Constrained connectivity for the processing of very-high-resolution satellite images
Pierre Soille · International Journal of Remote Sensing · 2010
For complex very-high-resolution satellite images, relevant image structures do not all correspond to image regional extrema, but also to plateaus with intermediate intensity values. In this case, dual or even self-dual morphological operations must be applied with care because they assume that the targeted image regions are either brighter or darker than their adjoining regions. A solution to this problem is to consider the image plateaus as the atomic image regions, and then cluster them so as to produce relevant segmentations and edge-preserving simplifications. This paper shows that this can be achieved by considering the concept of constrained connectivity. The problem of transition regions is emphasized, and a solution based on the concept of local extrema mosaic is proposed. An iterative area filter, ensuring that all connected components reach a minimum area threshold, is also described. All transformations are illustrated on sample IKONOS images.