Support for N-dimensional processing

Helical filtering Claerbout (1997a) opens new possibilities for multidimensional filtering applications. In particular, it becomes possible to write completely dimension-independent programs for deconvolution, missing data interpolation, least-square signal and noise separation, etc. Claerbout (1997b). Of course, going beyond the traditional one, two, and three dimensions has not only the theoretical value. Prestack seismic data are naturally five-dimensional (with a ``3-D'' acquisition), and additonal dimensions (depth, velocity, etc.) are often added by processing. Besides, dimension-independent programming avoids a large amount of an error-prone code duplication, required for going from one to two to three dimensions.

The sep module provides some additional functionality to support N-dimensional programming. The function sep_dimension  takes the file tag (defaulted to "in" - the standard input) and returns the effective number of dimensions in the SEPlib file. Additionaly, from_par, from_history, to_history, and other parameter handling routines accept not only scalar arguments, but also arrays of values. If iarr is an array of integer values, then

call from_par ("arr",iarr)

whill fill iarr with parameter values 1, 5, 3, when the command line (or a parameter file) has either a comma-separated list arr=1,5,3 or distinct values arr1=1, arr2=5, arr3=3. This convention can be handy for reading all the dimensions from a file:

allocate (n (sep_dimension ()))
call from_history ("n",n)

It also works with to_history: short arrays are written to the history file as comma-separated lists, while long arrays are expanded as distinct parameters.

A suite of N-dimensional programs (using the sep module) accompany the current revision of Claerbout's Geophysical Estimation Mapping .