We present a method based on LSI to directly image the subsurface using blended data sets.
This method does not require any pre-separation of the blended shot gathers, and the crosstalk
is effectively removed by formulating the imaging problem as a least-squares inverse problem.
The inversion examples on the Marmousi model show that LSI can successfully remove the crosstalk
that migration generates and optimally reconstruct the reflectivity in the subsurface.