bit confused at DVinfo.net

Steve Witt · April 1st, 2005, 12:58 PM

I did a search on this but I am still struggling. When we are talking about "COLOR", what size bit system are we dealing with in Digital Video. 24bit or 8bit or what? The reason I even came across this subject is that I was doing research on GAMMA CURVES. look here to see what I have read.

http://www.siggraph.org/education/ma...n/palette.html

Does this bit value difference show up when comparing 3CCD cameras versus single CCD? Is this color bit value something we can control? Thankyou!!

Andre De Clercq · April 2nd, 2005, 03:18 AM

DV is 8bits/color, means a potential of 16 mio levels of 'gray'. No difference between single and 3ccd. Gamma in videocams is performed using higher bit levels after a first A/D conversion.

Steve Witt · April 3rd, 2005, 05:25 PM

Thanks Andre. What is A/D conversion? Thanks.

Kyle Ringin · April 3rd, 2005, 08:09 PM

Andre,
Wouldn't 8 bit/channel give you 2^8=256 levels of grey?
Not sure what mio means though.

The distinction has to be made between bits/channel and total bits. DV is 24 bits (8 per channel - RGB) IIRC so there is 2^24=16.8 million colours

Steve,
The signal from the camera's CCD is analogue. The camera then runs it through and A/D (analogue to digital) converter so the signal is then digital.
The gamma curve (when speaking in the context of video, DV, etc) is basically a graph of input value to output value and changes the colour response. You can reduce the brightness of highlights, crush blacks, etc.
This is often done in an NLE in post, but some cameras (the more high end ones) can do it in camera.

Someone correct me if I'm wrong.

Hope it helps.

Glenn Chan · April 3rd, 2005, 08:24 PM

Many cameras have processors in them that determine the gamma curve/response of the camera. Cameras can perform this processing at higher bit-depths to avoid banding/rounding error.

Newer cameras tend towards higher bit processing for this.

(Sorry I don't think this makes sense.)

Chris Hurd · April 3rd, 2005, 08:41 PM

Steve:

<< What is A/D conversion? >>

Analog to digital conversion.

Many folks are not aware that a CCD is an analog device which puts out an analog signal. This is converted to a digital bitstream by the A/D converter.

Oops, sorry Kyle, didn't see your previous explanation there. Nevermind the echo please.

Steve Witt · April 3rd, 2005, 08:46 PM

Thanks Kyle and Glenn & Chris. 24bits colour makes more sense to me. I do remember reading a little about the analog to digital conversion in DV cameras but I guess the A/D expression threw me. Are the 24bits always divided equally between the red, green and Blue? Thanks again.

Kyle Ringin · April 3rd, 2005, 08:47 PM

BTW my ramblings above about 24bit, etc were to do with the processing of the video in an NLE. It doesn't take into account 4:2:0 and 4:1:1 colour sampling of DV.
I'd recommend Adam Wilt's site for that:
http://www.adamwilt.com/DV.html

Kyle Ringin · April 3rd, 2005, 09:00 PM

Steve,
When talking about RGB colourspace the bits are divided up equally between the colours. The total number of bits may not be a multiple of three though (eg 16bit, 32bit), due to extra bits (usually 1 or 8) used for the alpha channel.

When processing video in an NLE each video frame is basically decoded into an uncompressed image, processed, then recompressed.
When speaking of the DV compressed video, the colour sampling business comes into it as well.

Steve Witt · April 3rd, 2005, 09:21 PM

thanks for the help Kyle. I'm trying to remember the name of a book I checked out from the library on digital movie making a couple of months ago. At the time it was a little too advanced for me so I took it back and got more basic level books to start off with. Anyway, this book covered this very subject we are discussing here and I believe if I got my hands on this book again, it would now make more sense. I think I am ready to go more indepth on color and sound bits and other subjects related to DV that I considered unimportant at first. thanks again.

Andre De Clercq · April 4th, 2005, 03:59 AM

Kyle, you are right about the 256 gray levels for the luma part in DV. This Y value is a decimated (into 8 bits/sample) version of weighted linear 3x8bits/sample RGB combinations which before decimation, results potentially in 16 million values where most of them (depending on the RGB weighting factors) would result in different gray levels. So, if one would create a B&W picture straight out of an RGB 8bit/sample structure the grayscale would have 16 mio shades. That's why I mentioned a "potential" number of gray shades of a 3x8bit RGB signal...

Steve Witt · April 4th, 2005, 05:40 PM

I found the book that I had once before. It's called "The Filmaker's Handbook (a comprehensive guide for the digital age)"
by Steven Ascher and Edward Pincus. It covers fairly broad, the issues and subjects about DV that I am now trying to learn about.
I will keep this book near my computer for now on and use it like a Dictionary when I get stuck. Thanks.

April 1st, 2005, 12:58 PM	#1
Steve Witt Major Player Join Date: Feb 2005 Location: kentucky, USA Posts: 429	bit confused I did a search on this but I am still struggling. When we are talking about "COLOR", what size bit system are we dealing with in Digital Video. 24bit or 8bit or what? The reason I even came across this subject is that I was doing research on GAMMA CURVES. look here to see what I have read. http://www.siggraph.org/education/ma...n/palette.html Does this bit value difference show up when comparing 3CCD cameras versus single CCD? Is this color bit value something we can control? Thankyou!!

