pixel replication is straightforward. 

linear interpolation: by default when doubling the resolution this option will want to increase the resolution by 1 less than 2X the resolution. The reason for this is seen in the diagram below. Increasing from 4 to 7 points allows the interpolated points to be equ-spaced between existing points.

No matter what resolution is chosen, mrewave will begin interpolation centered at the 1st point and finish centered on the last point. So if you were to interpolate from 4 to 8 points the resulting grid would look like:

Fourier interpolation: Since phase rolls in an image cause shifts in its transform, mrewave needs to know if data set is chopped. For real images always choose "smooth". For complex data the following example illustrates the choice between 'smooth' and 'chopped'. Consider selecting 'chopped' in two different  cases.

the original magnitude image

The data is not chopped as seen in the phase image

After the inverse ft, the data is shifted out to the edges of k space.

Zero padding

After the final forward ft a high frequency oscillation is seen in the image

The data is chopped as seen in the phase image

After the inverse ft, the data is nicely centered in k space

Zero padding

After the final forward ft the same motion artifacts as seen in the original image are still present.