| Abstract: |
| This work devises a fast method for estimating the covariance matrix of tomographic projection images affected by high levels of noise in cryo-electron microscopy. The mathematical model for these images consists of observing noisy samples of projections of randomly rotated variables, convolved with oscillatory functions. The method relies on a novel algorithm to expand discretized images in the Fourier-Bessel basis (the harmonics on the disk), which enables a compressed representation and fast estimation of the covariance matrix. |
|