Accession Number : ADA288996

Title :   FBm InterpoIator.

Descriptive Note : Final rept.,

Corporate Author : TULANE UNIV NEW ORLEANS LA

Personal Author(s) : Martinez, Andrew B.

PDF Url : ADA288996

Report Date : 04 NOV 1994

Pagination or Media Count : 16

Abstract : A technique for interpolating an irregularly sampled multidimensional field is presented. The method assumes that the field to be interpolated is a fractional Brownian motion surface with additive measurement error. There are several advantages to this method: data sets with different mean-squared measurement errors can be optimally merged, and the variance of the estimator can be easily computed at each interpolation point. An iterative algorithm is given for implementing the estimator with the advantage of allowing a stopping rule to be used to terminate the iteration after a predetermined level of performance is achieved. Furthermore, at each stage, multiple points can be tested, the resulting performance evaluated, and that point which most decreases the variance of the estimator selected. Using the iterative algorithm produces a low order estimator based on a small, carefully selected set of points with a reasonable level of performance.

Descriptors :   *INTERPOLATION, *FIELD THEORY(ALGEBRA), *BROWNIAN MOTION, DATA BASES, ALGORITHMS, MEASUREMENT, MAXIMUM LIKELIHOOD ESTIMATION, ESTIMATES, VARIATIONS, ERRORS, SAMPLING, ADDITIVES, ITERATIONS, FITTING FUNCTIONS(MATHEMATICS).

Subject Categories : Theoretical Mathematics

Distribution Statement : APPROVED FOR PUBLIC RELEASE