April 3rd, 2005, 08:41 PM	#6
Chris Hurd Obstreperous Rex Join Date: Jan 2001 Location: San Marcos, TX Posts: 27,366 Images: 513	Steve: << What is A/D conversion? >> Analog to digital conversion. Many folks are not aware that a CCD is an analog device which puts out an analog signal. This is converted to a digital bitstream by the A/D converter. Oops, sorry Kyle, didn't see your previous explanation there. Nevermind the echo please. __________________ CH Search DV Info Net \| 20 years of DVi \| ...Tuesday is Soylent Green Day!

April 2nd, 2005, 03:18 AM	#2
Andre De Clercq Major Player Join Date: Sep 2002 Location: Belgium Posts: 804	DV is 8bits/color, means a potential of 16 mio levels of 'gray'. No difference between single and 3ccd. Gamma in videocams is performed using higher bit levels after a first A/D conversion.

April 3rd, 2005, 05:25 PM	#3
Steve Witt Major Player Join Date: Feb 2005 Location: kentucky, USA Posts: 429	Thanks Andre. What is A/D conversion? Thanks.

April 3rd, 2005, 08:09 PM	#4
Kyle Ringin Major Player Join Date: Oct 2004 Location: Melbourne, Australia Posts: 241	Andre, Wouldn't 8 bit/channel give you 2^8=256 levels of grey? Not sure what mio means though. The distinction has to be made between bits/channel and total bits. DV is 24 bits (8 per channel - RGB) IIRC so there is 2^24=16.8 million colours Steve, The signal from the camera's CCD is analogue. The camera then runs it through and A/D (analogue to digital) converter so the signal is then digital. The gamma curve (when speaking in the context of video, DV, etc) is basically a graph of input value to output value and changes the colour response. You can reduce the brightness of highlights, crush blacks, etc. This is often done in an NLE in post, but some cameras (the more high end ones) can do it in camera. Someone correct me if I'm wrong. Hope it helps.

April 3rd, 2005, 08:24 PM	#5
Glenn Chan Inner Circle Join Date: Jun 2003 Location: Toronto, Canada Posts: 4,750	Many cameras have processors in them that determine the gamma curve/response of the camera. Cameras can perform this processing at higher bit-depths to avoid banding/rounding error. Newer cameras tend towards higher bit processing for this. (Sorry I don't think this makes sense.)

April 3rd, 2005, 08:46 PM	#7
Steve Witt Major Player Join Date: Feb 2005 Location: kentucky, USA Posts: 429	Thanks Kyle and Glenn & Chris. 24bits colour makes more sense to me. I do remember reading a little about the analog to digital conversion in DV cameras but I guess the A/D expression threw me. Are the 24bits always divided equally between the red, green and Blue? Thanks again.

April 3rd, 2005, 08:47 PM	#8
Kyle Ringin Major Player Join Date: Oct 2004 Location: Melbourne, Australia Posts: 241	BTW my ramblings above about 24bit, etc were to do with the processing of the video in an NLE. It doesn't take into account 4:2:0 and 4:1:1 colour sampling of DV. I'd recommend Adam Wilt's site for that: http://www.adamwilt.com/DV.html

April 3rd, 2005, 09:00 PM	#9
Kyle Ringin Major Player Join Date: Oct 2004 Location: Melbourne, Australia Posts: 241	Steve, When talking about RGB colourspace the bits are divided up equally between the colours. The total number of bits may not be a multiple of three though (eg 16bit, 32bit), due to extra bits (usually 1 or 8) used for the alpha channel. When processing video in an NLE each video frame is basically decoded into an uncompressed image, processed, then recompressed. When speaking of the DV compressed video, the colour sampling business comes into it as well.

April 3rd, 2005, 09:21 PM	#10
Steve Witt Major Player Join Date: Feb 2005 Location: kentucky, USA Posts: 429	thanks for the help Kyle. I'm trying to remember the name of a book I checked out from the library on digital movie making a couple of months ago. At the time it was a little too advanced for me so I took it back and got more basic level books to start off with. Anyway, this book covered this very subject we are discussing here and I believe if I got my hands on this book again, it would now make more sense. I think I am ready to go more indepth on color and sound bits and other subjects related to DV that I considered unimportant at first. thanks again.

April 4th, 2005, 03:59 AM	#11
Andre De Clercq Major Player Join Date: Sep 2002 Location: Belgium Posts: 804	Kyle, you are right about the 256 gray levels for the luma part in DV. This Y value is a decimated (into 8 bits/sample) version of weighted linear 3x8bits/sample RGB combinations which before decimation, results potentially in 16 million values where most of them (depending on the RGB weighting factors) would result in different gray levels. So, if one would create a B&W picture straight out of an RGB 8bit/sample structure the grayscale would have 16 mio shades. That's why I mentioned a "potential" number of gray shades of a 3x8bit RGB signal...

April 4th, 2005, 05:40 PM	#12
Steve Witt Major Player Join Date: Feb 2005 Location: kentucky, USA Posts: 429	I found the book that I had once before. It's called "The Filmaker's Handbook (a comprehensive guide for the digital age)" by Steven Ascher and Edward Pincus. It covers fairly broad, the issues and subjects about DV that I am now trying to learn about. I will keep this book near my computer for now on and use it like a Dictionary when I get stuck. Thanks.