scientific article; zbMATH DE number 4000635

From MaRDI portal
Publication:4726786





[For the entire collection see Zbl 0607.00009.] A method is described for computing least-squares bivariate spline approximants to function values specified at the vertices of a very large rectangular mesh. The mesh is considered to be partitioned into a system of (slightly) overlapping rectangular subarrays or blocks, each of which can be held in memory. A constrained tensor-product spline approximant is formed for each block. The constraints imposed are such that the individual approximants blend together with a desired order of continuity. The method is designed to have local properties: modification to function values within a block will affect only the approximant for the block itself (and, possibly, the approximants for immediately neighbouring blocks.) Knot refinement is permitted at block boundaries. Thus the number of knots can be selected locally (i.e. at the block level) to account for local behaviour of the underlying function.











This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4726786)