MPEG-4 Encoding - Resize Before Encode or During Playback?

[thejonesyboy]

I'm converting cropped (720x432 with DAR 2.35:1) anamorphic PAL video to MPEG-4, at a fixed size (2CD's).
I could:

• Resize according to display aspect while maintaining height (1024x432 lanzcos), and then encode, or

• Leave it at 720x432 and then encode (will be higher bits per pixel due to same file size), but then resize during playback.

Between these two options the same amount of data is encoded in each frame, but resizing is done before or afterwards. I was wondering if there was a quality difference, as in the first method we are encoding at a higher res, but lower bits per pixel, and the second a lower res but higher bits per pixel.

I was thinking that the second option could be better due to a relatively higher amount of blocking (if some what indiscernable it's still there) with the first method at the higher res, and a lower amount of blocking in the second method then scaled up to the higher res without the blocking...?

Thoughts and critisism very welcome, as it seems to be the norm to resize beforehand, so I guess I want to see what it's like to go against the grain!

[guns1inger]

Yes, you have the same amount of data per frame, but in one method you spread it over a much larger area. Coupled with resizing up - a quality reducing process in itself), the best option would be to resize during playback.

That said, the usually process is to resize down when encoding DVD to mpeg4, in order to maximise the reduced bitrate over a smaller image size.

[jagabo]

You also need to consider your playback device. Most set-top players will not play MPEG4 files over 720x480. Many will also not repect the PAR flag. If you want the greatest compatability you should reduce the frame size to 720x304 or 640x272.

[thejonesyboy]

Yeah great I was probably gonna go with just a vertical resize to 720x304 anyway, so what would be an advisable downscaling resize filter?

[jagabo]

For downsizing you almost always want to use a Lanczos filter.