On adding "diffraction spikes" post-aquisition

[Rick]

Me again, the pre COSTAR images used de-convolution algorithms to remove as much as the optical effect as possible caused by the original mirror miss figuring.  That is now fixed and has nothing to do with diffraction spikes.

Deconvolution is a process by which a system's impulse response can be removed (up to a point). The diffraction spikes are part of the system's impulse response. The way deconvolution was applied to the pre-COSTAR Hubble images was by using a desired impulse response as an end target. There are other ways deconvolution can be applied. There are also a great many other signal processing algorithms which can be applied to enhance an image, and the Hubble team have made use of quite a few of them. They've even done a fair bit of the original research into them. Some are more successful than others.

None of this makes the addition of fake "diffraction spikes" to an image which didn't have them originally any more acceptable.

(Moderator note: I'm going to split this discussion out of this thread because it bears very little reference to the original thread.)

(Moderator apology: Sorry, I had to lock the thread long enough to split the topic. Normal service is now resumed.)

[Rick]

I was watching a Hubble/ESA disc last night & noticed the spikes on some of the images.

Those are genuine artefacts of the optical arrangement. Generally you don't want to remove them because they tell you something about the nature of the object. The Hubble folk know this.

[Rocket Pooch]

Actually, deconvolution (restoring) is an algorithm-based process used to reverse the effects of convolution (mucking up of) on recorded data.  In the case of the Hubble scope the convolution of the final image was caused by the miss figuring of the primary.  Therefore, the Hubble guys wrote bespoke deconvolution routines to restore the picture back to what it should have been given a perfect mirror, or as near as possible anyway.

Have a look at http://en.wikipedia.org/wiki/Deconvolution under "Deconvolution and Optics", it points out that a) you need to know what the reference point of the original image should have been and b) the routine needs to be specific to the source data for it to work well.

This has nothing to do with the Hubble guys removing diffraction spikes, if you look at the images of Supernova 87a (I think that's what it was called) the early pre-COSTAR images did not have well defined diffraction spikes, but post COSTAR they did.  This does not meant they removed them in the past; just that the routines written could not restore the images 100%.

Quote thingy "Deconvolution is a process by which a system's impulse response can be removed (up to a point)."  I assume you agree with the above with this? If not what does it mean?

Quote thingy "None of this makes the addition of fake "diffraction spikes" to an image which didn't have them originally any more acceptable."

To whom, just because I like my other half it does not mean that you do?  I believe someone probably wiser than both of us said "Beauty is in the Eye of the beholder", someone else said "don't over criticize other peoples work until you have tried it yourself" (me yesterday to one of my coders) I don't think Fay's image was meant to be technically significant, if it was then Flats, Bias no filters etc would come into play, its just a pretty picture with a bit of added bling!

Mike, is that ok?  FIGHT FIGHT FIGHT!!!!

[JohnP]

Well I'm actually learning quite a lot about deconvolution/ convolution - never understood any of it before but starting to make a little sense.. I'm looking forward to when someone starts talking about my best mates Lucy-Richardson & Maximum Entropy......

John

[Rick]

Actually, deconvolution (restoring) is an algorithm-based process used to reverse the effects of convolution (mucking up of) on recorded data.

Yep, that's a very good way of describing it. I was trying to say exactly the same thing, but fell into the trap of using the terminology folks at work would use.

Diffraction spikes are one of the ways an image gets "mucked up" by an optical system. However, they can be very useful. The Hubble folk knew what sort of diffraction pattern (spikes and all) they were expecting to see from the pre-COSTAR scope. They looked at the pattern the scope was returning, and tried to produce an algorithm which would take that "mucked up" diffraction pattern and bring it back into sharp focus. The trouble is the algorithm is very sensitive to noise, so it's not possible to do the job perfectly. That's at least part of the reason the pre-COSTAR diffraction spikes are a bit fuzzy.

