7.75
#...
·
·
1
like
|
---|
David Payne:Thank you David. In addition to your tips I really like the video by PaulymanAstro on the "Eric Coles Method" to balance the color chanels: The Eric Coles Method using GHS 2.0 - YouTube Murphy's law, whenever you have a new idea or new equipment, immediately the clouds roll in ... Happy to share my experiences and pick up suggestions by others. Clear skies Wolfgang |
4.69
#...
·
|
---|
Mike Cranfield: How is this different to EZ Soft Stretch (which is what I always use) |
10.10
#...
·
·
2
likes
|
---|
Hi Stuart, To be honest, I don't really know EZ Stretch, but they may be targeting different user needs. The claim Mike and I make is that the nature and properties of the GHS transforms, the degrees of freedom provided the parameters to control the stretch, and the environment that encourages the examination of target and preview images, target and preview histograms in log and linear scale, together with the interpretation of the transform makes it ideal for both beginner and advanced users. My suggestion is to do some reading and try them both (and your other options) out - I believe they are all without cost. There is a lot of information and links available at the ghsastro.co.uk site to answer your question. Whatever makes you enjoy the processing and the gratification from the image the most wins. Dave |
23.42
#...
·
·
5
likes
|
---|
Stuart Taylor:Mike Cranfield: Hi Stuart I don't use EZ soft stretch but my understanding is that it should only be used on linear data - it provides a single stretch taking you from linear to non-linear. Essentially it uses the Histogram Transformation approach, ie it doesn't do anything you couldn't do yourself with the standard Pixinsight Histogram Transformation process but it gives you some help in choosing the HT parameters based on your image characteristics and some parameters that you input. The GHS script is offering something fundamentally different from this. The Histogram Transformation process (and hence also EZ soft stretch) uses an intensity transformation known as the midtones transfer function (mtf). The GHS script allows you to use an mtf if you wish but it also offers an entirely new family of intensity transformations called generalised hyperbolic equations, from which the script derives its name. These equations have extremely desirable qualities when stretching images. These qualities have been discussed at length on this forum and elsewhere but one important feature is that they tend to be very "kind" to stars (star bloat is a common issue with the mtf used by Histogram Transformation). In addition to offering this new family of intensity transformations, the GHS script also provides some important additional parameters/degrees of freedom in specifying the stretch. This means, for example, you have significantly greater control over where contrast is to be added than with any other currently available script or process (so far as I am aware). This provides considerable flexibility to tailor your stretch to bring out the detail you want to highlight within your image. Careful application of successive stretches using GHS can really enhance and fine tune your image to your precise requirements. I could go on, but of course the best way to see the difference is to download the script and have a go - it's free so you have nothing to lose! |
5.70
#...
·
·
1
like
|
---|
Where would be a logical place to insert this into my workflow? Patrick |
10.10
#...
·
·
4
likes
|
---|
Patrick Graham: Hi Patrick When you want to stretch the image from linear to non-linear which is necessary for just about any deep sky target, or when you want to make brightness or contrast adjustments to your image. If you aren't taking deep sky images, you may not be converting your images from linear to non-linear; if you are doing EAA - your data capture program my be doing the stretching of linear to non-linear for you. In these cases GHS is only for brightness or contrast adjustments to your near-final image. For deep sky, you should do all of your linear operations first. This may require a non-linear mask - in which you can of course create with GHS. Or it may making an image clone stretched somewhat to "help" get your stars with Starnet. Other than that you should do any deconvolution, linear NR, colour calibration, gradient removal etc. or linear operations (where the process counts on pixel values being additive or linear) FIRST. Then use GHS to stretch your image to non-linear. Then finish off with your non-linear processes where the "additive" linear properties such as superposition are no longer required. Cheers, Dave |
5.70
#...
·
·
1
like
|
---|
Thanks, Dave. Very informative. I appreciate you taking the time to respond. CS Patrick |
7.75
#...
·
·
2
likes
|
---|
Patrick Graham: One of the many great things about GHS is not only that it's a fantastic tool. In addition Dave and Mike really care about anyone using it, even humble beginners like me who take their first steps in AP. A very capable tool with first rate support by its inventors. What else to hope for ... Clear skies and good luck with the tool. Give it a try and I'm sure you're going to be hooked up to it. Wolfgang |
2.97
#...
