A Generic Framework for Image Processing

Image processing example

The MeVis Image Processing Library

A fundamental part of MeVisLab is the object-oriented MeVis Image Processing Library (ML) providing a generic framework for image processing.

Each algorithm is represented as a self-descriptive module inside the development environment. Via an intuitive graphical user interface, these functional units can be combined to form complex and powerful image processing networks that are executed by a core image processing controller, the ML host.

Some important ML characteristics are:

  • Request-driven, page-based, multi-threaded image processing controller
  • Priority-controlled page cache to avoid unnecessary recalculations
  • Advanced support for various types of kernel filters and other local and global image filters
  • Special "Virtual Volume" support for global access to large images
  • Support for higher-dimensional images, time series, as well as tensor and complex valued images
  • Various optimization options for different types of image processing algorithms

Image Processing Modules

Currently, more than 5000 image processing modules are available, including image filtering, segmentation, and statistical analysis.

This includes, for example:

  • Filters: Diffusion filters, morphology filters, kernel filters, Hessian, and vesselness filters
  • Segmentation: Region growing, live wire, fuzzy connectedness, threshold, manual contours
  • Transformations: Affine transformations, distance transformations, projection and Radon transforms, manual registration
  • Statistics: Histograms, global image statistics, box counting dimension
  • Other: Unary/binary arithmetic, resampling/reformatting, dynamic data analysis, noise/test pattern generators

Other Data Processing Modules

Besides image processing, MeVisLab offers a rich toolkit of modules for processing different types of data objects, including time series data, histograms, surface meshes, marker lists, contour sets, and other geometrical objects.

Winged Edge Meshes

The Winged Edge Mesh (WEM) library provides unified algorithms for the generation, processing, and rendering of geometrical surface representations.

Some of the functionalities offered for common mesh processing tasks are:

  • Iso-surface generation
  • Surface refinement, smoothing, or complexity reduction
  • Assigning labels, colors or other types of information to submeshes
  • Surface rendering with full color and transparency support
  • Conversion from and to a variety of different file formats that are compatible with standard 3D applications

Contour Segmentation Objects

The Contour Segmentation Objects (CSO) library provides common data structures and modules for automatic generation, processing, rendering, and interactive manipulation of contour sets in voxel images.

This includes:

  • Extraction of iso-contours
  • Manual and semi-automatic contour generation (e.g., LiveWire)
  • Contour smoothing and refinement
  • Conversion from and to other geometric representations

DICOM Support, Image File Formats

MeVisLab provides sophisticated support for working with images stored in the widely used DICOM standard:

  • DICOM import tool automatically selecting and sorting slices that belong to the same 3D/4D image data block
  • Configurable DICOM browser, PACS query, and DICOM storage to PACS
  • Comprehensive access to information stored in the DICOM header of imported images, for example:

    • DICOM coordinate system, image orientation, voxel sizeChannel/time information
    • Patient information
    • Modality specific image acquisition parameters

  • Support for various non-image DICOM objects, including radiotherapy dose plans, registration and landmark objects, structured reports

Besides DICOM, many other file formats are supported, including:

  • TIFF (2D/3D, RGBA)
  • Analyze
  • RAW
  • PNG
  • JPG
  • BMP