The routine matches the pixel-index bounds of a number of NDFs so that their array components may be compared pixel-for-pixel during subsequent processing. Matching is performed by selecting an appropriate section from each NDF, the method used to define this section being determined by the value given for the OPTION argument.

CALL NDF_MBNDN( OPTION, N, NDFS, STATUS )

OPTION = CHARACTER $\ast$ ( $\ast$ ) (Given)

N = INTEGER (Given)

NDFS( N ) = INTEGER (Given and Returned)

STATUS = INTEGER (Given and Returned)

• If OPTION=’PAD’ is specified, then the NDF bounds will be matched by "padding"; i.e. each NDF will be extended by selecting the smallest section from it which encompasses all the pixels in all the NDFs. In effect, the pixel-index bounds of the NDFs are "maximised" and the "union" of all N sets of pixels is selected. Any new pixels introduced into an NDF will be padded with the "bad" value. If the NDFs have different numbers of dimensions, then the dimensionality of all the returned sections will match the NDF with the highest dimensionality.

• If OPTION=’TRIM’ is specified, then the NDF bounds will be matched by "trimming"; i.e. each NDF will be restricted in extent by selecting a section from it which encompasses only those pixels which are present in all the NDFs. In effect, the pixel-index bounds of the NDFs are "minimised" and the "intersection" of all N sets of pixels is selected. An error will result if the NDFs have no pixels in common. If the NDFs have different numbers of dimensions, then the dimensionality of all the returned sections will match the NDF with the lowest dimensionality.

• Note that the initial NDF identifier values will be annulled by this routine and replaced with identifiers describing appropriate new sections from the original NDFs. If access to the original data is still required, then the initial identifiers may be cloned with the routine NDF_CLONE before calling this routine.