General resizing question

[droolian01]

Hi there.

can i resize my dvb-t captures from 720x576 down to 480x576 so the encoded xvid can be resized during playback (using AR signalling) to the correct 16:9 ratio?

I am asking this as i am currently resizing with lanczosresize(688,384) and am getting low average quants (i=2.6,p=2.69,b=4.45) so thought i could up the frame size a bit.

What i really want to know is that by shrinking my encode ONLY in the horizontal direction does this leave the vertical direction unchanged? It is after all the vertical resolution which gives the 'detail' (like anamorphic dvds)

I have the feeling that this is a really dumb question but i can't see how yet, maybe someont could enlighten me?

Thanks.

[foxyshadis]

It'd make more sense to leave the horizontal and shrink the vertical. The more you scale one direction the worse it looks, even if the other is full of detail. It starts looking noticeably rough around 2x. For instance, the original requires a 42% expansion, which isn't too bad, but 480 leads to a 113% expansion. You can see how it looks yourself by using bicubicresize(1024,576) on both the original and the shrunk version. (Most renderers would use bilinear, vmr9 and haali's can optionally use bicubic, or you can do it in software with ffdshow and use lanczos.)

You might see how it looks without resizing at all. Otherwise, try a few combinations of shrinking the vertical and/or horizontal and scaling them back up, see how it looks.

[droolian01]

Thanks for your reply.

I think it was a bit of a dumb question after all as its been such a long time since i did any encoding i had forgotten how resizing algorithms work. Even if i leave the vertical resolution unchanged (as in 720x576 to 480x576) it is still altered as an averaging process has occured on 2x2 or 4x4 (or whatever) pixels which means the vertical information is changed.

I think your point about the amount you scale having a much more significant effect is excellent. So for 2 very good reasons my idea of a 480x576 resize is poor.

I made short encodes of the same sequence, one at my original 688x384 res and the other at 480x576 and played them in mediaplayerclassic. Both have to be resized during playback (doesn't everyone watch their video in full screen!) and i could see NO difference at all in the amount of detail present during playback (or freeze framing) - therefore no PRACTICAL difference.

So your point about the amount of change in each dimension is a usefull one to remember, except most people are (unconciously) doing that when resizing by picking the largest (and therefore least shrinkage!) framesize they can.

I knew it was a dumb question after all :o