[X] My Assistant

[Guest_Sunspot_*]

http://hirise.lpl.arizona.edu/nea.php


Rover tracks at Victoria
http://hirise.lpl.arizona.edu/PSP_004289_1780

[OWW]

Good to know that someting can still be done about the noise.
But will these improvements to the calibration algorithms really remove the noise, or just act as some sort of blur filter to smear out the noise?

I also noticed that the 'vertical lines'-artifacts are more noticable in shaded areas. (see picture).
Is this also the result of calibration problems? Is the S/N ratio for the bright terrain high enough for the current calibration algorithms to work, but not for the dark surfaces?

Attached thumbnail(s)

 

[DataMiner]

Good to know that someting can still be done about the noise.
But will these improvements to the calibration algorithms really remove the noise, or just act as some sort of blur filter to smear out the noise?

I also noticed that the 'vertical lines'-artifacts are more noticable in shaded areas. (see picture).
Is this also the result of calibration problems? Is the S/N ratio for the bright terrain high enough for the current calibration algorithms to work, but not for the dark surfaces?

Well, noise removal is usually done at the pixel level, meaning that you won't generally see smearing from removing systematic noise like the dark current. Generally speaking a bad noise removal algorithm will either not remove enough noise or actually add more noise to an image making it more "grainy" rather than smearing out the image. We're trying to avoid the use of smoothing algorithms for noise removal since those can smear things out a bit, although in our tests, we've found that smoothing doesn't seem to smear things out much at all, so if it comes down to it, it is option for extreme cases, but smoothing isn't part of our current plans.

The vertical lines are affected by the calibration problems, and there are some additive and multiplicative systematics that can make them more apparent (especially when you throw in the need to match the signal levels of the different CCD's). Generally, the lower the signal to noise the worse the current calibration works, although you can also see vertical striping in images were the signal to noise is still ok, but the image itself is rather flat. In these cases there isn't much contrast between the image data and the striping so low level pixel to pixel differences tend to be enhanced in these images.