of hot pixel effects (no outlier rejection was used) so a bad pixel map was also used to remove the odd warm pixels from the images after the initial calibration.

the results: Dither really can do the job just as well as a full calibration - left image is dither only, right image is full calibration. There is a slight brightness difference where I did not get the offsets exactly equivalent, but that's it. The SNRs from the two processes were effectively identical as measured by the sky background variance.

[Poita]

Great stuff Ray, perhaps we should start a new thread inviting people to upload their data and compare results.

[multiweb]

Hi Ray, no doubt dithering is a proven solution and a very powerful tool to remove noise and to some extend dust motes and other static patterns.
All the processes highlighted in the video aren't bad, but it is a bit rich to talk about noise reduction with non linear data. We might as well stack JPEG files and call it a day hey? :nerd:

[gregbradley]

This is similar to the debate about no need for calibration with Sony ccds.

Until you do the experiment and see that a flat actually adds noise to the image.

So this is relevant to noisy sensors more than anything. I would not have though Tony spent a lot of time with DSLRs. Perhaps his recent 60Da got him interested or he is trying to appeal to a wider audience than the CCD mob.

His DVDs on image processing are very good. I have watched them all. I don't particularly like his highly contrasted image processing but at least I know how he does it. He's like the original Pix Insight style image processor. He also several useful techniques that I use from time to time but some of it is getting a little dated now with advances in some software making the tricks redundant.

Greg.

[Shiraz]

Hey, don't get me wrong, I am not advocating using the Hallas workflow, just saying that the main thrust of his presentation (and the title of this thread) is to do with DSLR noise reduction without using darks and flats - this part may be very sensible.

It would very sad if rejection of some parts of his presentation resulted in an overall impression that this means that the idea of using drizzle to get rid of noise (as an alternative to darks and flats) is equally questionable - it clearly is not. I wonder if drizzle stacking with hot pixel removal might possibly be even better than full calibration in the situation where darks, flats and lights are not fully temperature matched and where there is likely to be some form of low level manipulation of RAW data (eg a modern DSLR). I don't recall anyone ever posting any results either way.

[Amaranthus]

Ray, why would you want/need to temperature match the flats?

Re: Noise, one can also do Gaussian pixel blurring to smooth out flat noise (but still retain sufficient definition of dust motes). I generally take 100 flats per filter/session to minimise noise, which is possible when using a light box.

[multiweb]

Not at all. I know you don't. :)

[Shiraz]

Hi Barry. The flats need to be dark subtracted (or at least bias subtracted). in either case, would expect that dark/bias level would vary with the temperature of the camera, so you could be subtracting a different bias from that inherent in the flat. Likely to be minor cf the mismatch of dark and lights, but still possible.

agree that if you do dither, hot pix rejection plus dither stacking gets rid of noise and then you can use smoothed flats to clean up any dust etc without adding more noise.