DVCPRO HD - Intra DC Precision question at DVinfo.net

John Jay · August 5th, 2005, 03:57 PM

Can someone say how many bits are allocated to the coefficients of the Discreet Cosine Transform in the in-camera compression scheme for DVCPRO HD ?

My question is related to the use of 35mm adapters whereby a high bit count can improve very fine shadings (Bokeh) in the encoded video

John Jay · March 8th, 2006, 04:41 PM

Anyone know for sure?

Chris Hurd · March 8th, 2006, 04:44 PM

Wow, over half a year and no responses? We've got to do something about that.

John Jay · March 8th, 2006, 04:51 PM

I'm sure you'll nail it Chris :)

A. J. deLange · March 9th, 2006, 03:55 PM

This is probably not a very satisfying answer but it is determined adaptively in order to make the output bitstream fit into the available bandwidth. The dc (0,0) coefficient is allocated more bits than the ac terms and the number of bits decreases as the frequency increases.

John Jay · March 10th, 2006, 10:37 AM

AJ; I read an article which says that in M-JPEG the DC coefficient of each macroblock (the average luminance of the macroblock) is defined with only eight bits of precision while with MPEG-2, the DC precision is defined to be 10 bits. This results in a perceivable image improvement with certain type of pictures.

more here >> http://www.broadcastpapers.com/editi...0-%20print.htm

I am presuming M-JPEG is the scheme used in DVCPRO-HD ?

So with MPEG-2 the edges of the 8x8 pixel blocks are that much smoother and more appropriate for images where subtle gradients are present and therefore less prominent

I am surprised Panasonic's white paper on DVCPRO HD says very little in this regard

Bob Grant · March 10th, 2006, 04:26 PM

There's another issue which is what I thought you were originally on about.
The effeciency of the encoder. Even though the data rates maybe 25 or 100Mbits/sec for DV25 or DVCPro100 not all of those bits are necesaarily used to carry useful data.
The encoder may simply run out of time to encode a frame due to it's complexity and therefore just pad the data. The article that I read was pretty old so maybe encoders have improved but back then measured actual data could run as low as 30% of available bandwidth.

August 5th, 2005, 03:57 PM	#1
John Jay Major Player Join Date: Sep 2002 Location: US & THEM Posts: 827	DVCPRO HD - Intra DC Precision question Can someone say how many bits are allocated to the coefficients of the Discreet Cosine Transform in the in-camera compression scheme for DVCPRO HD ? My question is related to the use of 35mm adapters whereby a high bit count can improve very fine shadings (Bokeh) in the encoded video __________________ John Jay Beware *PLUGGER-BYTES*

March 8th, 2006, 04:41 PM	#2
John Jay Major Player Join Date: Sep 2002 Location: US & THEM Posts: 827	Anyone know for sure? __________________ John Jay Beware *PLUGGER-BYTES*

March 8th, 2006, 04:44 PM	#3
Chris Hurd Obstreperous Rex Join Date: Jan 2001 Location: San Marcos, TX Posts: 27,366 Images: 513	Wow, over half a year and no responses? We've got to do something about that. __________________ CH Search DV Info Net \| 20 years of DVi \| ...Tuesday is Soylent Green Day!

March 8th, 2006, 04:51 PM	#4
John Jay Major Player Join Date: Sep 2002 Location: US & THEM Posts: 827	I'm sure you'll nail it Chris :) __________________ John Jay Beware *PLUGGER-BYTES*

March 10th, 2006, 10:37 AM	#6
John Jay Major Player Join Date: Sep 2002 Location: US & THEM Posts: 827	AJ; I read an article which says that in M-JPEG the DC coefficient of each macroblock (the average luminance of the macroblock) is defined with only eight bits of precision while with MPEG-2, the DC precision is defined to be 10 bits. This results in a perceivable image improvement with certain type of pictures. more here >> http://www.broadcastpapers.com/editi...0-%20print.htm I am presuming M-JPEG is the scheme used in DVCPRO-HD ? So with MPEG-2 the edges of the 8x8 pixel blocks are that much smoother and more appropriate for images where subtle gradients are present and therefore less prominent I am surprised Panasonic's white paper on DVCPRO HD says very little in this regard __________________ John Jay Beware *PLUGGER-BYTES*

March 9th, 2006, 03:55 PM	#5
A. J. deLange Major Player Join Date: Jun 2004 Location: McLean, VA United States Posts: 749	This is probably not a very satisfying answer but it is determined adaptively in order to make the output bitstream fit into the available bandwidth. The dc (0,0) coefficient is allocated more bits than the ac terms and the number of bits decreases as the frequency increases.

March 10th, 2006, 04:26 PM	#7
Bob Grant Trustee Join Date: Nov 2005 Location: Sydney Australia Posts: 1,570	There's another issue which is what I thought you were originally on about. The effeciency of the encoder. Even though the data rates maybe 25 or 100Mbits/sec for DV25 or DVCPro100 not all of those bits are necesaarily used to carry useful data. The encoder may simply run out of time to encode a frame due to it's complexity and therefore just pad the data. The article that I read was pretty old so maybe encoders have improved but back then measured actual data could run as low as 30% of available bandwidth.