AS!2 Drizzle Comparison.

[Space Cowboy]

Decided to run my best Jupiter avi through AS!2. Here is a comparison (drizzle first) of using 1.5x drizzle against none drizzle. Personally I prefer the drizzle version as it has a more refined quality though less of a detail "pop".

1.5x Drizzle:

No Drizzle :

[neil phillips]

were the wavelets all the sharpening processing ect exactly the same, also prefer the drizzle

[Space Cowboy]

Yes same wavelets Neil though did use more de-noise on the drizzle version.

[neil phillips]

When doing comparisons like this its important to keep the procesing the same, more denoise wont likely make the differences im seeing. The colour balance looks slightly different between the two ? that could certainly change its appearance

[WayBig]

The drizzle version is definitely sharper. Very nice.

Do you find as2 (is that Astro snap2) gives better results than registaxx, for general processing and for drizzle?

[si@nite]

I agree the drizzle version is nicer, lovely shots. SIMON.

[Space Cowboy]

Sorry guys I've realised the drizzle image had indeed been given more wavelets and extra saturation.

Below are the 2 images with identical wavelets and nothing else. The drizzle image clearly responds better to extra wavelets which would over-cook the non drizzle image. This is what I've found with Mars.

Chris AS!2 is Autostakkert. I find it far superior to Reg 6 especially when seeing is average or below. Reg 5 compares very similar to AS!2 but is so much slower. I've rarely used drizzle with Registax because of the speed issue.

1.5x Drizzle :

No Drizzle :

[Alan]

Yep definately better fine detail in the drizzle image Stuart. You realise now that I will have to go back and re-do all my images

Alan

[neil phillips]

Sorry guys I've realised the drizzle image had indeed been given more wavelets and extra saturation.

Below are the 2 images with identical wavelets and nothing else. The drizzle image clearly responds better to extra wavelets which would over-cook the non drizzle image. This is what I've found with Mars.

Chris AS!2 is Autostakkert. I find it far superior to Reg 6 especially when seeing is average or below. Reg 5 compares very similar to AS!2 but is so much slower. I've rarely used drizzle with Registax because of the speed issue.

1.5x Drizzle :

No Drizzle :

Not exactly sure what to make of the differences, any resampled image will behave exactly as you have mentioned,

would over-cook the non drizzle image. This is what I've found with Mars.

when i do a x2 resample avi on k3 and run it through reg. the wavelets behave as youve mentioned requiring more use of them to get the same amount of sharpening, as per the unsampled avi. Its a effect the sampling has on the wavelets, not just a AS2 drizzle.So again apart from the fact i do indeed prefer the drizzle version.Im not sure what to make of the comparison ? one could likely get a similar result by using less wavelets on the unsampled image, or more on the upsampled image if you get my drift

[beamish]

I'm gonna buck the trend here and say I prefer the no-drizzle version, I'd work a bit more with the wavelets I think, but the drizle version loses detail in the cloud swirls and colour esp. in the central Eq belt

my 2p

[neil phillips]

I'm gonna buck the trend here and say I prefer the no-drizzle version, I'd work a bit more with the wavelets I think, but the drizle version loses detail in the cloud swirls and colour esp. in the central Eq belt

my 2p

This is the question detail versus natural isnt it, your right it loses detail, but looks far less processed. I would likely go with the drizzle wavelet, and try to find a way to bring out the detail a bit more.

Extra sharpening, what ever that is, for hopefully somewhere between the two points The original version looked better because of the extra sharpening and saturation.

It looks weaker on the same as processing, second time around doesnt it. its now too laid back agreed karlo

[Space Cowboy]

I'm not suggesting there is anything special about the AS!2 drizzle over other re-sample methods Neil.

The wavelets do not behave in the same way when used on the drizzle image i.e using less wavelets on the non-drizzle image does not produce a similar result.

[neil phillips]

I'm not suggesting there is anything special about the AS!2 drizzle over other re-sample methods Neil.

The wavelets do not behave in the same way when used on the drizzle image i.e using less wavelets on the non-drizzle image does not produce a similar result.

