Dithering

[Carole]

I've read a lot of threads about dithering recently, both on here and other forums.  I think I've sort of got the idea what it is, but how to achieve it is another matter.

I gather this is to shift the position of the object being imaged fractionally in your field of view in order to erase/overlap hot/pixels and other blemishes, I guess in much the same way as taking an AVI would do when it is stacked in Registax. 

In a very simplistic way would turning the guiding off for a short while and then re-starting achieve this if there is slight drift without guiding?

Carole

[MarkS]

In a very simplistic way would turning the guiding off for a short while and then re-starting achieve this if there is slight drift without guiding?

Yes - that would work - your understanding is correct.

Before I discovered Nebulosity I used to shoot a few frames; switch off guiding; manually slew the mount by a few random pixels; switch guiding on at the new star position and shoot another few frames.  Very laborious but it did the trick.

Mark

[Carole]

Thanks Mark. 

How often would you say this needs to be done to be effective?

I have sometimes had to turn the guiding off anyway because a cloud came over or something, but it's only been about once an hour on average.

I guess finding some software that would do it for you would make life easier.

Carole

[mickw]

Your guiding turns itself off anyway  

I think nebulosity will do dithering  during capture

Nope it doesn't - but when you link phd to nebulosity you can pause guiding during image download - I assume that would work the same.

I have seen a capture program that does dithering but can't remember it 

[Carole]

Your guiding turns itself off anyway

Actually it does continue guiding now unless there is a cloud or something, but not all the guiding is good.

Still it's much better than it was before.  Just means I waste some of the subs which is annoying.  

Have now got myself a regulated power supply, so I am hoping that might help.  Not had a chance to use it yet.  

Carole

[mickw]

Just checked - Link Nebulosity to phd and you can dither

[Carole]

Just checked - Link Nebulosity to phd and you can dither

Interesting, will bear that in mind.

Carole

[MarkS]

Just checked - Link Nebulosity to phd and you can dither

Interesting, will bear that in mind.

That's exactly how I currently dither!

[Carole]

If I try to link Nebulosity to PHD, does this mean Nebulosity has to be my capture programme? 
I am quite happy with Canon Utility which I currently use.

Carole

[mickw]

Yes

[The Thing]

Hi Carole,
On this thread we have been discussing alternative capture programs that do or will support PHD dithering. Canon Utility is fine but it's much more convenient to be able to see your images immediately not to mention the other useful features these programs offer.

http://forum.orpington-astronomy.org.uk/index.php?topic=7166.msg45439#msg45439

Duncan

[RobertM]

On the subject of dithering, Fay did a lot of that on Saturday but I don't think she used PHD. 

[doug]

This is a new one for me.  As I`m not an imager, I always thought that dithering was to be uncertain as to something, or to be hesitant. I didn`t know it was an astronomical term; is it a new one??   

Doug.   

[Carole]

Hi Duncan, I think that thread was one of the ones I had been referring to.  

it's much more convenient to be able to see your images immediately

I can see my images immediately with Canon Utility, as soon as the 5mins sub (or whatever) has finished they download in a couple fo secs and I can see them straight away.

I think I need to ponder on this.

or to be hesitant

Doug, I think in effect Dithering in imaging does mean a slight hesitancy (or rather pause and wait) in guiding.  Probably where it got its name.  

On the subject of dithering, Fay did a lot of that on Saturday but I don't think she used PHD

Carole

[MarkS]

Doug, I think in effect Dithering in imaging does mean a slight hesitancy (or rather pause and wait) in guiding.  Probably where it got its name.  

A pause and wait in guiding would have no effect if your mount had perfect tracking. Instead, a dither is a deliberate slight movement of the star field across the image frame between subs.

But where does the word come from?

According to: http://en.wikipedia.org/wiki/Dither
"Dither is an intentionally applied form of noise used to randomize quantization error, preventing large-scale patterns such as "banding" in images."

I think the term was first used in printing - where you wish to create a wider grey scale or a greater number of colours from a limited palette.  Its usage in astro-photography is slightly different - it is something that happens at the acquisition phase.

