[NUUG video] Resolution on compressed video (384x288 vs 360x288)?

Herman Robak herman at skolelinux.no
Sun Sep 21 17:10:21 CEST 2008


On Sun, 21 Sep 2008 11:06:45 +0200, Petter Reinholdtsen <pere at hungry.com> wrote:

> At the moment, the resolution on the DV recordings are 720x576.  The
> compressed MPEG movies are 384x288.  Why is the ratio on the moves
> changes slightly?  If the idea was to half the height and width, it
> would be 360x288.

384/288 = 4/3

Since the aspect ratio is supposed to be 4/3, this means 384x288 gets square (well, almost...)* pixels.  Square pixels can be displayed without stretching on a computer screen. 360x288 would need to be scaled _up_ to 384x288, or the image would be too narrow; squashed and noticably distorted.


> The background for my question is the coding to Ogg Theora.
> Compressing from 720x576 to 360x288 is quicker, and I can only assume
> lead to less rounding errors. 

Only if the scaler cheats to work faster or to achieve higher resolution than the Nyquist theorem would permit.  And that gain lost again if you want correct display dimensions on a computer screen.


> This made me wonder why we compress to 384x288 for MPEG.  
> Anyone know why that is a good resolution to use?

If you are recoding from the high quality 720x576 video, this is the least lossy stage where you can resize the pixels to a sqare shape.  And most target devices have square pixels now.

- Naive players may fail to honour the aspect ratio header.
- Naive or misconfigured encoders may fail to set the aspect ratio header properly.
- Some formats don't even support aspect ratio metadata.
- Only players with the same native pixel aspect ratio (analog TV sets) will benefit from a preserved pixel aspect ratio (non-square pixels, that is), and only in the best cases.


*) Due to overscan, it's 392x288, unless you crop to 704x576 before scaling.
http://lipas.uwasa.fi/~f76998/video/conversion/

-- 
Herman Robak


More information about the video mailing list