PixInsight Image Alignment - what have I done wrong

[groberts]

Assisted by the Light Vortex tutorial + after some wobbles, I successfully managed to integrate (stack) a set of LRGBHa subs for the first time, which looked OK when stretched.  I've now moved onto the next step of alignment of the said images for subsequent combination in StarAlignment which on processing gives the the message:  

*** Error: Variant::ToDouble(): Invalid conversion from ByteVector type

<* failed *>

Any thoughts on what's wrong?

I suspect the problem might be file type.  On completing the aforemention stacking out of habit from DSS I thought I'd saved the finished stacks as Tiff files but I now see they were FITS - obvioulsy not paying attention.  Strangely I can't see the output file typein settings but suspect that if it's still PI .xisf this could be a conflict? Below are the settings used and a full screnn grab of the error.

Graham 

 

StarAlignment: Global context

Loading reference image:

C:/Users/Graham/Desktop/PixInsight Image Tests/M101 March 2019/Stack2d/L Stack2d.fit

Reading FITS image: 32-bit floating point, 1 channel(s), 9312x7040 pixels: done

BLOB property extracted: 'Instrument:Camera:Gain', 8 bytes.

BLOB property extracted: 'Instrument:Camera:Name', 28 bytes.

BLOB property extracted: 'Instrument:Camera:XBinning', 4 bytes.

BLOB property extracted: 'Instrument:Camera:YBinning', 4 bytes.

BLOB property extracted: 'Instrument:Filter:Name', 2 bytes.

BLOB property extracted: 'Instrument:Sensor:XPixelSize', 8 bytes.

BLOB property extracted: 'Instrument:Sensor:YPixelSize', 8 bytes.

BLOB property extracted: 'Instrument:Telescope:FocalLength', 8 bytes.

BLOB property extracted: 'Instrument:Telescope:Name', 24 bytes.

BLOB property extracted: 'Observation:Center:Dec', 8 bytes.

BLOB property extracted: 'Observation:Center:RA', 8 bytes.

BLOB property extracted: 'Observation:Equinox', 8 bytes.

BLOB property extracted: 'Observation:Location:Latitude', 8 bytes.

BLOB property extracted: 'Observation:Location:Longitude', 8 bytes.

BLOB property extracted: 'Observation:Object:Name', 8 bytes.

BLOB property extracted: 'Observation:Time:End', 12 bytes.

BLOB property extracted: 'Observation:Time:Start', 12 bytes.

63 FITS keywords extracted.

*** Error: Variant::ToDouble(): Invalid conversion from ByteVector type

<* failed *>

Edited June 4, 2020 by groberts

[Wyvernp]

Did you use a single sub to align all of the stacks? if so you should not need to run alignment again and can just combine the channels.

If not at a guess i'd untick "generate drizzle data"

[david_taurus83]

As above, you shouldn't need to align a second time. How did you do the initial alignment? A small criticism of your workflow, especially as you are learning the PI ropes, is that you appear to have pre processed all your individual channels at the same time before moving on to the next step in the process. Personally I would have done one channel from initial calibration through to integration. I tended to learn the process quicker this way as I was repeating the entire cycle 4 times (LRGB) whereas you have processed all your data but have only complete 1 full cycle. That is, of course, if I have interpreted your workflow correctly!

[groberts]

Thanks I appreciate the comments. 

I did align all the waveglengths using a single 'best' luminance sub fo LRGB & Ha and was therefore surprised to have to do this a second time but that's what the Light Vortex Tutorial seems to say or have I misunderstood?  https://www.lightvortexastronomy.com/tutorial-preparing-monochrome-images-for-colour-combination-and-further-post-processing.html#Section1

I have tried again + unticking the 'generate drizzle data' but with the same result.

David I take your point but I'm doing every filter to arrive at a final image (a) repeating each stage for each filter reinforces the learning, though it does take longer, and (b) I would like to end up with a final colour image to compare with my original DSS + Photoshop image.

Just to clarify the other point - is there a preferred file format I should have save the stacked images in or would either FITS, TIFF or xisf be OK at this stage?

Graham   

Edited June 4, 2020 by groberts

[Wyvernp]

I’ve just had a brief read of the previous section on image integration and I believe that in the tutorial it does not say to align all yours stacks to a single frame like we suggest here and thus requiring an additional alignment after the stacks are created. You should be able to skip this step as your stars will already be aligned. 
 

As for file format, I always keep the xisf file for reprocessing in PI as it holds additional metadata. And will only export to tif/fit for further processing in photoshop. 

Edited June 5, 2020 by Wyvernp

[groberts]

4 hours ago, Wyvernp said:

I’ve just had a brief read of the previous section on image integration and I believe that in the tutorial it does not say to align all yours stacks to a single frame like we suggest here and thus requiring an additional alignment after the stacks are created. You should be able to skip this step as your stars will already be aligned. 
 

As for file format, I always keep the xisf file for reprocessing in PI as it holds additional metadata. And will only export to tif/fit for further processing in photoshop. 