In astro-photography, dithering can prevent the "square star" look in images or banding across images.

Mark

[MarkS]

Nope it doesn't - but when you link PHD to Nebulosity you can pause guiding during image download - I assume that would work the same.

Pause of guiding during download is not dithering.  The reason guiding can be paused during download is to help prevent electrical noise degrading sensitive operations such as reading from the CCD.  When guiding is resumed after download, it goes back to the same guidestar in the original position - i.e. there is no interframe movement between subs.

Dithering is where the guide star is moved a tiny distance across the frame before the next sub is acquired.

In Nebulosity both functions are accessed from the same menu because both functions require the link to PHD.  

Mark

[doug]

     Mark ..... I was in the printing trade for forty nine years.  The only dithering we did was when we had a cuppa and a cheese roll !!!!!
But thanks for the technical explanation. 

    Doug.

[mickw]

Mark
Thanks for that and the Wiki link - very interesting and explains a lot.

Should dithering be applied as a matter of course or only if needed - a case of if it ain't broke don't fix it ?

[MarkS]

Should dithering be applied as a matter of course or only if needed - a case of if it ain't broke don't fix it ?

The problem is that you don't usually know if dithering is required until you process your data and by then it is too late.

Generally speaking, in a non-observatory setup, slight inaccuracies in polar alignment and flexures in the mount/rings/scopes mean that you get slight drift and rotation from image frame to image frame even if the guiding is accurate. This has a very similar effect to dithering.  But is is not guaranteed that this will always happen.  So I use dithering as a matter of course for the Canon.  It may not be so important for dedicated astro-CCDs.

Mark

[Rick]

I think the term was first used in printing - where you wish to create a wider grey scale or a greater number of colours from a limited palette.  Its usage in astro-photography is slightly different - it is something that happens at the acquisition phase.

In printing (including in a lot of computer printers) it relates to the way a screen is applied to a photograph so that it can be printed using ink of one colour (or, for colour, three or more), so each pixel turns into a very fine-scaled pattern of dots (or sometimes dots of varying sizes). Astrophotographers are just doing it backwards...

[mickw]

The problem is that you don't usually know if dithering is required until you process your data and by then it is too late.

That was in the back of my mind 

So in the field (or any iffy obs.setup) it would help to use it as part of the capture process

I think the term was first used in printing - where you wish to create a wider grey scale or a greater number of colours from a limited palette.  Its usage in astro-photography is slightly different - it is something that happens at the acquisition phase.

In printing (including in a lot of computer printers) it relates to the way a screen is applied to a photograph so that it can be printed using ink of one colour (or, for colour, three or more), so each pixel turns into a very fine-scaled pattern of dots (or sometimes dots of varying sizes). Astrophotographers are just doing it backwards...

I think this demonstrates the effect using 3 colours (which could be greyscale) - they are dithered (blended) to produce a different value.
Although this wouldn't be the first time I was wrong 

[RobertM]

Dithering is where the guide star is moved a tiny distance across the frame before the next sub is acquired.

That's not strictly correct.  The guider is used to remove tracking errors so the image remains in the same place on the capture chip.  Dithering is independent of guiding and does not need a guide camera.  If you are using a guide camera then a random offset is applied to the guide star from it's actual position, the guide software treats that as a guiding error and acts to center the guide star again.  The offset positions used in the dither operation are randomised about the original guide star position so acting to reduce noise without loosing any image area (around the edge).  The one disadvantage of this is that if guiding is stopped and restarted the dithering will be about the new guide star position (which was offset by the previous dither operation).  There will also be a short time after the dither operation has started where the guiding system will have to settle down to a steady state again before the next exposure can be taken.

It still has a noticable effect with cooled cameras though the results can be quite subtle.  Also unlike dark and bias frames the technique acts to reduce all captured noise.

I use it by default if at all possible.

Hope that helps
Robert

[MarkS]

Dithering is where the guide star is moved a tiny distance across the frame before the next sub is acquired.

