I use avisynth for frameserving and CCE 2.62 and TMGEnc for encoding. Source video is 480x320 at 23.976, CVD is 352x480. Problem that for one dimension need use downsizing and for other upsizing. Downsizing without prefiltering give aliased noise for sharp edge (step series). I try convert to RGB and use GeneralConvolution filter, but this filter work very slow and after I try increase frame in both dimension and downsize horizantal use HorizontalReduceBy2 filter which use prefilter before downsizing. Result after encoding became better.

video=BilinearResize(video,704,480)# Source has 480x320 resolution
video=HorizontalReduceBy2(video)#after 352x480 resolution

I can not find big difference between BilinearResize and BicubicResize.
I find that important relation for upsizing and best result give rational relation between source and destination (for width 479 and 718)
If any try make same thing let me know.
If any have question I try answer.
With kind regards Yury.