Structure and View Estimation for Tomographic Reconstruction: A Bayesian Approach

摘要: This paper addresses the problem of reconstructing density a scene from multiple projection images produced by modalities such as x-ray, electron microscopy, etc. where an image value is related to integral along 3D line segment between radiation source and point on plane. While computed tomography (CT) this when absolute orientation plane directions are known, orientations unknown - it akin structure-from-motion (SFM) extrinsic camera parameters unknown. We study within context protein macro-molecules in Cryogenic Electron Microscopy (cryo-EM), very noisy existing techniques use several thousands images. In non-degenerate configuration, viewing planes corresponding two projections, intersect 3D. Using geometry imaging setup, possible determine projections planes. turn, can be formulated type orthographic structure motion correspondences views unreliable due noise. formulate task denoising correspondence matrix present Bayesian solution it. Subsequently, each determined followed reconstruction. show results cryo-EM proteins compare our that Micrograph Analysis (EMAN) widely used reconstruction tool cryo-EM.