Wyvernp thanks for taking the time looking at this further. 

Such are the number of items required to deal with in PI that despite my very best efforts to stick to the excellent script (no pun intened) in Light Vortex, I might well have missed / misread one of the steps - as a long time user of DSS I also suspect I'm too condtioned by aspects of the protocol used therein.  Your observation would certainly explain my confusion.   

Based on this I will endeavour to continue with what I have. Can I therfore use the FITS files I now have or do I need to either restack and save as .xisf files or open and ressave the FITS stacks as .xisf files?

Given the above, I presume I therefore skip item 1.0 in this section and continue with 2.0 through to 4.0? 

Contents

1. Registering the Images with StarAlignment

2. Cropping the Black Edges with DynamicCrop

3. Subtracting Background Gradients with DynamicBackgroundExtraction

4. Matching the Images with LinearFit

  Graham

  

[Wyvernp]

Hi Graham, no problem at all. I'm still learning it myself so have been through similar pains. I can really recommend Chris Woodhouse's videos on pixinsight They've really helped me. 

https://www.youtube.com/channel/UCEd4A5H8HcV8ybp96WRGmsA

So to address your questions 

You could go through the restack process and save as xisf although you should still be able to staralign on fits files as that is the starting point with the raw subs from your camera, If you want to share your stacked files I'd happily have a play with them to see if I can get them working and I'll share the setting I use if I succeed.

I'd go ahead and skip step 1 and continue with the tutorial.

Cheers,

Wyvern

 

It actually looks like we have almost identical imaging setups so I look forward to see what you've captured.

Edited June 5, 2020 by Wyvernp

[groberts]

Hi Wyvern, haven't seen those videos so will take a look - thanks. 

It's taken a while to work up the courage / enthusiasm to tackle PI and despite it's supporters, I've got to say that this is probably the most user unfriendly software I've ever come across but will for the moment persevere! Ah yes hadn't noticed your gear is similar - I like to stick to the KISS principle and can't think how I ended up doing astrophtography.  You'll find my story and images on my website given in the signature at the bottom.  Processing to date has been almost entirely done with DSS and Photoshop.

I've been back and looked at the Light Vortex workflow which shortly after the start + after downloading lights in the section Registering & Intergrating Lights says:

 We need to set one of these light frames as the reference image so that all the others are aligned in respect to it. From the small list box that says View at the top of StarAlignment, select File instead, then click the downward arrow button to the right of it. Select one of your calibrated, cosmetically corrected and weighting-optimised (and Debayered, if applicable) light frames. It is generally recommended you choose the image that has the highest optimised weighting (SSWEIGHT) as this will act as the best reference, especially if FWHM and Eccentricity were included in the optimised weighting calculation. As aforementioned in the previous section, do make sure this reference is free from satellite tracks and star blooming. 

So not sure why you're saying (a) this was not correct and (b) having done this we're asked to subseqently align the stacks? 

Thanks for the offer, as suggested + FYI I've attached just my L stack as an example - not sure if there's something wrong but the file sizes are enormous!

Graham    

L Stack2d.fit

[Wyvernp]

1 hour ago, groberts said:

We need to set one of these light frames as the reference image so that all the others are aligned in respect to it. From the small list box that says View at the top of StarAlignment, select File instead, then click the downward arrow button to the right of it. Select one of your calibrated, cosmetically corrected and weighting-optimised (and Debayered, if applicable) light frames. It is generally recommended you choose the image that has the highest optimised weighting (SSWEIGHT) as this will act as the best reference, especially if FWHM and Eccentricity were included in the optimised weighting calculation. As aforementioned in the previous section, do make sure this reference is free from satellite tracks and star blooming. 

Ahh I must have missed that, it is indeed correct. But I am also unsure why they ask to realign.

Just as a quick test I duplicated that image a couple of times and tried to do a staralignment and it succeeded, Could you post a screenshot of your alignment settings, that may help.

Just took a look at your site and you have taken some beautiful images! Do you focus manually or do you use an autofocuser?

 

[groberts]

Thanks for your kind words Wyvern.  I've been very happy with the WO GT81 - though always thinking about something bigger, I'm put off by increased difficulties given our weather / limited imaging time and various other problems at my location.  Hopefully one day I'll be able to establish an dedicated observatory and move on to something more ambitious.  My images are all manual focus partly to keep it simple though I do own an electronic autofocuser - these things tend to sit in a cupboard for a couple of years and then possibly get used!

I remain baffled as to what's happening here + why a second alignment is required?  Did you use the given FITs file to align or convert to .xisf?

The aligment settings are shown at the top of this thread in my original post.

Graham     

[Wyvernp]

Oh of course! Not much sleep last night. 

This sounds like the cause of the issue, https://pixinsight.com/forum/index.php?threads/variant-todouble-invalid-conversion-from-bytevector-type.13914/ 

As you suspected it was to do with the fits files. It looks like someone on that thread managed to do a conversion and fixed the issue.

