The HIPS images are stored frame by frame while the LVQ datafiles are stored vector by vector (the elements in the vectors are pixelvalues from corresponding pixel positions in the frames). This implies that the program must build the LVQ datafile column by column as the frames are read in. For large images, this means that it needs a lot of disk space temporarily.
Alternately, the program can take decimated (as created by
qmf -d ) files as input (by using either of the filter
structure parameters - -s -sn -sb).
Then the first input frame is assumed to be in
decimated format, while the rest are assumed to be full rate frames.
This operation is much faster than the full rate operation.