Registering JWST NIRCam images of M74 Challenge

GaryP

GaryP

Well-known member
I have been trying to register six JWST NIRCam images of M74 with only partial success. Using StarAlignment with f187n as my reference (because it has lowest eccentricity), I have been able to register three images only, and only the f200w showed no rotation or translation on comparison. Using DynamicAlignment with reference star f115w, I can register all the images, but on inspection I can see that the rotations and translations are slightly off. Only f335 looked satisfactory on inspection. If anyone wishes to try and share their results, the files are available on OneDrive at folder M74. Note that the f187n is heavily banded---another problem I will have to deal with. Meanwhile, I will keep trying.

M74 XISF FILES

To thicken the plot, I downloaded four MIRI files of M74. I ran star alignment with f2100w as the reference file and had no success until I increased the Noise Reduction to 5. That produced one aligned file: f1000w. When I compared the two views, there was obvious rotation.

* Projective transformation matrix:
+1.014955 +0.006645 -8.464607
+0.004849 +1.017255 -8.987534
+0.000005 +0.000008 +1.000000
translation : 12.35 px
translation_x : -8.46 px
translation_y : -8.99 px
rotation : +0.01 deg
scale : 0.979
scale_x : 0.985
scale_y : 0.983
Generating registered image
Homographic Projection / Lanczos-3 interpolation, c=0.30: done
Registration successful.
Writing output file: /Users/garypalmer/Documents/ASTRONOMY/JWST/M74/MIRI/MIRIREG/jw02107-o039_t018_miri_f1000w_i2d_r.xisf

The MIRI files are also posted.
 
Last edited:
Lack of Response Perfect So Far

I like to think that the PI Forum offers a strong argument for purchasing PI. Good advice from members augments good support for the software. But it is hard to make the argument when a processing challenge goes uncommented, perhaps unnoticed, for three days. The challenge seemed to me a reasonable one. I researched it on this forum and Cloudy Nights; I read my manuals and tried to apply their hints; and I posted the problematic files as repeatedly recommended by Juan in other threads. Photos published by NASA and others on the internet suggest that someone knows how to solve the problem of registering JWST images of M74. So why has my challenge gone unremarked and unsolved? Are there just too few people interested in processing James Webb images? Did I pose the problem awkwardly? Is the solution so well known that the challenge is too easy? Is the challenge too hard? Are the active problem solvers overworked? Should I have waited longer? Will no one step up to the plate to enhance the content of our forum and the reputation of our software?
 
I think there are a couple of factors. It was mid week so most people were working and the number of forum members processing JWST is pretty small.

I have not looked at the M74 images but other JWST data sets I have looked at have been pretty tricky to work with. I ended up with something but nothing like the published image. There are papers online that describe the pre-processing pipeline but most of that went over my head.
 
I got the Mira Data to register using F1000W as a reference with the settings in the following screenshot.

Desktop.jpg
 
@chris.bailey, Thanks. That’s encouraging and I will try it with F1000W. Did you try comparing Image08 with F1000W using CMD-fn-pagedown or something equivalent?

You are probably right on both points regarding the lack of response.
 
I added all four together and they align pretty well. As I understand it, the published images are upsampled using AI NASA is training and F1000W was used as a luminance channel. The others are a strange mix of red, orange and cyan. Haven’t tried all of that yet but may do when the rain comes in later.
 
I used the parameters showing on @chris.bailey’s screen shot, but my choices were not exactly the same. We evidently have different versions of PI, and my result was different.

Console:

StarAlignment: Processing view: jw02107_o039_t018_miri_f770w_i2d
Loading reference image:
<path>/JWST/M74/MIRI/MIRIXISF/jw02107-o039_t018_miri_f1000w_i2d.xisf
Loading image: w=2379 h=1178 n=1 Gray Float32
4 image properties
86 FITS keyword(s) extracted.
<path>/JWST/M74/MIRI/MIRIXISF/jw02107-o039_t018_miri_f1000w_i2d.xisf:
Noise reduction: done
Structure map: done
Detecting stars: done
0 PSF fits.
*** Error: No stars found
<* failed *>

The console readout and the screen shot were edited after I noticed that Chris used a different registration model.
 