Quote

I found that I could do a batch conversion from fits to xisf and star alignment would work.  I just take my lrgb masters and apply the conversion from fits to xisf.

What I did probably wasn't a fair test as it was the same image. I'm also baffled about the second alignment.

[Wyvernp]

32 minutes ago, groberts said:

Thanks for your kind words Wyvern.  I've been very happy with the WO GT81 - though always thinking about something bigger, I'm put off by increased difficulties given our weather / limited imaging time and various other problems at my location.  Hopefully one day I'll be able to establish an dedicated observatory and move on to something more ambitious.  My images are all manual focus partly to keep it simple though I do own an electronic autofocuser - these things tend to sit in a cupboard for a couple of years and then possibly get used!

It's always great to see what others are achieving with similar equipment, Its lust such a shame I only managed to cobble my gear together just as there is very little darkness.

I'd also love an imaging scope with a bit more focal length for galaxies and planetary nebulae.

[groberts]

37 minutes ago, Wyvernp said:

Oh of course! Not much sleep last night. 

This sounds like the cause of the issue, https://pixinsight.com/forum/index.php?threads/variant-todouble-invalid-conversion-from-bytevector-type.13914/ 

As you suspected it was to do with the fits files. It looks like someone on that thread managed to do a conversion and fixed the issue.

What I did probably wasn't a fair test as it was the same image. I'm also baffled about the second alignment.

Thanks for the guidance but I've tried the file conversion to .xisf and still get the same problem = something seems very wrong here?

Edited June 5, 2020 by groberts

[Wyvernp]

Yeah I think I'm at the end of my limited expertise. Hopefully someone more experienced will be able to chime in.

[groberts]

I've just downloaded these stacks to PI again which on the face of it went OK but gave the following warnings:  

**Warning: Variant::ToDouble(): Invalid conversion from ByteVector type

and 

Warning Deprecated format: FITS

 

This seems to clearly indicate there's something wrong with the said files but what is it?  These were created in PI and then manually saved after registration - did I do something wrong in the saving  process perhaps?

 

[groberts]

For the record below is an answer I got from the PI Forum, which confirms my original suspicion that the problem was using FITs files. 

I had followed to the letter the excellent Light Vortex Tutorial which is shown working with FITs files but apparently things have changed with more recent versions of PI which will now only use .xisf files.  I've therefore had to restack to produce new .xisf files which have just now worked fine in the aligment routine,  which was where I originally crashed!

Onwards and upwards to the next phase of my PixInsight initiation.

Graham

Answer from PI forum:

 The problem is as you suspected, the .fit format is too restricted for advanced processing techniques and Star Alignment does not handle drizzle integrated images in .fit format in the latest PI platform.

This wasn't the case with earlier PI platforms and if you are following older tutorials they won't always point this out.

In general the authors of PI don't consider .fit to be a suitable format for advanced processing techniques and the platform doesn't really support .fit except to import raw or export completed images.

The error message is rather vague and unhelpful but that is the cause.

You can import the raw data as .fit or pretty much any other image format but once you are into advanced processing you should only use .xisf until the completion of all the processing stages.

Once the image has been successfully processed you can export as .fit or any of the other standard formats.

There was a previous question on the same error earlier this year and that was the reason given by Juan to that questioner.
 

Variant::ToDouble(): Invalid conversion from ByteVector type

Hi, StarAlignment process shows this bug error: "Variant::ToDouble(): Invalid conversion from ByteVector type". It's important because I can't align the images. Curious thing is that it appears also alignment the picture with itself !!! The versi?n of Pixinsight is 1.8.8-2 Ripley (x64). Trying...

pixinsight.com

Sorry to be the bearer of unwanted news but the only way forward is to start again and stick to .xisf throughout the process.

If possible, try to carry out the whole processing sequence in PI. It's not good practice to mix-and-match software during data processing as it makes it much more difficult to pin down where something went wrong in the sequence.

 

    

Edited June 9, 2020 by groberts

[Alxtrnk]

Hi to all, 

I am facing now the same problem as described in this post. I can not use star alignment process and I get the same error: Invalid conversion from byte vector type. I have tried directly saving all master fits files (L,R,G,B,Ha) into xisf files with command Save as, but when I try to carry out alignment with these xisf files I have no success and get the same error. 

You say that It is required to repeat the entire stacking saving all in xisf format, but as far as I know, this is not selectable. When image is integrated, this one appears in pixinsight screen directly in fits format (no option to select output format for integrated image). If I save it as xisf format, Am I not doing the same as I have tried, saving all masters as xisf at the end of the stacking? Do you maybe mean that when Lights are calibrated, Do I have to use master flat, master dark and master bias also in xisf format (now I have used these masters for lights calibration in fits format)?

Any help is appreciated, because I am stuck and can not align my masters. 

Regards. 

[Alxtrnk]

I reply myself.

Saving master flat/master dark/master bias as xisf files it works. I have repeated image calibration/integration per two channels and I have repeated stacking process saving all masters (also master lights) in xisf format and no problem with star alignment process.