Multiple Timescale Methods in Biomolecular Simulations and Nonlinear Diffusion of Images
Biomolecular dynamics (MD) simulations play an important role in elucidating essential functional properties in key biological processes. Because of the crucial link between structure and function, structural information derived from experimental studies can be refined and interpreted at atomic resolution using all-atom computer simulations. I will present a novel multiple timestep (MTS) integrator, designed to speed up the calculation in modeling Protein/DNA complexes by marching with large timesteps, describing its advantages and limitations.
Next, I will mention how MTS methods may be used to target specific scales within a digital image. I will show how marching with large timesteps in the nonlinear diffusion equation where the variable is an image leads to nonlinear filtering approaches, culminating with bilateral and mean-shift filtering.
Finally, I will present examples and future applications in imaging and biology for these methods.
Joint work with T. Schlick, H.H. Gan, D. Comaniciu, M. Israeli, R. Kimmel and other collaborators.