Ringing artifacts introduced near bright point sources using sparse_deconv_condatvu

[MTLeist255]

Hello,

I was wondering if there was some parameter I could vary using the sparse_deconv_condatvu deconvolution method to reduce ringing artifacts found near bright point sources in my images? I'm currently calling the method as is via:

sparse_deconv_condatvu(image, psf, n_iter=100)

Attached is a screenshot of the result (please ignore the iteration number inset, this was just an example showing the ringing). I've also attached the original image (before deconvolution) for reference. As you can see there are some very strong ringing artifacts found near the bright central point source in my image, I was wondering if there was some parameters I could vary with this method to reduce this ringing effect?

Best, Mason

[sfarrens]

Hi @MTLeist255, I will have to think about this.

@jstarck, have you come across something like this before? It's just a starlet deconvolution using Conda-Vu.

[MTLeist255]

Hi @sfarrens thank you! Out of curiosity, is the proper name for this method starlet deconvolution using Conda-Vu? My understanding from the PySAP paper (https://arxiv.org/pdf/1910.08465.pdf) and the deconvolve.py source code was that this method was called Sparse regularization using Condat-Vu, what would be the proper title for this method?

[sfarrens]

Hi @MTLeist255, yes indeed. The starlet is just the specific wavelet transform that we have implemented for this method and hence the aspect that regulates the sparsity of the deconvolution. Conda-Vu is the specific primal-dual splitting algorithm we have implemented to solve the inverse problem.