Main

CITATION NOTE

If you use XMSFilter for your research, please cite it: http://www.ncbi.nlm.nih.gov/pubmed/20802209


Introduction


XMSF (Microscopy Mean-Shift Filtering) is a tool for denoising and pre-segmenting noisy electron microscopy tomograms. It could also be successfully applied to X-ray tomography and other imaging techniques where gray scale images are used. XMS supports both 2D and 3D images and is able to deal with different image formats including SPIDER, MRC, IMAGIC, etc.. File name extensions are used to infer which file format is being used for both reading and writing.

Getting the software


XMSF can be obtained in two different ways:
  • As a stand-alone statically compiled executable: executables are provided for Linux. Just copy to any destination folder you want, add specific environment variables and use it.
  • As part of the Xmipp image processing software package: Go to Xmipp page, download the software and compile it following instructions. The program will the be invoked as 'xmipp_mean_shift'.

Usage


XMSF is a command-line application. Whether you are invoking it from the Xmipp installation or from the stand-alone application, the set of arguments to pass to it are:


Compulsory arguments:

  • -i [image in] OR [volume in] Input image or volume to be filtered. Different image formats are admitted (image format is inferred from file name extension)
  • -o [image out] OR -o [volume out] Where the result is written. Different image formats are admitted (image format is inferred from file name extension)

Optional arguments:

  • -hs Maximum distance between neighbor pixels to be taken into account. If not provided, a default value of '6' is used.
  • -hr Maximum color distance between adjacent pixels to be taken into account. If not provided, stddev is calculated for the input image and is used as the default value.
  • -thr Number of threads to use. If not provided, one (sequential algorithm) is used
  • -use_gauss Use gaussian normal kernels instead of the faster computing method. This method implements the original mean-shift algorithms without performance improvements.
  • -iters Number of iterations to be performed. If not provided one is used.
  • -save_iters Save resulting filtered volume after each iteration. If not provided, just an output volume is created at the end of the process.