search

lod / seek-thermal-documentation

Unofficial documentation on iteracting with the Seek Thermal hardware
Creative Commons Zero v1.0 Universal
76 stars 16 forks source link

Figuring out Initial Frames and Fixed Pattern Noise Corrections #2

Closed zougloub closed 9 years ago

zougloub commented 9 years ago

The imager sends a series of initial frames:

  1. ID 4 - Looks like it at least contains recorded info about dead pixels (the zero-valued pixels) and the other pixels might be indicative of (pure speculation) integration rates frame-000-raw
  2. ID 9 - Two gradients.. mystery frame-001-raw
  3. ID 8 - Almost looks like a shutter frame frame-002-raw
  4. ID 7 - Something rather uniform... and light frame-003-raw
  5. ID 10 - Something rather uniform... and dark frame-004-raw
  6. ID 5 - Weird frame-005-raw

whose meaning and use is not exactly clear (Frame.md), but they should be here for a reason.

Then there are the frequent sequence of ID 6 / ID 1 frames, and everybody agrees that ID 1 should be used as a reference for the pixel baseline values, and its frame should be substracted from regular (ID 3) frames.

Once we perform the reference frame substraction, and perform dead pixel interpolation, we can see that there is fixed pattern noise, with some pixels integrating faster than others.

There is that 207th column which provides information used to correct for horizontal bands, how it does that is a bit of magic at the moment.

We can see that images exhibit more than just these horizontal bands, but how would we correct for them (except from performing median/bilateral/fancier filtering, which are just lesser solutions if a clean one exists)...

lod commented 9 years ago

Thanks for the images, I have incorporated them into the wiki