Code review #521


Check SWarp's weight scaling when input and output images have different pixel scales

Added by Emmanuel Bertin almost 7 years ago. Updated almost 7 years ago.

In Progress
Start date:
Due date:
09/21/2015 (over 6 years late)
% Done:


Estimated time:
0.50 h
Spent time:
Actions #1

Updated by Emmanuel Bertin almost 7 years ago

  • Status changed from New to In Progress
  • % Done changed from 0 to 80
The astrometric scaling of pixel fluxes works in the following way in SWarp:
  • If $a$ is the output pixel area divided by the input pixel area, then the variance map pixel is multiplied by $a^2$. This is correct in the sense that it represents the variance of individual output pixels.
  • However, when the output pixel scale is much smaller than the input one, the values of neighbouring output pixels are totally correlated, hence the actual variance of the pixel flux integrated over some (large) image area is much larger than the one that would be computed by summing output variance map pixels over the same area. For this purpose, it would make sense to apply a factor $a$ to input variance map pixels instead of the current $a^2$.
Both approaches to variance scaling make sense, depending on the use case. So one could either:
  1. Leave things as they are now, being clear in the documentation about the choice that was made, or
  2. Replace $a^2$ with $a$ in the code, or
  3. Add a new variance scaling option.
    Option 2 should be preferable to option 1, as it fits with the notion that resampled images should not be considered at the scale of individual pixels, but rather at the scale of objects. It must now be decided whether Option 3 would really be useful.

Also available in: Atom PDF