Electron tomography provides opportunities to determine three-dimensional cellular architecture at resolutions high enough to identify individual macromolecules such as proteins. Image analysis of such data poses a challenging problem due to the extremely low signal-to-noise ratios that makes individual volumes simply too noisy to allow reliable structural interpretation. This requires using averaging techniques to boost the signal-to-noise ratios, a common practice in electron microscopy single particle analysis where they have proven to be very powerful in elucidating high resolution molecular structure. Although there are significant similarities in the way data is processed, several new problems arise in the tomography case that have to be properly dealt with. Such problems involve dealing with the missing wedge characteristic of limited angle tomography, the need for robust and efficient 3D alignment routines, and design of methods that account for diverse conformations through the use of classification. We hereby present a computational framework for alignment, classification and averaging of volumes obtained from limited angle electron tomography, providing a powerful tool for elucidation of high resolution structure and description of conformational variability in a biological context. Index Terms Tomography, Image registration, Image classification, Clustering methods