Question about the image processing

This is the track of a geostationary satellite recorded in HSC DR2.

Close inspection shows that adjacent to the bright dashed streak caused by solar reflection off the satellite body, there is a dark band where the hazy background that is present in the rest of the image field is missing. This dark band also has a slightly brighter edge at the point where it rejoins the hazy background. This edge brightening shows a periodicity that matches the periodicity of the main bright streak. So it could either be another reflection from a different part of the satellite, such as the ends of the solar panels that are obscuring the background view and hence causing the dark band, or it could be an artefact of the image processing software.

Image processing to combine multiple layers and build up the complete image is likely to soften the edges of imaged features, not to sharpen them, so I don’t believe this is a processing artefact. I think this track is actually pin-sharp imaging of the satellite itself which, only being present in one layer of the multi-layer final image, would not be softened by the image processing software overlaying multiple images of the same thing.

If that is correct, then I believe my interpretation of many of the short perfectly linear streaks that were discussed previously as micro-meteors rather than cosmic rays is supported, because the level of detail present here indicates that it would be perfectly possible for the image sensor to record a micro-meteor track as a single-pixel wide streak.


I don’t know all the details of the HSC image processing, but I would bet that if you looked at the raw images, you would see a solid streak the full width of the central ‘blink’ plus the empty region plus the brighter wing. The processing software has masked (ignored – hence no noise) the blank region, but somehow failed to mask the central blinking region. The fainter lines are just the edges of that wide streak.

In these pipelines that are tuned to recover accurate measurements of extremely faint objects, bright things just wreak havok, and processing pipelines tend to do different things with fully saturated pixels versus those affected by bleed trailing.

There’s no way that’s a resolved structure :slight_smile:


I think HSC images are coadded, so its possible thats the culprit.

Seems a bit pointless to mask part of a bright object and not the whole thing! But based on what you say the image processing is designed to do, this is what I think is happening with the above image:

There are possibly two reflections from this satellite, slightly out of phase (the smaller reflection above the main streak brightens a little after the main one). The software is designed to remove pixel-wide streaks from cosmic ray strikes on the detector, so if the reflection is small enough the software will blank out the weaker reflection almost completely. The software also detects the edge of the main reflection and blanks out a pixel-wide band of data. But it then detects the main reflection and considers that it may be a star, so it stops blanking the data after the first pass. This leaves a dark band all around the main reflection, and a dark band just above and to the right of it where the secondary reflection has been almost completely removed. The bright “halo” alongside the blanked region is due to the spreading of the image beyond a single digital pixel, hence it is not fully removed by the software.

That would explain this phenomenon.

If you’re really interested, I would suggest trying to track down the single exposure / raw image. It should be available at
though these HSC images are coadds of many exposures, so there may be lots to look through.

It’s also worth computing how big a pixel of HSC (0.168") projects to at geosync distance :wink:


Hi Dustin

Unfortunately it says “STARS is a private archive for astronomers with approved observation-proposals on the Subaru Telescope, and their staff.”

So I can’t browse their data. :triumph:

Oh – you could also try the SMOKA archive, – looks like their user registration form has a checkbox for amateur astronomers.
(Can you tell I’ve never tried to track down raw HSC images?)

1 Like

Hmmm… that looks trickier than I thought. Maybe if I have a few idle hours to spare I might try again!