search

jeff-regier / KillerAsteroids.jl

Detects asteroids that might kill us
Other
0 stars 1 forks source link

take image masks into account #24

Open ameisner opened 8 years ago

ameisner commented 8 years ago

Each WISE exposure comes with a corresponding bitmask image flagging bad pixels. I think it makes sense to keep track of these bad pixels and simply leave them out of the likelihood computation altogether. We could handle this by having type Image also store a mask. I can try to add this.

jeff-regier commented 8 years ago

Yep, makes sense. I think just add mask::Bool to Image (or is a mask not just binary?) and then skip the "masked" pixels in the loop around line 54 of model_probability.jl.

ameisner commented 8 years ago

Well, the published mask file has many different bits, e.g. 2^0 for cosmic rays, 2^1 for static bad pixels, 2^2 for saturated pixels, etc...When I write the utilities to load in the mask files, I may want to enforce some predefined definition of "good" versus "bad" pixels based on a combination of all bits, and then downstream analyses would only see binary mask images. But I guess that's not set in stone until I put the basic WISE utilities in place...