Hi, Sungjin-

Offhand, I wouldn't expect the "deobliquing" with 3drefit to make any appreciable change to the memory demands of the data. Note the difference between deobliquing with those programs:
+ with 3dWarp, "-deoblique" means to apply the obliquity matrix info to resample the dset and put the brain back into its physical coordinate (i.e., scanner coordinate) location.
+ with 3drefit, "-deoblique" means to purge the obliquity matrix info, so the data is *not* resampled (only header info changes), and from hereon we pretend like the oblique coordinates *were* the scanner ones.

That doesn't sound like "too much" data either, although that measure is necessarily relative to the specific machine that you have. If it were particular high-spatial resolution, that would increase memory demands.

Is that a laptop that it is being run on? Do you have a larger machine for it?

--pt