SIAM Journal on Computing
A Metric for Distributions with Applications to Image Databases
ICCV '98 Proceedings of the Sixth International Conference on Computer Vision
Algorithms for the transportation problem in geometric settings
Proceedings of the twenty-third annual ACM-SIAM symposium on Discrete Algorithms
| Hi-index | 0.00 |
We consider the following variant of the well-known Monge-Kantorovich transportation problem. Let S be a set of n point sites in R^d. A bounded set C@?R^d is to be distributed among the sites p@?S such that (i) each p receives a subset C"p of prescribed volume and (ii) the average distance of all points z of C from their respective sites p is minimized. In our model, volume is quantified by a measure @m, and the distance between a site p and a point z is given by a function d"p(z). Under quite liberal technical assumptions on C and on the functions d"p(@?) we show that a solution of minimum total cost can be obtained by intersecting with C the Voronoi diagram of the sites in S, based on the functions d"p(@?) equipped with suitable additive weights. Moreover, this optimum partition is unique, up to sets of measure zero. Unlike the deep analytic methods of classical transportation theory, our proof is based directly on simple geometric arguments.