In aperture synthesis images noise is partially correlated across all pixels of the image.
This effect is incorporated in the error bars on the source parameters in a heuristic manner: any variance further away than the FWHM clean beam is considered uncorrelated, while any image noise at a pixel position nearer than the FWHM clean beam is considered fully correlated.
This is effectuated in calculate_correlation_lengths:
return 2.0 * semimajor, 2.0 * semiminor
This is hard-coded now - and copied from the NVSS paper (equations 25/26) - but we should offer the option of changing the correlation lengths through an ImgConf setting.
For example, for images from observation with very poor uv-coverage, one may want to increase the correlation lengths beyond the FWHM clean beam.
In aperture synthesis images noise is partially correlated across all pixels of the image.
This effect is incorporated in the error bars on the source parameters in a heuristic manner: any variance further away than the FWHM clean beam is considered uncorrelated, while any image noise at a pixel position nearer than the FWHM clean beam is considered fully correlated.
This is effectuated in calculate_correlation_lengths:
This is hard-coded now - and copied from the NVSS paper (equations 25/26) - but we should offer the option of changing the correlation lengths through an ImgConf setting.
For example, for images from observation with very poor uv-coverage, one may want to increase the correlation lengths beyond the FWHM clean beam.