Back when the first amateur astronomers to experiment with digital imaging were presenting their first results, there was a short talk at the BAA about the possible ethical problems digoital imaging might pose. The biggest fear was that it would be easy to add fake detail to a digital image; certainly much easier than it is to add fake detail to a photographic plate, and harder to detect. These days we rely entirely on the integrity of our observers to present an accurate image of the phenomena they've observed. Adding fake diffraction spikes is a very minor deception. Folks who've used the Astronomy Tools software may know it has a variety of "add spikes" options, and may recognise the effect in others' images. I didn't know, and didn't recognise the effect for what it was. I was fooled into thinking it was a genuine effect. Others who come to look back on the images years from now may not know, or may have forgotten, and may end up being fooled too. As I say, it's a very small deception, but it is a deception, and to do good science we should avoid even the hint of deception.

[Rocket Pooch]

Well I'm actually learning quite a lot about deconvolution/ convolution - never understood any of it before but starting to make a little sense.. I'm looking forward to when someone starts talking about my best mates Lucy-Richardson & Maximum Entropy......

John

Next week mate after half a bottle of whiskey.

[Mike]

Mike, is that ok?  FIGHT FIGHT FIGHT!!!

Hey all I said was ".....Not sure about the Christmas Card stars you've 'created' there though......."

[Rick]

To be honest, I'm still completely boggled that the author of Astronomy Tools thought that options to add fake spikes would be a good idea.  In a paint package, where you go to let your inner artist is have a play, weird visual effects like spikes, lens flares, halos and such can work wonders on fantasy scenes, but they really shouldn't have any use when you're trying to enhance an image to bring out the real detail hidden in the raw data (which, as I understand it, is the aim of Astronomy Tools (or is it?) )...     

[Tony G]

Still after all that, I personally still like the diffraction spikes

Tony G

[JohnP]

Rick... can't believe you are still going on about this...... Just FYI here is what the author of Astronomy tools Noel Carboni (a brilliant astrophotography by the way- some of the best images on the web) says about them...

Star Diffraction Spikes - High-end Ritchey-Cretien telescopes (like the Hubble) make them because of the cross-shaped vanes that hold the secondary mirror in position... Some folks make them by tying strings across the aperture... One thing's for sure - they look cool! And the highest quality astroimages often have them. I own a Schmidt-Cassegrain telescope that doesn't make them optically, so I decided to craft a way to make spikes digitally. These spikes are created proportionate to the size/brightness of the stars and are the color of the stars, so they look very natural.
If your telescope doesn't make diffraction spikes and you want them in your images - or you'd like to enhance the ones you get optically - I've included four different actions for making them. You have the option of adding them to your images in the quantity you want, from spikes on even the smallest stars to spikes on only the biggest, brightest ones.  

Tony - I'm with you.....

John

[RobertM]

Aren't we missing the point here ?  If we're out to produce pretty pictures (aren't we ?) then we adjust the colours by eye so they look pleasing (that means subjective in my book).  Doesn't adding diffraction spikes come under the same heading i.e. a personal thing.

At the end of the day it would be difficult to do much real science with any of my images - I don't use Johnson or any other astrometric filters or even perform proper calibration.  As for knowing the set point temperature, does anyone know what their CCD cools down to or even whether it's stable from image to image ?

Personally I like the spikes, the image looks better for them.

Just my penneth ...

[Rick]

Quoting Noel Carboni on diffraction spikes: I own a Schmidt-Cassegrain telescope that doesn't make them optically, so I decided to craft a way to make spikes digitally. These spikes are created proportionate to the size/brightness of the stars and are the color of the stars, so they look very natural.

That's exactly why I have a problem with it. It's not revealing something that's there already. It's adding something that wasn't there before. Once you start down that road, where do you stop? Will you sneak in a fake comet, supernova, TLP, or some such? I hope not, but once I know you've added something that wasn't there to an image, even if it's as innocuous as a fake diffraction spike, then you're sowing the seeds of doubt in my mind. That's why I am still going on about this. I want to know whose images I can trust. One thing's for sure, having read that quote I'll certainly trust Noel Carboni's images a lot less in future, however pretty they look.

Aren't we missing the point here ?  If we're out to produce pretty pictures (aren't we ?)

Are you really investing all that cash and effort in kit and imaging time just to produce pretty piccies, without even the slightest hope that maybe one day you'll image something nobody else has yet seen?

then we adjust the colours by eye so they look pleasing (that means subjective in my book).  Doesn't adding diffraction spikes come under the same heading i.e. a personal thing.

