EMAN software is a collection of scientific image processing tools created primarily for the transmission electron microscopy (TEM) community.
EMAN is a suite of scientific image processing tools that is primarily intended for the transmission electron microscopy community, although it is increasingly being used in other fields. The software has a specific focus on the process of single particle reconstruction, which involves collecting images of nanoscale molecules and molecular assemblies embedded in vitreous ice on a transmission electron microscope, and then processing them using EMAN to create a full 3-D reconstruction.
EMAN's processing capability allows low resolution structures of around 2 nm to require about eight hours of computer processing and several thousand particles. However, structures aimed at around 0.5 nm or better resolution could require hundreds of thousands of particles and hundreds of thousands of CPU-hours on large computational clusters. In fact, EMAN is often used in supercomputing facilities as a test application for large-scale computing.
Scientific image processing entails analytical processing, which differs significantly from typical 'Photoshop' image processing. EMAN processes images as floating point grayscale images, with pixel values in the images represented as real numbers rather than small integers. Processing often includes complex image processing operations in Fourier or Wavelet space.
EMAN was first released in 1999 and has undergone continuous development since then. It comprises a C++ library of hundreds of different image and volume processing algorithms with bindings into the popular Python scripting language. In new EMAN development, all user-level programs that are over 200 in EMAN 1.8 are developed in Python. Thus, a knowledgeable end-user can make modifications without downloading or compiling any of the C++ source code.
In the latest release, EMAN has made substantial improvements in refine2d.py, and refine2d.py is highly recommended for every new data set. Some programs use the EMAN2 style of arguments rather than the old style, and there is a new HDF5 format compatible with EMAN2. New AIRS programs have been added, such as "skeleton", and new options have been added to make refinement work better on large icosahedral objects. Although network-related problems may still exist, the parallelism infrastructure has been improved. Lastly, random model generation in makeinitialmodel.py has been fixed.
EMAN's processing capability allows low resolution structures of around 2 nm to require about eight hours of computer processing and several thousand particles. However, structures aimed at around 0.5 nm or better resolution could require hundreds of thousands of particles and hundreds of thousands of CPU-hours on large computational clusters. In fact, EMAN is often used in supercomputing facilities as a test application for large-scale computing.
Scientific image processing entails analytical processing, which differs significantly from typical 'Photoshop' image processing. EMAN processes images as floating point grayscale images, with pixel values in the images represented as real numbers rather than small integers. Processing often includes complex image processing operations in Fourier or Wavelet space.
EMAN was first released in 1999 and has undergone continuous development since then. It comprises a C++ library of hundreds of different image and volume processing algorithms with bindings into the popular Python scripting language. In new EMAN development, all user-level programs that are over 200 in EMAN 1.8 are developed in Python. Thus, a knowledgeable end-user can make modifications without downloading or compiling any of the C++ source code.
In the latest release, EMAN has made substantial improvements in refine2d.py, and refine2d.py is highly recommended for every new data set. Some programs use the EMAN2 style of arguments rather than the old style, and there is a new HDF5 format compatible with EMAN2. New AIRS programs have been added, such as "skeleton", and new options have been added to make refinement work better on large icosahedral objects. Although network-related problems may still exist, the parallelism infrastructure has been improved. Lastly, random model generation in makeinitialmodel.py has been fixed.
What's New
Version 1.8: N/A