Edge and line detection in multidimensional noisy, imagery data

A machine process for analysis of multidimensional data for detection of edge and line distinctions, comprising the steps of:

• inputting multi-dimensional digital image data of line dimensions m.sub.1 .times.m.sub.2 .times.N.sub.3 ;
• determining denominators of coefficients for fitting a hypersurface to function values in a neighborhood of dimension N.sub.1 .times.N.sub.2 .times.N.sub.3 ;
• computing global parameters for N.sub.1 lines of m.sub.1 lines of data of size m.sub.2 .times.N.sub.3 ;
• examining a first pixel neighborhood of N.sub.1 .times.N.sub.2 .times.N.sub.3 dimension and computing the coefficients of a fitting surface;
• computing for a central pixel of the neighborhood a direction perpendicular to a direction of generally linear discontinuity;
• performing tests of significance on directional derivatives of a fitted surface along a direction perpendicular to the direction of discontinuity to derive edge and line data;
• deriving such edge and line data for central pixels of successive neighborhoods by computing the coefficients of the fitting hypersurface through updating, estimating the directions of edges and lines, and performing the tests of significance on the directional derivatives;
• repeating the process for all sequential N.sub.1 lines of said m.sub.1 lines of data by updating the global parameters;
• and
• outputting said edge and line data for visual display.