Is your feature request related to a problem? Please describe.
Most real astronomical images have some covariance between adjacent pixels. This can be from charge diffusion, or other effects.
Describe the solution you'd like
While a full covariance matrix for all pixels would be astronomically large, a covariance kernel can represent the covariance of a small window of neighboring pixels. This can then be applied int he likelihood calculation as a convolution operation if formatted correctly.
Describe alternatives you've considered
Currently no covariance is accounted for and this works fairly well, though it may be ill suited for pushing the detector limits in very faint cases.
Is your feature request related to a problem? Please describe.
Most real astronomical images have some covariance between adjacent pixels. This can be from charge diffusion, or other effects.
Describe the solution you'd like
While a full covariance matrix for all pixels would be astronomically large, a covariance kernel can represent the covariance of a small window of neighboring pixels. This can then be applied int he likelihood calculation as a convolution operation if formatted correctly.
Describe alternatives you've considered
Currently no covariance is accounted for and this works fairly well, though it may be ill suited for pushing the detector limits in very faint cases.