Increased picture quality with increased framesize (during transcode)..?

[Gew]

Hiya!

Just a bit curious.

I have this cheapshot analogue USB2 tv-tuner adapter which I use to record stuff from cable tv, preserve old VHS/Hi8-tapes of personal value, etc. However, the tv-recording is limited to MPEG2(8Mbits) w/ VCD-like resolution (352x288). No 3rd-party software works, so I'm stuck with this, no work-arounds on the software-part, let's continue..

I want to know if it's my imagination, that I get better end-result quality (when re-authoring to ~700kbps MPEG4) by coding with resize.

My imagination(?) tells me that I get better video by doing a resize i.e. from 352x288.mpg to 512x384.avi on transcode, than simply encoding it down to a Xvid of original size (352x288). Isn't it somewhat a main rule? That you can never achieve better (frame-size) results than the first video was written in? But on the other hands, isn't MPEG2 sort of complex way of encoding? Drawn by the fact that a tiny 352x288 MPEG2 may appear almost sharp when played in fullscreen mode, whereas i.e. Xvid often needs higher frame sizes. Does this have something to do with packed bitstream? Interpolation?

Also, when doing the resize in VDB, there are a couple of way to do it. The only method that won't give me pleasing results is the defualt one, "Nearest neighbour". Resizing with this gives me glitches during scene movement on playback, wack! The other methods (Bicubic, Bilinar, Precis belinar, Lanczos) all works fine. However, could someone tell me witch of these other methods I should prefer? I'm pretty novice and my eye can honestly not tell any difference in quality them between, so I could use an expert's opinion here. Stick to the old Lanczos, huh?

So, final wrap-up of my questions:

- Is resizing MPEG2 i.e 352x288 to Xvid 512x384 just a total bitrate waste? Then please just tell me so.
- Which resize method is to prefer, really?

[jagabo]

Frame size vs bitrate is a compromise. The smaller the frame the less bitrate it takes to keep the image from breaking up into macroblocks. But the smaller the frame size the blurrier the video will look when displayed full screen. Here are some severe macroblocks from too little bitrate:



Generally, you don't gain much by enlarging the frame before encoding. But you might find that a Lanczos resize delivers sharper results than whatever you graphics card does (probably bilinear or bicubic).