Obviously i cant be sure, but i know that im correct when i say upsampled images need more wavelets to get a simlair sharpened effect, ive seen myself a million times on my own experiments. I tell you what send me 2 images with no wavelets drizzle no drizzle, and ill see if using less wavelets ( or even different ones, because upsampling will also efffect that. ) will not produce similar results with just wavlets alone. im not convinced your right about that. If i find that is the case, then i agree theres a lot more too it, than seeing the effects of the wavlets being affected by upsampled or not, images. Care to test to it out, im game

The reason it can overcook it on unsampled image is because with such images the wavlet effect is more powerful, as mentioned ive done x2 k3 avis, first time i tried i was surprised to find i had to use not just a little more wavelets to get it sharpened but probably about TWICE as much. yeah that subtle

I realize its a different type of upsampling, and it was x2 not 1.5. so effect was likely even more pronounced. But your comments tally exactly with what i found with my own experiments, and even the images apperance too. the only thing that doesnt tally is your assertion that the AS/2 upsampled image isnt affected in the same way the k3 upsampled avi is. I agree thats possible, and i could be wrong. I would like to test that out if pos. At the moment i think its mostly the effect i mention. But am willing to discover thats not the case. In which case the AS2 drizzle does indeed behave completely different to the k3 x2 effect on wavlets. Alternatively i could run these experiments for myself and post the results, but will take more time. if you have a couple of images handy. as im working 5 saturn rgbs from the 28th that i may make a movie from.

[Tim]

That's a cracking image!

I'd be very interested to see what would come out of this sequence if you used the Atrouswavelet tool in Pixinsight and followed the tutorial for it.

PS, prefer the drizzle version too

[Space Cowboy]

No problem Neil I've attached both TIFF files.

DFK0043 11-11-27 23-09-14_castr_g2_b3_ap52.tif

Drizzle1.5_DFK0043 11-11-27 23-09-14_castr_g2_b3_ap52.tif

[neil phillips]

No problem Neil I've attached both TIFF files.

Ok i had a look at these images, and if you read exactly what i was suggesting, that upsampled images can affect wavelet strength ( and possibly even there fineness ) when you suggested You noticed the Mars images did not overcook using the drizzle 150 resizes.

I remembered the effect my k3 x2 upsampled images had on wavelets, seriously effecting there strength, by a large margin.

Upon putting both your drizzle and non upsampled images on registax, i tried a small amount of wavelet sharpening on both.

straight away i could see that the drizzle image was reacting to the strength of the top wavelet far less, than the unsampled image.

Basically proving some upsampled images ( as yet i havent determined which types, but AS2 drizzle and k3ccd x2 avi ) Are affected in similair ways as i was suggesting.

Heres a example bothe exactly the same 37.4 top wavelet ONLY.

Notice how the drizzle image looks like it has less wavelet applied. it doesnt, its the same. The unsampled image looks sharper and more processed.

Now i also wondered if i could get the images to look very similar by manipulating the wavlets to try to take into account the image strength ( and possibly fineness ) altering effects of some upsampled images.

Now of course i am not going to be able to make them look EXACTLY the same, after all, one is upsampled using drizzle, theres different amounts of wavelets being used ( to try to rebalance the upsampled effect on wavelet strength )

I used 37.4 top wavelet on the unsampled image

And 37.4 top wavlet plus 38.6 second wavelet on the upsampled image to try to rebalance this effect on wavelets, some upsampled images appear to be having.

I noticed just increasing the top wavelet above 37.4 on the drizzle 150 image also worked quite well.

but i belive ( might be wrong im unsure ) but belive the fineness of the wavelet is also affected by some upsampling, meaning by using a bit of the second wavelet, as oppossed to just increasing the top one, seemed to work better in counteracting the effect of the wavelet strength and fineness the upsampled image was having.

Yes theres differences, but ive mentioned why of course there will be.

I certainly think ive proved to my own satisfaction that most of the differences ( possibly not all granted ? ) but most of the differences are purely a affect the upsampled image is haveing on your wavlet choice.

Its why i gave up doing x2 k3 avis on registax, as i didnt like the effect it was having on the wavelets. and found i could get a similair result ( as here ) just by using more or less, or different wavelets.

You may not agree. But i am right about the effect on wavlets.

I think its pretty clear in these examples using k3 and now your drizzle images, like i suggested.