·
|
---|
Dave, Thanks for the description of your initial stretch technique. The documentation also discusses then recovering DR lost on the left side of the histogram and then following that up with a "final initial" stretch to restore background levels. The first of these can be done with a linear pre-stretch or (preferably, so that it's reversible) GHS SP/HP. However, I haven't found a good protocol for this like what you have written here. Do you have any specific suggestions? Cheers, Roman |
23.42
#...
·
·
2
likes
|
---|
Hi Roman I’m sure Dave can give a better answer to your question but let me share from my own experience. The first stretch does, as you say leave a fair amount of DR at the left side of the histogram. However, my second and subsequent stretches tend to focus on adding contrast to areas of the image that I want to bring out. As we know, contrast can only be added if it is taken from somewhere else and this tends to be at least in part from that left hand “wasted” DR. The end effect after a few stretches often ends up with a result I am happy with without having had to do anything specific to reclaim that left hand DR after the first stretch. It would be great to hear others’ experiences and suggestions on this too. Cheers Mike |
10.10
#...
·
·
3
likes
|
---|
Roman Pearah: Hi Roman, I believe I know what you are asking but forgive me if I misread your question. Let's say with all the contrast you have been adding, you happened to have compressed the background into darkness too much and you would like to see it again. This issue you might have is that when you put SP near/at this data (at a low value, say 0.0 to 0.15) and a strong D & b to brighten and add contrast to it, you end up brightening everything to the right too much and destroying the work. In this case, try moving the HP slider lower, towards SP. You will note, that as you pass 0.5 or so, that the stretch taking place to the right of HP is actually linear - ie a straight line. The transform line above HP will initially be above the identity line, meaning that these pixels will still be brightnened, but less and less as HP is move too the left. The slope is less than the the identity line too, because you are taking contrast from the RHS of HP to feed the curve on the LHS. So to the right of HP, the image will be brightened, but in a linear way (without clipping) but by an amount that allows for the brightening of your dark data. If the effect is not enough, or too much, you should adjust D,b (or even SP), but your "work" to the right of HP will be "preserved" When adjusting HP is this manner, you may want to watch the identity line and the transform. Lowering HP will push down the amount of stretch occurring to the left of HP and you may want to be careful that it doesn't drop below the identity (45 degree line), or you will actually be further darkening the areas where it is below. You may find that as you lower HP, you can actually place the straight line to the right of HP directly upon the identity line. In this case you will get the opposite of what you are trying to achieve - further darkening/compression of the of pixels to the left of HP, but all pixels to the right of HP will remain exactly as they were - no stretch - because the line is indeed on the identity. Further movement of HP towards SP may move the line below the identity, which is a good way to compress dark pixels anyways. So play with this technique to see if it work for you. Also not that this works in reverse by setting a high SP and moving LP towards it. FInally, you can play "surgeon" to some extent, by putting SP in the middle some where and pushing LP towards it from the left and HP from the right. You will be perming esssentially a "bi-linear" stretch, with the GHS equation forming a well behaved transition in the middle. I hope this helps, Dave |
9.03
#...
·
·
1
like
|
---|
Hi Guys, i have worked away on this and have a couple images up there. Sh2 119 HOO Abell 39 M 5 Using Generalized Hyperbolic Stretch Sh2 115 using Generalized Hyperbolic Stretch I am afraid you have created a monster.... I am hooked. I have tried to put some GHS notes up there to make this simpler.... I find this is extremely useful when combined with destarring and then denoising the images with StarXterminator and NoiseXterminator... no reason to stretch the noise or the stars if you can avoid it.... anyway I am squeezing a lot of signal out of my new ZWO ASI 6200M camera. Constructive comment on the images appreciated. I am working on Abell 43 now and have reviewed the video (Have done all the Abell PNs and Sh2 objects once) now going back for revenge.... I am getting a lot of detail in Abell 43. Work in progress contrasting stretch 1 versus 2: https://www.astrobin.com/muvint/ |
2.11
#...
·
·
1
like
|
---|
Hi Mike - thanks for your amazing script, it gives excellent and progressive control of the stretching intensity and behavior on different parts of the image, that with other tools is something like a crazy horse. Not to be forgotten your support site that includes great tutorials, necessary to use the script properly. GHS is, since I discovered it, my #1 stretching tool. |
5.25
#...
·
·
1
like
|
---|
And thx to @Adrian Knagg-Baugh , @David Payne and @Mike Cranfield Siril users will have the benefit of this tool too . Thanks guys. |
10.10
#...
