Removing outliers during frame ranking?

[johnmyleswhite]

I have several videos I've taken with my ASIAir Plus that have 1-2 corrupt frames in them. These corrupt frames look like some slices of the image have been left-shifted or right-shifted. I've reported the issue to ZWO already to resolve my problem during data collection, but I wonder if the PSS frame ranking process could be improved by accounting for outliers when measuring frame quality. In particular, a single frame exhibiting this form of corruption creates problems for the entire frame ranking step in PSS because the quality measures heavily reward the corrupt frames over all the other frames. I'm attaching images to show what's wrong in the frames and to show how it skews the entire quality distribution.

I'm happy to submit a PR to evaluate potential outliers and make it possible to exclude them from frame ranking if that's something you'd be open to.

[Rolf-Hempel]

Hi John,

Thank you for your comment on handling corrupted frames. I must admit that so far I did not pay much attention to this problem since this problem does not show in my videos. I agree, however, that an automatic exclusion of those frames could help some users. But we have to make sure that the exclusion algorithm is very robust, so that it will not exclude good frames,

A code change which solves this problem would be very welcome. Since your camera produces bad frames occasionally, you have the right test material at hand. Before submitting the PR, please test your algorithm very carefully that it does neither produce false positives nor false negatives.

I'm looking forward to hearing from you!

All the best, Rolf