6 ms·

Why do stars appear spike-y in the new images?

perihelions 4y ago
Diffraction spikes from occluding support poles: https://bigthink.com/starts-with-a-bang/james-webb-spikes/ https://bigthink.com/starts-with-a-bang/james-webb-spikes/
- monkeybutton 4y ago
  That's a great explanation, thanks!
  - Ono-Sendai 4y ago
    I wonder why they don't seem to attempt to deconvolve the image.
    - supernova87a 4y ago
      I don't think that's possible, to recover the true image from the light that's been through the diffraction pattern. It would be like if you were able to resharpen an image that had been passed through a blur filter. You would be reconstructing / creating information that had been lost.
      - NavinF 4y ago
        > resharpen an image that had been passed through a blur filter This is trivial if you know what filter was used and it’s possible even if you don’t know the filter: https://en.m.wikipedia.org/wiki/Blind_deconvolution https://en.m.wikipedia.org/wiki/Blind_deconvolution Note that when you blur an image in software, information is only lost when you round each channel to the nearest u8. Each of the NxN output pixels is a linear combination of the NxN input pixels. There are only N^2 equations with N^2 unknown variables.
      - HPsquared 4y ago
        There wouldn't be any more information, but it'd look nicer and might perhaps be easier to process / interpret without the spikes.
      - dTal 4y ago
        It's absolutely possible, and it may surprise you to learn that blurring is not (by itself) a lossy operation and you can, in theory, unblur an image if you know the blurring function. The practical difficulty in doing so is that the blurred image has a lot of information in the low bits, which tend to get distorted through other processes (quantization, sensor noise, lossy compression). I too am surprised that no attempt appears to be made to remove the diffraction spikes from JWST's images, since the function should be known with great precision.
    - kurumi 4y ago
      These are calibration test images. It wouldn't make sense to use software to improve the image before the hardware is well calibrated.

Diffraction spikes from occluding support poles: https://bigthink.com/starts-with-a-bang/james-webb-spikes/ https://bigthink.com/starts-with-a-bang/james-webb-spikes/

monkeybutton 4y ago
That's a great explanation, thanks!
- Ono-Sendai 4y ago
  I wonder why they don't seem to attempt to deconvolve the image.
  - supernova87a 4y ago
    I don't think that's possible, to recover the true image from the light that's been through the diffraction pattern. It would be like if you were able to resharpen an image that had been passed through a blur filter. You would be reconstructing / creating information that had been lost.
    - NavinF 4y ago
      > resharpen an image that had been passed through a blur filter This is trivial if you know what filter was used and it’s possible even if you don’t know the filter: https://en.m.wikipedia.org/wiki/Blind_deconvolution https://en.m.wikipedia.org/wiki/Blind_deconvolution Note that when you blur an image in software, information is only lost when you round each channel to the nearest u8. Each of the NxN output pixels is a linear combination of the NxN input pixels. There are only N^2 equations with N^2 unknown variables.
    - HPsquared 4y ago
      There wouldn't be any more information, but it'd look nicer and might perhaps be easier to process / interpret without the spikes.
    - dTal 4y ago
      It's absolutely possible, and it may surprise you to learn that blurring is not (by itself) a lossy operation and you can, in theory, unblur an image if you know the blurring function. The practical difficulty in doing so is that the blurred image has a lot of information in the low bits, which tend to get distorted through other processes (quantization, sensor noise, lossy compression). I too am surprised that no attempt appears to be made to remove the diffraction spikes from JWST's images, since the function should be known with great precision.
  - kurumi 4y ago
    These are calibration test images. It wouldn't make sense to use software to improve the image before the hardware is well calibrated.