In-Person Event

*This event is in-person and open only to Princeton University ID holders*

The contrasts of Cryo-EM images vary from one to another, primarily caused by uneven thickness of ice layers. These various contrasts often affect the results of image-denoising and CTF correction, and eventually lead to inaccuracy in 3D reconstruction. Currently, the contrasts of images are often estimated in the iterative refinement stage, but correcting the contrasts at earlier stages is believed to be more beneficial. However, with low SNR and CTF, estimating contrasts from noisy images is challenging. In this talk, I will first talk about the previous work on covariance wiener filtering (CWF). Then, I will introduce our method that jointly denoise image, correct CTF and estimate contrast. This method is based on a variant of CWF and assumes contrasts as i.i.d. random variables that are independent of the underlying clean images. We demonstrate that our method achieves better results on both image denoising and contrast estimation compared to those of CWF with image normalization.