That's not strictly correct. 

Maybe it's not strictly correct but it is exactly what you see when observing the output from the guide camera chip (at least in PHD).  The guide star jumps to a new random position in the guide camera frame calculated by the dither algorithm just before the start of each exposure.  It is equivalent to how I used to do it manually - stop guiding; slew scope randomly by a few pixels; begin guiding at new guide star position.

Mark

[RobertM]

Yes, but my point is that you don't need guider to do it.  Offset of the guide star is the easiest one way to achieve it with a guider.  For example, in MaximDL dithering can be achieved either by direct mount movements or via the guider if connected.

As well as multipixel dithering, dithering at the sub-pixel level can also be used to improve spatial resolution for undersampled images.

[MarkS]

Yes, but my point is that you don't need guider to do it.  

True, but realistically we all do guiding because of periodic error and polar misalignments.   And once we have guiding switched on, the only way to perform dithering is to interface to the guider in some way shape or form.

Mark

[doug]

     The image that Mick has put here with the red,  green and blue colours on the wall .................  where they overlap the colours change to yellow, magenta and cyan.  These are the colours of the inks that printers use for basic colour printing with the addition of black to usually give intensity of shadow areas.  Of course, one can always add other tints to enhance the reproduced picture, such as pink, grey, light greens, etc.  We have even used a transparent white ........

     It`s a long time since I thought about the printing process ........

     Doug.... with ink still on his fingers. 

[Rocket Pooch]

I'm enjoying this, but I would like to point out that under sampling will not get fixed by dithering.

[MarkS]

I'm enjoying this, but I would like to point out that under sampling will not get fixed by dithering.

Actually, if you use dithering during acquisition and drizzle during processing it goes a long way to fixing undersampling.  Add a bit of post-processing deconvolution, then "hey presto!"

Mark

[Rocket Pooch]

Hi,

Bit busy at the moment, but if you have undersampled a small star to less than one pixel, it does not make any difference what you do with it, its still the shape of the pixel. Which is one of the reasons I don't like uba wide field images, they just look like a field of little blocks, not stars.

Chris

P.S. http://www.ccd.com/ccd113.html

[mickw]

Wouldn't dithering round off the blockiness by distorting the square pixel (assuming a star the size of one pixel) ?

[Carole]

this has been an interesting discussion, 

drizzle during processing

But we need one now on Drizzling which I haven't a clue about. 
think we need to start a new thread for that, so will start one. 

Carole

[RobertM]

Hi,

Bit busy at the moment, but if you have undersampled a small star to less than one pixel, it does not make any difference what you do with it, its still the shape of the pixel. Which is one of the reasons I don't like uba wide field images, they just look like a field of little blocks, not stars.

Chris

P.S. http://www.ccd.com/ccd113.html

Sorry Chris, usually I agree with you but in this case that will  only be the case if all the light from a star falls exactly on one pixel (which is rare but maybe that's what you meant in which case I fully agree).  If it falls on more than one pixel then the star centroid is offset by a sub pixel amount and can therefore be derived at a greater accuracy.  It's this that we rely on for sub pixel guiding accuracy.

Robert

PS. Carole, you are a whole heap of trouble !!!

[Carole]

 

[MarkS]

Sorry Chris, usually I agree with you but in this case that will  only be the case if all the light from a star falls exactly on one pixel (which is rare but maybe that's what you meant in which case I fully agree).  If it falls on more than one pixel then the star centroid is offset by a sub pixel amount and can therefore be derived at a greater accuracy.  It's this that we rely on for sub pixel guiding accuracy.

Indeed, and this is why many people recommend guiding on a slightly defocused star - to ensure the star covers a greater number of pixels.

Mark

[Rocket Pooch]

Rob and Mark,

I agree with you both, if the star is on one pixel, as I suggest a larger box is a larger box. 

Still holding out on the defocused guide star, but I do understand why people do that as well, esspecially if they are guiding with a very short focal length guide scope, then spreading the star over at least 4 pixels is a good move.

Chris