In a shallow water environment, mud-rolls are often dominant and appear as a prevailing coherent linear noise in OBC seismic data. Their complex properties make the noise removal notably challenging in seismic processing. To address the challenges, we propose a method of dispersive multi-modal mud-roll estimation and removal, using the feedback-loop approach with a sparse inversion of 3D Radon transform. In this paper, we illustrate the proposed method, and will show some real data examples in our presentation at ADIPEC 2013 to demonstrate the virtues.