Hacker News Reader

Comment by pppone | Hacker News Reader

Yes, and it is already happening in professional astronomy. For example, the "Antarctic Tianmu Plan" [0] have shown that you can successfully capture non-trailed images without using tracking mounts by using drift-scanning CCDs—basically letting the sky move across your sensor while the detector is read out at the same rate.

[0] - https://doi.org/10.1117/12.3019468

View on HN

You can, but dark noise is a problem with this technique as your SNR per bucket ends up being low. The purpose of long exposures with tracking is to maximise your SNR.

Also, it helps significantly to be in Antarctica, where the relative movement is much slower than it is at lower latitudes — and to have multiple telescopes - and low noise CCDs, in a cold, dry environment.

Sadly, most of us don’t have those luxuries.

what about computational methods? i have always wondered how stacking many short exposures without tracking compares to deconvolution of a single long exposure. it seems that there is software able to do this by taking into account both motion blur and the PSF of the imaging system:

https://siril.readthedocs.io/en/stable/processing/deconvolut...

The problem is that the noise can swamp the signal. Another example of this would be doing astrophotography during the day. The sun doesn't block anything, it just makes the sky glow with "noise". Theoretically it has exactly as much signal from space as it does at night, but because the sun adds so much noise it's completely lost.

> "because the sun adds so much noise it's completely lost."

Do you mean that it would be conceptually possible to image planets or even deep-sky objects during the day with incredibly efficient denoising software? (I am a noob in astronomy)

I would be, yes. As early as the 1950’s, several avionics companies made daylight-capable star trackers (for jam-resistant long-distance airplane navigation) using chopper techniques. Those trackers were mostly mechanical, except for electronics to demodulate the star signal from the single pixel sensor.

I suspect diffusion models can shine at denoising single shot deep sky images. Will be attempting when I find bandwidth. I do a lot of deep sky landscape photography (IG: @dheeranet) and I want to do them in one go instead of stacking ground (untracked) and sky (tracked) separately.

Reply on news.ycombinator.com