Help needed

[alacant]

6 hours ago, bottletopburly said:

levels and curves taking forever and a day to get something then I would highly recommend Startools

+1

Change is hard. Many will stay with fossil fuel powered motor vehicles for ever.

If there are alternatives, let's have the freedom to recommend them.

It's not the 1990s any longer. Let's move forward.

Cheers 

 

Edited May 28, 2021 by alacant

[vlaiv]

10 minutes ago, alacant said:

Hi

IMO, it's better by design. You tell it what you want, rather than vica versa. It has an excellent choice of alternative stacking algorithms and comprehensive debayer routines along with some very good background and noise reduction modules.

Recommended:)

Do you know what interpolation algorithm is used for registration (does it have a choice)?

[alacant]

5 minutes ago, vlaiv said:

Do you know what interpolation algorithm is used

Yes. Which is your favourite?

 

 

[vlaiv]

1 minute ago, alacant said:

Yes. Which is your favourite?

 

 

I don't really have favorite, but I dislike using DSS because it uses simple bilinear interpolation which alters noise statistics and lowers resolution in the image - acts as strong low pass filter.

As long as it is decent filter - like Lanczos, Splines of higher order - is better than simple bilinear / bicubic filter

You can see that image has been stacked with DSS if you examine noise grain. Here is simple comparison to show you what I mean:

From left to right - just some Gaussian noise in first image, second image is first image translated by 0.5px, 0.5px using bilinear interpolation, third image is again - first image translated by 0.5px, 0.5px - using Quintic B-Spline as interpolation.

You can clearly see blur that results from use of bilinear interpolation. It not only impacts noise in the image - but also detail. FWHM of stars increases after registration with bilinear interpolation.

 

[alacant]

7 minutes ago, vlaiv said:

don't really have favorite,

I'm surprised.

Could any of these become your favourite perhaps?

[vlaiv]

3 minutes ago, alacant said:

Could any of these become your favourite perhaps?

There you go - Lanczos-4 will do

Must download latest version of Siril - I have old one that I tried some quite a bit long time ago.

Edited May 28, 2021 by vlaiv

[alacant]

26 minutes ago, vlaiv said:

Must download latest version

OMG vlaiv. We almost agree on something LOL!

[Lee_P]

Thought I'd join in the fun. About 15 minutes of editing time, mostly using PixInsight, then a little bit of Lightroom. I pushed it too far so it's noisy, but I reckon you've got some good data in there. I could do more but need to collect my daughter from nursery 😁

 

[alacant]

5 hours ago, Lee_P said:

I pushed it too far so it's noisy

Nice processing and thanks for being honest about it...

It illustrates the problem with old style apps. Stretch and you're now no longer working with linear data. Get that wrong and...

How many times have we all blown it by over (or under) doing it? Go back, start again and hope you make a better guess next time...

In StarTools we perform the important processing with linear data and so always have the opportunity of going back to change the past if we don't like the present. Stuff like stretching and deconvolution which need linear data can now be performed at any stage of the process. Data modelling at its best.

Cheers

Edited May 28, 2021 by alacant

[The Lazy Astronomer]

9 hours ago, alacant said:

Hi

IMO, it's better by design. You tell it what you want, rather than vica versa. It has an excellent choice of alternative stacking algorithms and comprehensive debayer routines along with some very good background and noise reduction modules.

Recommended:)

Interesting. I'll definitely be trying it out now. I tried ASTAP briefly, but went back to DSS because it was just so easy to use.

@vlaiv: out of curiosity, what is your stacker of choice, and why? 

[vlaiv]

48 minutes ago, The Lazy Astronomer said:

out of curiosity, what is your stacker of choice, and why? 

ImageJ.

I tried many, I even downloaded evaluation copy of PI, but I never warmed up to it.

ImageJ allows me freedom not many other software does. It is built for completely different purpose - scientific image processing for microscopy - that was original goal.

It has most of the functionality you'll need to calibrate, align and stack your data out of the box - but it also lets you program additional features in form of macro language or plugins.

It will probably be completely useless for most novice users, but if you have background in programming / image and signal processing - then it is valuable tool box.

I've written several plugins of my own that I use when stacking and processing data.

As far as I can see - Siril will do important part of the job for me, and I plan to include it in my workflow.  One of the problems with ImageJ that I have now is that I can't do automatic alignment in the way I would like to do. There is really nice plugin that uses RANSAC algorithm to do frame alignment - but it either produces transform matrix or performs actual alignment using bilinear interpolation.

If you want better interpolation algorithm - you need to manually align subs using that transform matrix and different plugin.

I also written plugin that uses Lanczos interpolation - but that was just to test out things - I did not code it to be fully functional alignment plugin (like load matrix data and align all subs). Due to work and other commitments, I don't have much time now to implement that - but hopefully I'll make software for stacking at some time in future - I have several very interesting and innovative algorithms that I tried out using ImageJ.

I've written couple of times about these ideas and presented results. For example here:

That one deals with different quality of subs in the way no other stacking software does. Most just assign "weight" to whole sub - but there is no single weight that is appropriate for whole sub (there is no single SNR for whole image - every pixel has its own SNR) - each portion of sub requires different weight depending on SNR of that particular part of the image - my algorithm deals with that and breaks image up in zones depending on SNR and assigns weight according to SNR of each sub compared to others.