We consider the problem of approximating a function on a d-dimensional voxel grid by a unimodal function to minimize the L2 approximation error with respect to a given measure distribution on the grid. The output unimodal function gives a layered structure on the voxel grid, and we give efficient algorithms for computing the optimal approximation under a reasonable assumption on the shape of each horizontal layer. Our main technique is a dominating cut algorithm for a graph.
|Number of pages||11|
|Journal||Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|
|Publication status||Published - 2004 Dec 1|
ASJC Scopus subject areas
- Theoretical Computer Science
- Computer Science(all)