I made no attempt to process your image. just to show what i was saying was correct, i wasnt certain to be honest. Untill i tried. But it looked that way to me, as your comparison looked very similar to my k3 x2 experiments. And your comments about overcooking.The differences now are far less clear, and likely due to a number of factors not least of which was my poor attempt to balance them. Looking at them here it looks like i could have used slightly less sharpening on the drizzle image to balance them even more. But it was becoming a pain lol. I could certainly do better likely if i spent time to the point where they look almost identical ? ps just tried another slight tweak and its damm hard to get exact. Im not suggesting its not worth using drizzle, just one has to be careful of the assumptions they make when differences appear

another tweak

[neil phillips]

Couldnt do all that without trying to do what i would do on my images

You likely wont like it, as i sometimes like to add zing to images, not neutral no, but The resize was more laid back. but i thought the life looked good.

[kev]

Interesting thread and comparisons.

I note your comment about Reg 5.....

Do you prefer that to Reg 6 on planetary?

[RobH]

I've just been delving into planetary imaging for the first time with pretty shoddy results!!....No idea how you guys produce such wonderful results

The 1.5x drizzle is definately the better image.

I downloaded AS!2 and it looks great...but....how do you save an image....there's no option. Or does it save automatically? If so, where?

Sorry for being thick!!

Cheers

Rob

[beamish]

It saves automatically in a separate folder within the source folder Rob.

re my "bucking".. i will still stick my head out and say the original is the better image but as I said in the post I'd tweak the wavelets- to me it does appear a bit strong and as Neil has suggested it needs to be somewhere between the two. I feel that the softer larger image benefits from simply being larger and people like large, we all do, but I do feel it has gone too soft and detail has been lost so the Q one has to ask oneself is -what is the reason behind the upscale ? since clearly trying to kill the grain induced from noise has had a detrimental effect on the information.

Yes I'd agree that put one against the other in a straight comparison the drizzle appears aesthetically more appealing but... when that means you give up data .....???

Unless the data can more or less remain intact when upscaled then I can't see the reason for it. [EDIT: I'll admit this is a more recent change in attitude , partly because natural scale is easy to get]

Upscaling and drizzle however you do it will, as Neil suggests, results in more sharpening required to pull the same detail but with added problem of noise so,... it will always be a balancing act.

Bottom line is IMHO is that the smaller image could be improved -and show more detail.

Great thread tho guys

[Kokatha man]

.....easy to get lost in these sorts of threads with so many images and variations thereof being posted: in short as far as I'm concerned drizzle in AS!2 works very well most of the time, but it is of course dependant upon the quality of the images fed in, being an interpolation procedure.....the better (more detailed) the data the better the outcome.

I never really bother with processing poor data unless times are rough and I've got nothing else:eek: - although doing another repro of some decent data is more satisfying and preferable then :) but if the data is reasonable I invariably do a drizzle, 3x if it'll handle said but 1.5x if it's borderline (ie, just worth processing!!!)

Rob, AS!2 will automatically save the end-image as either a .png or .tif which you nominate in the program.....you can create a folder for your avi and the output from AS!2 that will hold everything but AS!2 will create a folder anyway (eg, AS_f4000 etc) and the .png or tif will be inside this (the "4000" indicates you chose to stack 4000 frames and if you vary this the specific number will appear in the folder AS!2 creates - good as you can trial several different-sized stacks and not have to worry about them getting mixed up!)

[RobH]

Thanks....but where does it create the folder?

Rob

[beamish]

within the original source folder........................normally . From whence you got the .avi

[Anweniel]

Lot of comparisons and techniques debated there! I'm very unexperienced with the processing of my images but have always avoided the drizzle options as given the time it akes for the stacking process to work out a final image, to have it get to the end only to tell me that I am 'out of memory' is quite annoying! No such problems with resizing.

Problem with resized image is the loss of definition, not sure if its my unexperienced hand or a lack of finess with the wavelets but I cannot achieve anything like the detail of the original final stack after a 200% resize ( but then mine are a lot smaller images than these beasts!) Maybe theres something I do wrong though with the drizzle function - not understanding entirely what its doing, lol.

Must say though thats a cracking image of Jupiter being played with and I think theres some middle ground to achieve between the natural and heavily processed images. Its going to be great to get as much fine detail out of an image as possible, I do however prefer a more natural appearance. So, somewhere close to those Drizzled images with only top wavelet adjusted would appeal to me greatly.

Great image and interesting thread.

[RobH]

within the original source folder........................normally . From whence you got the .avi

Thanks Karlo