ParbatParbat

Parbat is a research prototype for remote sensing image processing and visualisation. The software tool was part of my PhD research at ITC in the Netherlands. Parbat incorporates novel techniques for classification and specifically segmentation of remotely sensed imagery. Visualisation of 3D feature space dynamically linked with imagery is one of the unique aspects of Parbat. A fuzzy visual classifier allows for visual classification of a remotely sensed image. There are no help files or tutorials for this software. It is meant to be used "as is". My PhD thesis explains the techniques implemented in Parbat in more detail. See screenshots here. Download Parbat here.

 

Features:

  • Single band and colour composite display, see screenshots
    • Scroll, Image and Zoom window
    • Zoom factor slider
    • Query window with coordinates and values
    • ENVI Regions of Interest display
  • ENVI file handling, import and export for all file types (band sequential binary image with ASCII header)
  • ENVI region of interest (ROI) ASCII format for display and classification/segmentation
  • 3D feature space plot linked with image display
    • rotation, zoom, translation, auto-rotation
    • pixel selection with auto-zoom and highlight in image display
  • Class visualization in 3D feature space plot (chapters 2 and 3 of my thesis), see screenshots
  • Image Segmentation, see screenshots
    • Unsupervised Split-and-merge segmentation based on mean and (co)variance (multivariate) (chapter 4 of my thesis)
    • Unsupervised texture-based split-and-merge segmentation based on the Local Binary Pattern (LBP) and local variance texture measure (univariate) (chapter 5 of my thesis)
    • Supervised texture-based split-and-merge segmentation based on the Local Binary Pattern (LBP) and local variance texture measure (univariate) (chapter 5 of my thesis)
    • Supervised texture-based split-and-merge segmentation based on the Local Binary Pattern (LBP) and RGB 3D histogram texture measure (multivariate) (chapter 6 of my thesis)
    • Supervised texture-based split-and-merge segmentation based on the Local Binary Pattern (LBP) and local variance texture measure at multiple spatial scale levels (chapter 7 of my thesis)
    • Unsupervised multivariate region growing segmentation (van der Werff and Lucieer, 2004)
  • Image Classification, see screenshots
    • Supervised Fuzzy c-means with Euclidean and Mahalanobis distance metric
    • Visual fuzzy classifier (chapter 2 of my thesis)
    • Isosurfaces for interactive visualisation of uncertainty in fuzzy classification results (chapter 8 of my thesis)
  • Extra Tools
    • Cooccurrence Texture image
    • Local Binary Pattern (LBP) texture image, see screenshots
  • Parallel Coordinate Plot (PCP). The PCP is a separate Java tool called Parvis (Hauser, 2000)
  • 3D Terrain visualizer (ArcGrid ASCII export format), see screenshots

 

 

Categories
Links