Attachments

  • M74_StarAlignment2.jpg
    M74_StarAlignment2.jpg
    472.1 KB · Views: 51
Last edited:
Suggest updating to 1.8.9-1 as it works fine for me. Don't bother ticking Local Distortion as that won't work

Bit of messing about and I got pretty close to the published images using the colour mapping described in the accompanying article i.e 2100 to red, 1130 to orange and 770 to cyan and then using 1000 as luminance.

Processed_Final.jpg
 
@chris.bailey That looks great! I wasn't aware of the update, so I will do that and try again. Do you have a link for the article? I'm sure others would be interested.
 
I checked again and changed another parameter to match @chris.bailey's window and obtained alignment on all four MIRI files. I can't match the rest of his processing, so I won't display my results here. Problem solved.
 
GaryP said:
I checked again and changed another parameter to match @chris.bailey's window and obtained alignment on all four MIRI files. I can't match the rest of his processing, so I won't display my results here. Problem solved.
Click to expand...
Feel free to ask for help though I’m kind of glass of red and click some buttons kind of image processor. The more wine … the better it looks!
 
I will thanks. Did you have any particular reasons for your settings in StarAlignment on Detection Scales, Peak Response, and Bright Threshold?
 
GaryP said:
I will thanks. Did you have any particular reasons for your settings in StarAlignment on Detection Scales, Peak Response, and Bright Threshold?
Click to expand...
As above … glass of red wine … fiddle until they all registered. On detection scales, a lot of small nonstellar details, noisy so upped noise settings the rest ………
 
chris.bailey said:
As above … glass of red wine … fiddle until they all registered. On detection scales, a lot of small nonstellar details, noisy so upped noise settings the rest ………
Click to expand...
Does StarAlignment remove noise or just work around it? I'm thinking it could be detail, like dust, that obscures stars, but would be nice to save.
 
To follow up on Chris’s information, I’m not sure what it means to assign f2100w to red, f1300w to orange, and f770w to blue. To integrate the images, I ran PixelMath with the following

RED: f2100w*0.7+f770w*0.3+f1130w*0.7
GREEN: f1130w*0.5
BLUE: f770w

The assignments are based roughly on the theory that red is red, orange is .7 red plus .5 green (on scale of 1), and cyan is .3 red and 1.0 blue. I base that on a color chart that I made using PixelMath. But this doesn’t take into account the large differences in flux among the three filters. I was reluctant to try to equalize them, because they are telling us only what they are seeing.

Then I ran ColorCalibration and let it do whatever it does. It seemed to do what it is designed to do. Without benefit of any relaxing beverages, I did some histogram transformations. The result is not as pretty as your posted image or Judy Schmidt’s image, but in my defense I would say that most of the light in this MIDI set comes from the farthest infrared filter, and the photons appear to be coming through and from interstellar gas and dust. So perhaps there is some value in displaying it in the reddish brown hues. It sacrifices the nice blue in the core, but one can see that the core is hotter than the dust because it shades to purple. I am also intrigued by the fact that a few of the stars (or clusters — I can’t tell) are red or deep orange. Are these old stars?

I also wonder about the gradient to lighter shades on the right hand side. Can we trust the JWST process to have removed any gradients created by the camera? Is that a real difference in the galaxy that we see?

I realize I’m causing this thread to veer off the topic of alignment, but I’m not sure how best to continue the discussion of how best to process this JWST image using PixInsight.
 

Attachments

  • M74_240415.jpg
    M74_240415.jpg
    472.1 KB · Views: 52
Last edited:
Hi @GaryP ,

Sorry if I am late to the JWST Processing party!

One of my favourite things to do is process JWST data as it has its own unique challenges.

Below is my attempt using PixInsight only (no other tools).

My Process:
  1. Star Alignment
  2. Dynamic Crop
  3. Linear Fit
  4. Combined the images
  5. Stretched the Image
  6. Extract the stars
  7. Color Masks
  8. Luminance Mask
  9. Local Histogram
  10. Dark Structure Enhance
  11. SCNR on the stars
  12. Combined the Stars Back
Looks pretty nice printed.

ngc628-m74.jpg
ngc628-photo-frame-template-straight-on-square.png
 
You must log in or register to reply here.
Back
Top