Fast imaging with multiples and source estimation

Multiples are usually treated as unwanted components in seismic data. However, if correctly used, they can provide valuable information about the subsurface. In this presentation, I will talk about how to make use of multiples in $\ell_1$ regularized least-squares imaging, and how to estimate the source wavelet on the fly using multiples. Synthetic examples show that using multiples not only helps to retrieve the true-amplitude seismic image and the source wavelet, it also increases the wavenumber contents in the seismic image. By using dimensionality reduction techniques with rerandomization, we also greatly decrease the simulation cost without compromising the image quality.