·
·
3
likes
|
---|
Hi Guys, Hi @jerryyyyy , Spectacular Images... It sounds like you are on the similar tack as I am taking. I use the same camera and often do deconvolution starless, as well as some non-linear processes. Also, I have NoiseXterminator and I like it a lot, but I agree with others who have indicated that the default 0.9 denoise slider is too high. The formulation of the transforms in GHS were designed to help bring weak signal out of noise by enabling the placement of contrast, and the amount of contrast placed. What I would encourage, and perhaps you do this already, is to prior to denoising, you should explore the background using the GHS script by looking for any dim signal in the "background", that you might otherwise overlook in your data. It is a step I am using now in my processing in that I just do a series of "exploratory" stretches just to find out what is in my data, particularly the dim stuff that may have a weak SNR ratio and then decide if that is something I want to show (in most cases I would want this). The challenge here is to place contrast in between this weak signal (that likely contains a lot of noise too) and the true background, which for sake of argument, we can say only contains noise because we found this out during our exploratory stretch. The second step after the exploratory stretch is to then employ NoiseXterminator or whatever denoise routine to see if you can sufficiently reduce the noise without having the denoiser interpret the signal as noise and take out the signal too. Similarly with background extraction, you want to make sure you don't take out the signal along with any gradients. This is an art that I am still trying to master - you may need to denoise, then test with GHS to see if you denoised too much or too little. GHS allows you to try and place contrast between weak signal and pure noise, but it doesn't know the difference itself and will stretch noise right along with the signal you are after. Ultimately, if you cant get to a place that is acceptable to you, it may indicate that you need additional total integration time to boost the SNR of any dim signal. So the results can be: A) You can satisfactority bring the dim signal out from the background without too much noise using your denoise tool and GHS B) You need to add frames to improve the SNR of the dim stuff you want to show C) You decide that you want to hide this signal, because it is too noisy, in the background by making it dark . (At least the no-clipping feature of GHS means that you may be able to bring it out in the future when you add more integration time.) My point is, unless you explore your data sufficiently before hand you may not even know it is there, then either denoising it away, or by setting your initial SP setting too high, too far to the right, just pushing it in the dark background . At least C above is a conscious decision and better than this. To me, the images of @Eric Coles (coles44) are amongst the best I have seen that have pulled this weak signal (nebulosity, galactic "atmospheres", or IFN) out from the data and displaying it, although there are many more better than myself at this. (My excuse is partly that I have had such little practice of late, due to this years' La Nina effecting Vancouver Island skies). Thanks for sharing your magnificent images, you are obviously a master of this yourself and I only wanted to give my take on this. Chears and Clear Skies! Dave |
10.10
#...
·
·
1
like
|
---|
Cyril Richard: Thanks @Cyril Richard and thanks for letting me "beta-test" it. The implementation by @Adrian Knagg-Baugh is brilliant, with the selfless help of @Mike Cranfield and of course yourself, this is a great addition to Siril, which itself is a fantastic program. Siril itself is already quite popular amongst many of my AP friends and fellow club members and I am now joining their ranks - giving me a choice of another highly competent and increasingly complete processing option. |
9.03
#...
·
·
2
likes
|
---|
David Payne:Hi Guys, Thnaks for the comments. I will explore the data more. On the Sh2 objects I was able to look at the H-alpha carefully and set SP accordingly. On both 115 and 119 I had to get more O-III as that is just noisier. The O-III in 119 just "popped out" with GHS... you could have looked at it for days as there is so much other H-alpha in the background... It is not a reflection as it has a structure. I like the tip on NoiseXterminator. I have set it down to 80%. Not sure if you are familiar with the Topaz products for PS, but they have a very powerful AI nenoiser, but I have left that expensive software for NoiseX in linear. I know Eric as we are "local" but he is up at SRO in the dark sky territory so he has a lot more to pull at in his LRGB images... I am stuck with H-alpha... I also can only fit a 5" refractor in my system... complaints complaints Here is the "final" version of Abell 43: General Hyperbolic Stretch of Abell 43 I put it up with both stretches overlapping. Key is where to put the SP. I know my new camera (ZWO ASI 6200M) is very very noise free and has a deep photon bucket... but if you do not process these linear images right, you will miss the detail that is there... it is very very finely separated in the linear images. PS I cross posted a line or two over at the Pixinsight Forum and people seem to be struggling with the basics.... I think you need a GHS Stretch Made Simple page... PPS One more small Abell, but this time I did not destar, just dealth with the stars in HOO and in GHS. Abell 66 using Generalized Hyperbolic Stretch |
9.03
#...
