Color Image Segmentation
Color Structure Code (CSC)
The CSC is an inherently parallel hierarchical color segmentation method that can operate on distributed data (Figure 1) on an optional number of processors.
Figure 1: An image on a 3x3 processor grid
The CSC operates on a hierarchical overlapping hexagonal topology (Figure 2).
Figure 2: The hexagonal topology
This leads to inherently parallel algorithms and combines the advantages of local region growing (simplicity and quickness) and global techniques (robustness and accuracy).
Figure 3: Example of an CSC-segmented scene, where all found objects (trees in the CSC data structure) are presented by the outer border and the involved pixels in the mean color of the object.
Further segmentation examples (CSC in comparison with other traditional color segmentation methods). If you want to find out more about these techniques read this document (only available in German): .ps (3.5MB) .ps.gz (180KB).
The tools for object extraction from the CSC data forest are problem dependent. Generally, primary features are color and geometrical shapes of the outer border, secondary features are their topological relations. xyys