Pixel replication and subsampling are done on a floating point indexing basis, where the data point index is obtained by truncating the cumulative floating point index. Therefore, when the calculated replication factor or the subsampling step size is non-integer, the resulting integer step size varies during the process.
Magnification Factors Magnification factors for each dimension, Width (wmag), Height (hmag) Depth (dmag), Time (tmag), and Elements (emag) can be specified. Magnification factors can be any value greater than zero.
Center of Action The magnification or reduction is done about a center point that can be specified to be anywhere in the object (even outside of the object). The correct way to think about this is by analogy to a magnifying glass where you place the magnifying glass over the point of interest and then choose the lens to get the magnification you want. The center of action coordinates can be specified individually for each dimension (wc, hc, dc, tc, ec) or, for convenience, the center of action can be automatically computed to be in the middle of all dimensions. This is accomplished by selecting the "center" flag. Automatic centering can be overridden on selected dimensions by explicitly specifying the centers of action for those dimensions.
Output Object Size The resultant output object size can either be the same as the original input object size if Resize dimensions is set to "No" (-resize 0), or it can be resized to fit the magnified or reduced data set if Resize dimensions is set to "Yes" (-resize 1).
If the user chooses not to resize the dimensionality of the output object, and the input object is reduced (magnification factor < 1.0) then unknown pixels are set to zero. Likewise, if the input object is magnified (magnification factor > 1.0) then input data that would be mapped outside of the output object are lost.
Data Type The data type of the output object is the same as the input object's data type. Internally, the data is processed using one of the following: unsigned byte, long, unsigned long, double, or double complex. Data will not be cast to a lower type for processing.
If there is map and value data in the source object, and zero
order interpolation (pixel replication) is used, the map is transferred
to the destination object. If any interpolation algorithm is used,
the value data is mapped before the operation can begin and the
destination will not have map data.
Failure Modes This program fails if the input object lacks both map data and value data.
This program fails if there is mask data and the operation uses anything
other than zero order interpolation. In other words, if value data
is being replicated or subsample on grid, then the mask can be
correspondingly replicated or subsampled. (Bilinear interpolation can not
be performed on mask data.)
If explicit location or time data are available, the sampling algorithm can not resample these data yet, and will fail. This will be implemented later ....