[Enhancement] Improve the Smap estimation process for Self Calibration reconstruction

[chaithyagr]

From reference : https://hal.inria.fr/hal-01782428v2/document,

we notice that the Smap estimation involves:

1. Choose a percentage of kspace center
2. Obtain Sensitivity by NUFFT adjoint operation or gridded Inverse fourier transform

This below step is mentioned in the reference , but we have not implemented this yet:

Noise attenuation in the image background is achieved by masking the estimated sensitivity maps. This binary mask is actually computed by thresholding the mask, where the actual value of the threshold is given by a 2-cluster k-means algorithm. The binary mask is eventually defined as the largest connected component.

This method seems to be vital for background noise separation. However, this could get tricky with actual data, however, it is worth implementing such an utility function and allowing users to use it if needed.

[chaithyagr]

@Daval-G , this is something you might need and what we were discussing about.