BNMC Seminar - In Promptu Ponere: Removing Noise in Cryoimaging

Alexandre Cunha
CACR and the Center for Integrative Study of Cell Regulation
Caltech

Electron cryomicroscopy is a remarkable technology enabling new discoveries at sub-cellular scales making the cryoelectron scope one of the most valuable and advanced assets in the toolbox of the contemporary structural biologist. This technology is still young and there is much to do to improve the cryoimaging pipeline, including the development of suitable image processing algorithms.

In this presentation I will describe my inroads into filtering typical, highly noisy cryoimages. I will review some recently proposed variational models for denoising, recall some classical filters which are still frequently used by practitioners, and talk about our new developments in nonlocal filters capable of producing significantly better results. I will illustrate with cases in Caulobacter crescentus tomograms, in large image sets used for single particle analysis, and in MRI.

I will also comment on the computational aspects for efficiently processing large, high resolution images containing millions to billions of pixels/voxels, and how we are using parallelism and multicore computing to achieve ten to hundred fold speed ups.

This is joint work with Grant Jensen and Jerome Darbon.