Colour balance is very subjective. Different screens show colours in different ways, and peoples' eyes see colours different ways too. I even see some colours slightly differently depending on whether I use my left eye or my right. So for astro images you (hopefully) note which filters went into which channels and let folk decide whether the colour mapping works. Feed the red into green and the green into red if you like. The underlying image data remains the same, though it'll look a little weird. Sometimes weird colour mapping is just what you need to reveal some otherwise invisible detail. So no, adjusting the colour balance isn't the same sort of thing.

At the end of the day it would be difficult to do much real science with any of my images - I don't use Johnson or any other astrometric filters or even perform proper calibration.  As for knowing the set point temperature, does anyone know what their CCD cools down to or even whether it's stable from image to image ?

One thing with astronomy is that you never know when something unexpected and spectacular might happen. Just suppose you were out imaging one night. A week or two later there's a headline on the news saying something like "Spectacular supernova in...", and you realise it might have been in one of the images you made that night. Let's say you'd used "add spikes" on the final image, and thrown the raw frames away because it was "just a pretty picture" and you'd got it as pretty as you wanted. One of those spikes just happens to obscure the supernova's position, so you figure you saw nothing. Who's been cheated? Suppose instead you realise you do have a pre-discovery image of the supernova. Your image would be of real scientific value even if you havn't got the whole set-up calibrated to N decimal places, provided you have a few basic notes (like exactly when and where you took the image) and you've still got the raw frames or you've not mucked the final image about too much.

Think it'll never happen to you? I was out photographing the southern sky from Kenya about 8 hours before SN1987A in the LMC was discovered. I didn't quite manage to get the LMC into my camera's field of view that night because there was a tree in the way and I didn't quite stay up late enough...

[Mike]

.....Let's say you'd used "add spikes" on the final image, and thrown the raw frames away because it was "just a pretty picture" ......

Oooh that's a big no no, even if you are just making pretty pictures. Despite the small number of images (approx 25?) I have posted, I have taken 2691 FITS frames over the years. Like with any photography, the majority of your pictures don't get used in the end as they are no good. However, i've still kept them all in case I come across a better way to process, etc. in the future, plus, as you say, there may well be some useful astronomical data in there.

[Ian]

I think a lot of the issue is around pretention (and I don't use that word in a negative way). I am realistic that my photos (if they happen  ) are not created with the care and attention that images taken for scientific purposes demand. They are more an enhancement to my ability to observe. In my back garden, which is about as dark as Sidcup gets, I get the slightest hint of some objects, but they're the bright Deep Sky objects. A CCD image is a bit like the consolation prize, "look what you could have won".
But, as Rick says, there is the slight, but real chance that an image may have real scientific value, and then the raw data becomes scientifically interesting, and it's technical shortcoming can be compensated for.

However, in the meantime, I like to create images that give an impression of what is up there to see, often to show non-observers (like my wife who really doesn't understand the cost and extreme cold our hobby involves sometimes). That's what I mean by "Coffee Table Image". In that case, a small amount of scientific "dishonesty" to make an image look nice is acceptable. I don't have astronomy tools, I might one day, and I don't know if I'd use that particular one, but I think if it makes an image look nicer, and that's my aim for that image, then I might use it. (Ooh, lots of commas. It's like a camper van convention)

But, as Mike said, raw data is just that, and is always retained...

[JohnP]

Rick - As an astrophotographer (not sure if you take images) you never throw raws away. The thing that inspired me was seeing all those wow images in the glossy mags etc... & yes at the end of the day my number 1 priority is to produce a 'pretty' image of some distant object - no way do I spend hours scanning my raws identifying all the stars etc. for some undiscovered object. Obviously, if something jumps out I will investigate further by that's about the level of science that I do & I believe that is true of a majority of imagers. If I wanted to present my images in a scientific manner then I would just post the raws (which I don't think would appeal to many people) - if you think about it even doing a simple curves adjustments is manipulating the data as you are changing the brightness of one area unproportiuonally to another.

Are you really investing all that cash and effort in kit and imaging time just to produce pretty piccies

YES - I'M AFFRAID SO..... & I'm also planning trips to south of France next year so I can get even PRETTIER ones

John