[Resolved] ML RAW - 14 bits by color channel or 14 bits at the totally ?

Started by FLoR1aN, October 17, 2016, 08:28:02 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

FLoR1aN

There is something I cannot figure out.

When we say ML can film in RAW 14 bits:
•         It's 14 bits by color channel
•         It's uncompressed
Right?

But when I look the files of my rushes, I'm disturbed by the file size.

For example, with a rush of 10.7s in 1824x764 / 25fps (268 frames):
The MLV (without sound) file size is around 623MB.
Thus, it gives
•         623 / 268 = 2.32MB by frame
•         2.32 / 1824 / 764 x 1024 x 1024 = 1.75 Byte by pixel by frame
•         1.75 x 8 = 14 bits exactly

Thus, I obtain at the totally 14 bits by pixel by frame, and not 14 bits by color channel as excepted...

Moreover, when extracted with raw2cDNG, the dng are always exactly the same size (in function of the output 12bits or 16bits).
With MLrawviewer (output 14bits), the size is different for each DNG, but I read in the soft specs that it includes lossless compression (approximately 40% of compression).
Thus, I think I can be sure that the MLV is not originally compressed.

So, I see 3 solutions:
•         The color depth is 14 bits at the totally (4 or 5 bits by color channel)
•         The MLV image is lossless compressed at exactly 33%, but like I said, I seriously doubt of it...
•         I made a mistake in my calculation somewhere but I can't see where


I'm a little confused...

Walter Schulz

You have to keep in mind all pixels are masked by a filter (CFA, Bayer-Matrix).
gbgbgbgb...
rgrgrgrg...

1/4 capture red
1/4 capture blue
1/2 capture green


FLoR1aN

Ok, that's explain a lot !

I knew a bit on Bayer Matrix, but I didn't think about it for my problem.

Thank you !