Idea/question about raw RGGB green cheat

[1024]

Hi all,
first of all i want to thank you for the amout of amazing work you guys put in magic lantern.

I have a question.. is there a posibility to remove information about one of the green values and restore it partially from three  remaining pixels? This would give us more banwidth/resolution/frame rate. Im aware that the restored green value would be bit cheated, but i'm sure that with some sneaky algorithm the results would be quite ok.

What do you think?

[1024]

Is there no one that could at least tell me, if there's a posibility to write such files? .. pretty please

[g3gg0]

Not possible in realtime

[Midphase]

Seems like everyone has some great idea that surely none of the developers could have possibly thought about to get 4K 120fps video on a T2i.

Problem is that as you start reducing data here, get rid of bits there, and so on and so forth, you might as well go back to recording in H.264 and forget about ML.

I have seen plenty of footage coming out of the lower end cameras which has needed to be resized pretty substantially, and IMHO the results aren't much of an improvement over the Canon codec.

We sometimes tend to forget that for many things, the Canon H.264 video looks quite good all things considered (many feature films have been shot on it).

There are two major bottlenecks at work here, CPU speed, and the write speed to the card.

The ML developers have unlocked something quite remarkable, but their discovery has real hardware limitations. My guess is that faster CF cards will render the 5D3 more reliable (not that it's not right now, I can consistently record 1920X1280 footage). Unfortunately faster CF cards will not help 5D2 and 50D users as they have already reached the limits of what their cameras can output.

The idea that reducing data will enable higher resolutions on lower cameras goes against both of those bottlenecks since they are outfitted with slower CPUs and slower SD card writers.