Reprocess of Jupiter "GRS Hoovering" image.....

[Kokatha man]

Hi all.....further to other comments I've made etc I am endeavouring to quantify/qualify processing techniques with a view to understand more fully not only the elementals of various programs' "modus operandi" (ie how things actually work) but also how the various controls (including those less commonly utilised) influence the image outcomes.....

Obviously a task which will be ongoing and require me to select various avi's I take which I expect will produce hi-res planetary images, applying the various controls selectively to develope a flexible procedure that offers variety and alternative options within these methodologies to "wring" the utmost out of any one avi.....

These will primarily centre around Registax 4&5, AstraImage Pro and P/shop CS3.....

As stated beforehand, hopefully this ongoing exercise will allow others to appreciate a little of the "why" of what they are doing as well as using reliable procedures.....I find I need a bit more of this "why" as opposed to adopting rote-like methods that create "brick-wall" outcomes when things don't seem to function as expected....!

Here is the first processing of Friday's Jupiter (which appears in another thread in this section of SGL, and then the latest processing of the same avi that I have just finished doing.....imho the 2nd is a better process except that with "exposure/gamma" control adjustments in CS3 I believe I slightly over-darkened the limb of the planet in this 2nd attempt - apart from that I believe it a better processing.

Let me know what you think.....constructive, positive and polite responses please - not "know-all" diatribes of the kind that only say " I use so and so setting in this adjustment" without rationalising or explaining it - that assists people very little unless you are able to expand upon it a bit....!

If you are interested in honing your planetary imaging (as I am) then throw a comment or two in about my perspectives or your own - it could help others too....!

I've decided to post this here - the moderators may deem it more suitable elsewhere but I feel that from my perspective that it is better to post a few processing variations on any one avi/image and have people comment or ask questions.....then move on to another image to keep the interest up and give opportunity to discuss more or different points/perspectives in a new thread with new images.....:)

[omneferuss]

the second is definitely sharper.. and stronger hue/saturation (??) shows finer detail hidden in the first ..that's all I'm qualified to say, having no idea about processing/softwares as such, myself

ripper progress Km

[minimalist]

Yep, definitely the second for me too, the extra detail without any extra noise is pretty impressive imo. Is the difference a result of the stacking or post processing?

If you can, a pic of a typical frame out of the camera and the final stacked image before any post processing would really help imo. This would give the rest of us a good baseline as to the quality of the data and the results of the various processing stages.

Btw, ever tried avistack? In my case, it's saved my bacon quite a few times, as it does dozens to hundreds of reference points in comparable times as it took me to do just a few in registax. The latest version boasts the ability to read just about any codec, although I haven't been successful getting that part to work. I end up using Vdub to transcode my avis, which is no biggie really, just takes a few minutes and since it's uncompressed, no artefacts are introduced in the frames.

Daniel

[AlphaTau]

Not that I know much but the second one for me too. Fantastic detail in the image!

Andrew

[Kokatha man]

Yep, definitely the second for me too, the extra detail without any extra noise is pretty impressive imo. Is the difference a result of the stacking or post processing?

If you can, a pic of a typical frame out of the camera and the final stacked image before any post processing would really help imo. This would give the rest of us a good baseline as to the quality of the data and the results of the various processing stages.

Btw, ever tried avistack? In my case, it's saved my bacon quite a few times, as it does dozens to hundreds of reference points in comparable times as it took me to do just a few in registax. The latest version boasts the ability to read just about any codec, although I haven't been successful getting that part to work. I end up using Vdub to transcode my avis, which is no biggie really, just takes a few minutes and since it's uncompressed, no artefacts are introduced in the frames.

Daniel

Hi Daniel - as I said, my own image processing analysis will centre around R/stax (4 or 5) AstraImage and CS3 programs.....and later today (Oz time ) I'll try and quantify the various steps in each of these programs so that anyone interested can see the variations I've made.

Also, I will "bone up" a little more on my scant working understanding of how the various procedures work if I can to (hopefully!) back up certain appraisals I have as to why "such and such" is worthwhile etc.....although I have to say that this aspect will be more speculatory than factual owing to my (current) extremely limited understanding of the "nuts & bolts" of such functions!

Perhaps others who may have an insight into what is actually happening in the "Optimise" stage (for example) may be able to have input therein to assist those who require more understanding of the "nitty-gritty" there (I certainly could!?!)

That's not to say that other programs with their own methodologies are any less important.....but obviously within the (more than large at present) ambit of my own experiences/experimentations it would be only leaning towards those areas I previously stated I wished to avoid to start canvassing (for example) avistack etc.....

Perhaps you might like to pick one of your own images that you feel is worth exampling and give a run-down of "A to Z" with maybe some variations and thoughts to illustrate your points....?

