question about 32-bit color

[compusic]

by searching this forum, can't find the answer.

It seems "32-bit color depth" is used to refer to different color depth, very confusing. can someone make it clear?

well, I just give some examples I have encountered.

for crt or lcd display, 32-bit means 24-bit RGB and the extra 8-bit is for transparence, right?

and some softwares process video data in 10-bit resolution for each channel, and this is also called by someone as 32-bit color, though only 30-bit is used, and the eatra 2-bit may or may not be used for transparence.

some file format allows 32-bit color for each channel, and the total depth without transparence is 96-bit.

but some NLE softwares have 32-bit color option for export or render, and don't tell whether it's 10-bit or 8-bit for each channel, such as Premiere.

[HyperHacker]

What are you asking? 32-bit colour is generally RGBA (or backward); that is, 8 bits of red, green, blue and alpha (translucency) respectively.

I like the sound of that 10-bit mode. It's not like my monitor is doing anything with the alpha channel.

[KoD]

"and some softwares process video data in 10-bit resolution for each channel"

Are you sure they're not using some nonlinear color representation like Cineon 10 bit ?

[Mug Funky]

there's a bit of confusion between bits-per-channel and bits-per-pixel.

32 bit RGBA is typically 8 each for red, green, blue, alpha. wheras 32 bit per channel is for HDR images and (usually) is a IEEE floating point number for each colour channel (making a total of 12 bytes per pixel).

10-bit depth is a broadcast thing. SDI (serial digital interface - a coaxial digital video standard used in pro gear) carries 10 bit per channel 4:2:2 YUV, and dual-link SDI (ie 2 cables) can carry 4:4:4 RGB or YUV in 10 bits.

"cineon" is a slightly different beast. it originally was the name of Kodak's scanning/grading/laser film out system, but the .cin files produced by the scanner became known as cineon files, and by association the newer, standardised Digital Picture Exchange (dpx) format is usually also called cineon. it is 10 bit-per-channel RGB, but the values are in a logarithmic scale that (approximately) maps to actual film density. the data can be sent through SDI and stored on tape, but will look dull and low-contrast on account of log data being shown in a linear scale. a simple lookup table corrects this of course. the good thing about log scale for film density is you can fit the majority of a correctly exposed film in the greatest range of numbers, and you can also preserve a much larger dynamic range - so overexposed film can still be saved to a large extent. everything that hit the film is represented, hence the term "digital negative".

maybe one day in the future we'll be watching everything in 32bit float, but for now 10 bits is king (16 bit linear may be the next thing, but currently film recorders don't handle linear as well as log).

@ HyperHacker:

the 3x 10 bit channels plus 2 bits padding thing is in a particular subset of the dpx format. usually it's not padded though.