Un-upscaled H.264 output

[laconiccrusader]

My question is maybe stupid, but as i heard the video was recorded at approximatly 1700px on most Canon DSLR like:
550D , 7D, 5DII, 600D etc
then the sensor is doing a realtime upscaling to 1920px.
Canon avoided some aliasing issues on the 5DIII by designing a native 1920 multiple sensor without interpolation,
Realtime interpolation are usually quite bad and leads to a lot of artefacts and loss of texture details, my question now,
is it possible to skip the interpolation process, doing it manually in post with a better algorythm ?
If it's possible, would it be useful? i think it could result in smaller, sharper files that avoid some of the edges artefacts we all know.

[heavygrafix]

if that would be possible, we already had 1DC like h264 output. But ML at current time is not able to interfere with the H264 processing in the Camera.

[feureau]

if that would be possible, we already had 1DC like h264 output. But ML at current time is not able to interfere with the H264 processing in the Camera.

I wonder why that is. Is it just harder to access or nobody has really focused on the h.264 chip or something?

[Audionut]

AFAIK, H.264 is all done on a separate chip.
There is currently no control of this chip.

[2blackbar]

video has to be upscaled because codec supports only resolutions that are divided by 16, thats why it has to be 1920x1088, only way around that is raw video or new compression but that would slow cpu down and wouldnt be worthy to use because h.264 compression is done by hardware.