·
·
1
like
|
---|
Trying to figure out how to use the Shadow Protection Point in this GC (M12). This is the second stretch after going from linear: Trying to get the background darker but at the same time make some of the dimmer star visible. Not sure I understand how to do this... normally in HT I would set the histoprgram black point just to the left of the data. |
10.10
#...
·
·
1
like
|
---|
Hi, In short - do the stretch you want to do, inserting contrast where & how much you want using D, b & SP, or smoothing out humps in the log histogram, or whatever you are trying to do. Then, control the brightness of the image using LP (to brighted the image - higher LP will shift the histogram to the right and brighten- at first the dims and then the overall image. The opposite for HP, as it is moved lower, at first it will just protect the stars, but as it is moved lower towards SP, it is dim the more of the whole image. You may then need to iterate a little bit between adjusting the stretch using D, b, & SP, and then adjusting the brightness profile using LP and SP. After you do this a few times, the relationship between the parameters will become more understandable and you will become proficient in setting the parameters - i.e. you will "get the knack of it". Hope this helps, Dave |
9.03
#...
·
|
---|
Thanks, I used the LP but it seemed to brighten the whole image too much. I checked the historgram in PI and ran the "set to sero" option on the back point and it found very few data to stretch on the left... see below. So when I just set to zero I get a nice imagem one on the right\, and I know no data are thrown away, maybe? |
7.75
#...
·
·
1
like
|
---|
Thanks, I used the LP but it seemed to brighten the whole image too much. I checked the historgram in PI and ran the "set to sero" option on the back point and it found very few data to stretch on the left... see below. So when I just set to zero I get a nice imagem one on the right\, and I know no data are thrown away, maybe? Apologies for repeating myself but I still get the best results with GHS when performing an iteration of GHS and linear stretches. Start with a GHS stretch with high b (around 10) and D value and an appropriate SP. Then adjust the black point with a linear stretch without clipping data. I leave a small offset for the black point since I don't want to move weak signal too close to zero. Then perform the second GHS stretch. Adjust the black point again in another linear stretch. Usually I'm done after 3 iterations. In rare cases a 4th iteration is in order. I then tweak the colours by adjusting the individual colour channels. In a final step I address colour saturation. I really appreciate that all of this can be done seamlessly within the script. Stretching images has never been easier ... Clear skies Wolfgang |
10.10
#...
·
·
3
likes
|
---|
Hi Jerry and Wolfgang, There is nothing wrong with linear stretches and adjusting black point or even clipping data - just that you don't need to.... The key is to watch the transform curve and where it is relative to the identity line. where the transform is below the identity line, those pixels will be dimmed, above will be brightnened and on it will remain the same... For example, to only adjust the background brightness while leaving subject matter unchanged... (note that HP was adjusted so that most of the red transform lies on top of the identity line.) To dim the whole image - like a blackpoint adjust without clipping: not that HP has been moved all the way down to the SP point. To dim the brightest parts the most.... Here we have used a logarithmic (low focus contrast addition) b of -1 and placed SP right at one as this is where the maximum contrast will be added (more dimming between 1 and any data). In this case, the dimming is further tempered by applying some LP. Of course, the same abilities apply in reverse for brightening. I hope this helps. Cheers, Dave |
7.75
#...
·
·
2
likes
|
---|
Very nice explanation, thank you Dave! So far I only used the highlight protection point to protect the highlights. Using it to push back the dark areas makes a lot of sense. Clear skies Wolfgang |
9.03
#...
·
·
1
like
|
---|
Yes, very interesting manipulation. I have been doing a number of GC and have one more in the hopper... not much else up early for me. I have a lot of data on M12 with about 2,700 stars in the frame. I will check but it would seem that dropping the transform below the original line would lose dim stars? Clipping from the leftin HT using the zero cliup option would not if the histogram is truely empty as PI indicates. |
1.51
#...
·
·
2
likes
|
---|
Thanks Mike, I've found the processes of learning about the tool a real eye-opener in terms of understanding histograms and the point of the various transformation possibilities. Many thanks for this, you have contributed to the discipline/art of astrophotography. Nice job, Paul |