R3D compression ratio question

[jussi rovanpera]

Hi,

What is the reference 1:1 "uncompressed" frame that the compression ratio refers to, is it 16bit RGB?

 

[Blair S. Paulsen]

I don't believe that the uncompressed data set is RGB, so it cannot be directly compared with any RGB format. AFAIK, the compression that happens when writing the data to the onboard mags is before decoding (de-bayering, demosaicing, decrypting, RGB re-construction via algorithm, etc). If you're really curious, I suggest you read up on entropy codecs like JPEG2000 that use wavelets.

Cheers - #19

 

[jussi rovanpera]

If I calculate from the ratios and the data rates, uncompressed 6K would be 30 MB per frame, so the reference is a monochromatic 12 bit file?

 

[jussi rovanpera]

If I calculate from the ratios and the data rates, uncompressed 6K would be 30 MB per frame, so the reference is a monochromatic 12 bit file?

Uncompressed = pixels on frame x 12bit? Can someone verify this?

 

[Terry VerHaar]

Read what Blair said as I think he has explained correctly that you can't make the comparison directly. The compression is applied to data; captured directly from the sensor and before any processing whatsoever.

One tidbit of information that plays into this is the fact that REDcode is mathematically lossless at about 2.5:1; meaning, of course, that the original signs; can be flawlessly recreated from the data that has a 40% footprint of the original, captured data.

My guess is that the data related to a single frame of be-bayered video would be much greater than the data stored in a single frame of REDcode; probably by at least 2.5 times.

 

[Mark Kern]

If you're trying to figure out the bit depth, that's not going to work. That said, a 16 stop dynamic range on a Dragon is going to require over 12 bits...16 probably, depending on how things are encoded.

 

[jussi rovanpera]

Read what Blair said as I think he has explained correctly that you can't make the comparison directly. The compression is applied to data; captured directly from the sensor and before any processing whatsoever.

One tidbit of information that plays into this is the fact that REDcode is mathematically lossless at about 2.5:1; meaning, of course, that the original signs; can be flawlessly recreated from the data that has a 40% footprint of the original, captured data.

My guess is that the data related to a single frame of be-bayered video would be much greater than the data stored in a single frame of REDcode; probably by at least 2.5 times.

Thats what I said on my later post, monochromatic 12 bit, not debayered RGB.
Lossless is said to be 2:1-3:1, so uncompressed is different from lossless. So what is 1:1?

 

[jussi rovanpera]

If you're trying to figure out the bit depth, that's not going to work. That said, a 16 stop dynamic range on a Dragon is going to require over 12 bits...16 probably, depending on how things are encoded.

I'm not trying to figure out the bit depth, I would just have asked for it. 8:1 datarate is 1/8 of the uncompressed data rate, and I know the size of the sensor, so in my calculations it equalled monochromatic 12bits.

 

[Sabovic Adis]

6144x3072(pix) x 16(bit) x 24(fr) = 864 MB per sec. @1:1;
36 MB per frame @1:1.

288 MB per sec. @3:1;
12 MB per frame @3:1.

That's what I calc.
After debayering it gets RGB...right?
:biggrin: