Hierarchical Image Segmentation Software

The HSEG software is an image analysis program that analyzes single band, multispectral, or hyperspectral image data. It can process any image with a resolution up to 8,000 x 8,000 pixels. The implementation of the software analyzes each pixel, groups the pixels that are have similar characteristics to form regions, and then combines regions based on their similarity, whether adjacent or disjointed. This grouping creates spatially disjoint regions. The greater the number of regions located, the higher the segmentation resolution, however some regions may be split into multiple regions when viewing data at its finest levels; the highest resolution is not always ideal.

Original RGB Imagery Data

Highest Number of Regions

Medium Number of Regions

Least Number of Regions

Least Number of Regions

Available along with the HSEG software is a viewing tool called HSEGViewer. Using this tool, the user can select from several levels of resolution to obtain the best results for a specific application. In the above example for and Landsat Thematic Mapper (TM) image from over downtown Washington, DC, HSEGViewer was used to perform a customize selection of regions.

The benefits of such a software application are:

  • Provision of finer resolution or detail than typical image-analysis tools
  • Maintains boundaries between image regions at full resolution at all levels of detail
  • Includes HSEGViewer, a companion visualization and segmentation selection tool that can be used to highlight and select data points from particular regions
  • Provision of convenient image representation for a variety of data-mining and knowledge discovery applications.

The applications for this software include:

  • Identifying population densities and areas with greatest expansion
  • Monitoring agricultural crops
  • Improving scan analyses for CAT scans, MRIs, ultrasound medical equipment