My whole purpose is really to "open up" the debate on methods to achieve/develope hi-res images and the "more the merrier" re other's own experiences.....but examples please....!:)

Briefly (becuase I'll try and put something comprehensive in as soon as possible) the amount of frames stacked in each is identical.....that is going to be a relative constant: and as to the Q about "stacking or post-processing" - obviously it is in these functions (and importantly I think the order therein) that some of my hunches for better processing and final image quality is found.....

As for VirtualDub and ninox (ppm-centre) I must confess that I have found neither of much extra benefit.....and yes, mulit-point alignment in R/stax is an extended process that (for planets) I have not found particularly helpfull.....:D

[brianb]

mulit-point alignment in R/stax is an extended process that (for planets) I have not found particularly helpfull...

Really this depends on your image scale. If you're able to get Jupiter images ~400 pixels across that contain sufficient detail to use multiple alignment points you will almost certainly get better alignment using multi point alignment than you will using a 512 pixel square alignment area to contain the whole planet. And the processing will be faster too. If you are working at an image scale of > 512 pixels to the planet diameter there really isn't a choice, default single alignment probably won't work well, "onion skinning" artifacts will almost certainly arise.

When I was processing my Jupiter AVIs of August 9th with an image scale of only ~160 pixels / diameter I found that the best results were obtained using two point alignment, using Callisto and its shadow as the alignment points, 32 pixels square.

[Kokatha man]

G'day Brian - As stated beforehand I didn't really want to get into any dogmatic assertions in my postings nor make it interpreted as some sort of didactic tome.....I respect Daniel's comments re his use of avistack but as I also said my comments and any ambit I canvass really revolves around Registax, AstraImage and CS3 processing programs.

Not because they are everyone's tools (nor through any association/pecuniary interests thereto:eek:) but because, as I also noted, this attempt by me to chronicle various processing strategies is going to be a lengthy, arduous task to say the least.....expanding to encompass other options would make it patently impossible for me....!

My own experiences and observations with both VirtualDub and ninox are just my own (honest) experiences with said.....as is the point you seem to wish to take issue with re multi-point alignment.

Frankly, I have never been able to enable the single 512 alignment box in any version of registax I've used, so I don't quite understand the validity of that particular statement.....but re multi-point alignment, if the pik you have posted above is your "August 9th" image you refer to then I can see why multi-point is easily able to fix upon those 2 (most) clearly-defined features to give a satisfactory and relatively speedy alignment in the circumstances.....possibly my problems with m/p may have been due to the much greater amount of detail involved, slowing the alignment process down very considerably amongst other issues.....

Onion rings are a contentious issue no doubt.....I haven't found any definitive single cause as the culprit but I do feel that there are ways to avoid it.....none of which seems to be an alignment issue imho.....

But look, feel free to construct your own methodologies with examples to illustrate your convictions Brian - I can only articulate my own experiences and appraisals thereof, and concentrate upon that which I know or discover, rather than that which I don't, if you appreciate my meaning.....but hey - aren't you making a habit of contradicting me and using it as an opportunity to slip your Jupiter image into my threads - or are they different ones each time?!?!?:)

No worries bro, cheers, Darryl.

ps: I don't think I'll get to post my "processing method #1 (1st installment) tonight now....!

[brianb]

.but hey - aren't you making a habit of contradicting me and using it as an opportunity to slip your Jupiter image into my threads

Sorry if that's the way you read me. No intention of offending. And certainly there is more than one way to skin the cat, whatever works best for you may not for me with different seeing / equipment issues ...

I wish I could make images as detailed as yours.

As for "onion ringing", my analysis has been that Registax is misidentifying minor projections from the belts causing some frames to be grossly misaligned. And I've had the worst problems with it when trying to put the whole planet into the alignment box when the individual frames have lacked consistent sharp features. Breaking the AVI down into frames using virtualdub then aligning & cropping them in ninox seems to help solve the issue for me, and my theory seems to be consistent with my observations. Your mileage may differ!

[michaelmorris]

As for "onion ringing", my analysis has been that Registax is misidentifying minor projections from the belts causing some frames to be grossly misaligned. And I've had the worst problems with it when trying to put the whole planet into the alignment box when the individual frames have lacked consistent sharp features. Breaking the AVI down into frames using virtualdub then aligning & cropping them in ninox seems to help solve the issue for me, and my theory seems to be consistent with my observations. Your mileage may differ!

Brian, Hi.

That's a really interesting observation about onion ringing. I find I get the same problem on all planets if the planet image has a poorly-defined edge and I try to push the sharpening. I'd always assumed that it was caused by continuous gradients in disc brightness caused by limb darkening being 'converted' into discrete bands by the sharpening process. This is where manipulating wavelets can come into it's own as, (from my experience) these onion ring effects tend to be large scale structures in the image and hence are made worse by increasing sharpening in the large order wavelets (e.g. 1:5 and 1:6), more than those wavelets working on smaller scale features (e.g. 1:2 and 1:1).

