Algorithms in question are mostly those which are more easily implemented for raster than vector data, like
methods that use cost-distances or involve neighborhood operations. For the implementation of a local transformation,
the minimum bounding rectangle of the desired polygons or area of interest is calculated. A margin is added to avoid
edge effects. After applying the generalization algorithms, data is re-converted to the vector model. A major drawback
of this method is that the computational effort might be considerable for large datasets and/or smaller portions where methods involving local conversion are applied. Furthermore it should be pointed out that, in principle, the same problems
and restrictions as noted for global transformation apply to local transformations as well. But since the amount of
data will be rather small in most cases and all relevant parameters can be selected for a particular local situation, loss of
information and positional precision can be better controlled.