RED code, RAW, lossless, lossy ??

[dchenevert]

I don't quite get it.

If you consider 3 data format options:

(a) all the data from the sensor (12 bits X fps X pixel-count)

(b) "visually lossless" but in reality a little lossy

(c) literally lossless, compressed to lower the disk usage but
without a single bit missing from (a)


Then...

(a) is called "UNCOMPRESSED RAW" or "RAW"
(b) is called "REDCODE RAW" or "REDCODE"
(c) does not exist yet (i.e. has not been actually announced)

Is this correct?

[andersh]

Do you mean something like zip/gzip? No, I don't think it has been explicitly mentioned. But I guess it could be embedded in the RAW spec. Based on a simple test on a half gig file I had it seems it could give a 1:2 compression ratio. Also, I don't think inter-frame compression has been mentioned. When filming high speed it would probably give high compression ratios.

[dchenevert]

andersh,

Right, something like gzip. Compressed internal to the camera, not
dropping any bits, then going out to the not-quite-refridgerator-sized-disk-array.

Compression might be better than gzip because of interframe
and interline self-similarity.

I infer that RED expects everyone will prefer (b) over (a), but
for the most conservative users, I wonder if (c) would be
of value.

Also, if a RED team member sees this, you might someday
want to clear up the ambiguity in the word "RAW". On your website,
RAW has two different uses on the "REDCINE Overview Chart",
but I'm guessing just one use (not clear which one) on the
"RED ONE FORMAT OPTIONS" page.

[stokestack]

(b) "visually lossless" but in reality a little lossy

(b) is called "REDCODE RAW" or "REDCODE"

Is this correct?

I don't think so. It's just Redcode, using wavelet compression. I don't think the term "raw" is used in conjunction with it.

It seems that raw camera data is often losslessly compressed, although I don't know exactly what kind of compression is used. Maybe it uses correlation between adjacent photosites. If you look at the size of a raw picture from a Canon SLR, for example, the file size is way smaller than you would expect of an uncompressed bitmap of equal size. Then again, with a single-chip camera, there aren't as many red, green, and blue photosites as there are pixels in the final image, so some savings are realized there.

[Anders Holck]

There is both REDCODE RAW and REDCODE.

REDCODE RAW
When shooting REDCODE RAW it's automatically 4k. It is lossy compression using VBR wavelets, at about 1:10. The data is the "raw output" of the CMOS so it's one plane bayer encoded (not RGB), 10 bit log at 4k. You need to process the data in REDCINE to get something usefull out of it.

REDCODE
Is a RGB codec you can shoot in 2k, 1080p or 720p. It's still using VBR wavelets but it will be usable without REDCINE

Lossless compression is standard in most DSLR RAW formats.
Efficiency is normally around 1:1.5, so not that usefull to bring the datarate down, compared to the CPU used.
Most codecs use Run-lenght encoding, Entropy coding or LZW

[dchenevert]

ah I stand corrected...

(a) all the data from the sensor -> UNCOMPRESSED RAW

(b) visually lossless (but, technically, "lossy") -> REDCODE RAW

(c) literally lossless compression of (a) -> no announced support yet

(d) RGB codec, different from (b) -> REDCODE


right?

[visionmind]

According to http://red.com/download/redcine.jpg it seems you're right.

In my mind I don't why they couldn't losslessly compress the UNCOMPRESSED RAW. Losslessly compressing an uncompressed image CAN save a lot of space. Perhaps it requires too much processing power.

visionmind