Perhaps one way around this might be to use the dodge tool in Photoshop to lighten the edge of the unsharpened imaged before importing this modifed image back into Registax for applying wavelets. Used carefully, this could have the effect of greatly lessening the drop off in brightness near the disc edge and hence, in theory at least, lessen to propensity to form 'onion rings' when sharpening the image. Some form of radial gradient-based gradient applied to the image brightness might also be used to effect this change. (I have no idea at present how this would be achieved). I'll try to get around to giving these ideas a try and let you know how I get on.

[brianb]

I find I get the same problem on all planets if the planet image has a poorly-defined edge and I try to push the sharpening. I'd always assumed that it was caused by continuous gradients in disc brightness caused by limb darkening being 'converted' into discrete bands by the sharpening process.

I'd tend to agree with your logic, except that I've occasionally had gross onion ringing when trying to stack full disk solar images - the FFT transformed data sometimes has one prominence more significant than the limb, sometimes a different one, and if the two proms are close enough to be "confused" by seeing effects then you can get an image that looks like a bunch of superimposed improperly registered images even before wavelet sharpening. And this is with overexposed solar disks that do have fairly sharp edges, much sharper than e.g. Jupiter's globe.

Of course, once there is a tendency for spurious data to be in the image, sharpening can pull it out into the open. Reducing sigma tolerance in stacking can reduce the tendency at the expense of more noise in the image as less frames will contribute to the image at critical points.

For images with limb darkening, increasing gamma in capture will increase the apparent sharpness of the disc edge if you think that would help - probably a damn sight easier than messing around with radial filtering; of course you would need to reduce gamma in the finished image to get the surface detail out again.

[michaelmorris]

Okay I've tried out my theory on a smooth gradient and it doesn't work! Here is the evidence. At the top is a simple smooth gradient created in photoshop. Next is the same gradient sharpened in Registax using 100% Gaussian sharpening on the 1:1, 2:1 and 3:1 wavelet layers. Below this is the original gradient sharpened in Registax using 100% Gaussian sharpening on the 4:1, 5:1 and 6:1 wavelet layers.

AS you can see, they are all identical.

Suggested reason:

Because it is a perfectly smooth gradient there is no edge for the sharpening algorithm to work on so all three images are the same.

I'll have to go away and try something else!

[Kokatha man]

Strewth fellas.....perhaps I should just forget about chronicling my various processing methodologies and just be content with posting images.....it is rather hard when you start a thread with said aspirations and it developes into a debate between a couple of people posting various images of theirs etc about the cause of "onion rings".....don't get me wrong, they are a phenomenum VERY worth a dozen threads (at least!!!) but I did post the 2 variations upon my latest Jupiter and ask for some sort of on-topic, constructive input.....

And before you counter with the "onion rings are artefacts encountered in processing" I agree - but they are really just veering this thread way aways from my hopes for it.....and whilst I wouldn't entirely disagree with both of you (in parts!) I believe that essentially they are more a capture defficiency manifesting itself in processing.....

AAAAAAAHHHHHH....! I give up - you've sucked me into this onion ring ranting - AND I DON'T EVEN HAVE ANY FLIPPIN' ONION RINGS, OR PEELS, OR ANY GARLIC FOR THAT MATTER!!! I think I'll go have a lay down and ponder the merits of attempting to continue with my original intentions....!:)

[Kokatha man]

If you are working at an image scale of > 512 pixels to the planet diameter there really isn't a choice, default single alignment probably won't work well, "onion skinning" artifacts will almost certainly arise.

As for "onion ringing", my analysis has been that Registax is misidentifying minor projections from the belts causing some frames to be grossly misaligned. And I've had the worst problems with it when trying to put the whole planet into the alignment box when the individual frames have lacked consistent sharp features. Breaking the AVI down into frames using virtualdub then aligning & cropping them in ninox seems to help solve the issue for me, and my theory seems to be consistent with my observations. Your mileage may differ!

.....Aplogies for that - I just needed a bottle of whiskey and now I'm fine....!:)

I've put 2 of your quotes in this thread together Brian 'cos initially when looking back I saw your "as for onion ringing" and thought.....hang on - I never mentioned onions!!!:eek: .....then continued looking and found that you had raised the subject too!!!

Very cunning bro.....you've not only found a way to post your own images in other posts (sorry, I deleted it from the quote here!) but you sneakily raise issues not canvassed and turn them into the "main debate."

You don't happen to work for an advertising agency by any chance....?

But I can see it all with good humour bro (especially now I'm drunk).....just going down to the local now - I'm out of whiskey!!!:)