Manifold denoising for cryo-EM data setsMS51

A key challenge in processing cryo-EM images is the overwhelming level of noise in them. We present an algorithm for denoising cryo-EM image sets, by exploiting the geometrical property that all underlying (unknown) clean images lie on a manifold of lower dimensionality. Each image is denoised by projecting it onto this (unknown) low-dimensional manifold. We show that all the quantities required to compute this projection can be estimated using only the two-dimensional images.

This presentation is part of Minisymposium “MS51 - Algorithms for Single Particle Reconstruction in Cryo-Electron Microscopy (cryo-EM). (3 parts)
organized by: Roy Lederman (Yale University) , Joakim Andén (Flatiron Institute) .

Yoel Shkolnisky (Tel Aviv University)
Boris Landa (Tel Aviv University)
computed tomography, graph laplacian, image reconstruction, image representation, manifold learning