A question over (possible) ML Raw video compressions

Started by gyf304, August 12, 2013, 06:24:33 AM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

gyf304

Well, one of my friends in 10th grade was working on an voice synthesis project and he developed something interesting: he first differentiated an audio(voice) file and then compressed it with deflate codes. The result is pretty good, he got a 70% reduction on the file size.
My questions are:
1. Is ML currently using any compression?
2. Since video is kind of a continuous format, can we just differentiate it over time(calculate the difference in RGB of the same pixel in neighboring frames) and then deflate it? Sound like it will reduce the file size as well as bandwidth requirement on the card, esp in some stationary scenes.

I'm just a high school student who happens to do a little programming, so point out if there's any thing wrong about my idea.

Greetings from China.

jameschens

Your idea is good, and this is a commonly asked question.  But any form of compression will require CPU cycles and unfortunately, the developers have stated that there are none to spare while copying the raw files from memory to card.

They stated it directly in the original RAW thread, though I can't find it atm.  But they've already marked other threads of this topic as impossible:

http://www.magiclantern.fm/forum/index.php?topic=7095.msg59225#msg59225
http://www.magiclantern.fm/forum/index.php?topic=7447.msg63508#msg63508

g3gg0

Ehat you described is one of the earliest compression methods available.
Its called ADPCM and is the same concept as the P-frame in h264: you just transmit differences
Help us with datasheets - Help us with register dumps
magic lantern: 1Magic9991E1eWbGvrsx186GovYCXFbppY, server expenses: [email protected]
ONLY donate for things we have done, not for things you expect!

gyf304

Quote from: g3gg0 on August 12, 2013, 08:53:55 AM
Ehat you described is one of the earliest compression methods available.
Its called ADPCM and is the same concept as the P-frame in h264: you just transmit differences
So is that possible? As it doesn't seem to be very intensive.

g3gg0

Quote from: jameschens on August 12, 2013, 07:56:05 AMBut any form of compression will require CPU cycles and unfortunately, the developers have stated that there are none to spare while copying the raw files from memory to card.

They stated it directly in the original RAW thread, though I can't find it atm.  ...

this!
Help us with datasheets - Help us with register dumps
magic lantern: 1Magic9991E1eWbGvrsx186GovYCXFbppY, server expenses: [email protected]
ONLY donate for things we have done, not for things you expect!