Log in

View Full Version : What is the native color space of a DV clip


kassandro
21st August 2005, 06:45
I am currently developing a dead pixel removal filter for Avisynth. To get the best quality, the removal should be done in the native color space of the DV clip (FourCC = dvsd). Currently I assume that YUY2 is the native color space. Am I right?

communist
21st August 2005, 07:17
No. DV is stored as YV12 (4:2:0 for PAL , 4:1:1 for NTSC).
I'm not sure how much it helps but you can take a look at A. Dittrichs OpenSource DV Codec for possibly more information.
http://forum.doom9.org/showthread.php?t=94458

kassandro
21st August 2005, 10:36
No. DV is stored as YV12 (4:2:0 for PAL , 4:1:1 for NTSC).
I'm not sure how much it helps but you can take a look at A. Dittrichs OpenSource DV Codec for possibly more information.
http://forum.doom9.org/showthread.php?t=94458
Thanks for the information and the link. I will make a look at it. I am surprised that YV12 is selected, because interlaced YV12 is really broken.

trevlac
23rd August 2005, 14:42
I am currently developing a dead pixel removal filter for Avisynth. To get the best quality, the removal should be done in the native color space of the DV clip (FourCC = dvsd). Currently I assume that YUY2 is the native color space. Am I right?

This may be a silly question ... but how are you going to get 'native' info out of a DV clip? :confused:

DV is of course compressed. Something has to de-compress it. The colorspace output of the de-compressor is dependant on what codec is being used on the machine. I believe this all happens before AVS gets the image.

So ... most DV codecs probably deliver RGB. Some may out YUY2. I'd be surprised if they output YV12 ... especially if we are talking about 4:1:1 NTSC.

Bottom line ... you probably have to write your own decoder to get at the 'native' values ... but then again, since these are encoded and quantizised at 8 bit, they are not really perfect either. Actually ... you are pretty far away from what that CCD sees.

BTW: Don't take my rant the wrong way. You probably know a bunch more than me. I just saw your post and thought ... cool idea ... be interesting to see how well it works. :D

scharfis_brain
23rd August 2005, 14:44
I am surprised that YV12 is selected, because interlaced YV12 is really broken.

How do you mean this?


besides A. Dittrichs OpenSource DV Codec, there are alternativel for getting the native YV12 for PAL-DV.

a) use ffdshow's DV-Decoder and select YV12-only Output
then use Directshowsource within AVISynth

b) use traditional AVISource and my ReYV12()
(it probably needs an update due to changed AVS-behaviours...)

henryho_hk
6th September 2005, 16:54
Currently I assume that YUY2 is the native color space.

No, DV is YV12 based. The bad news is that PAL-DV and NTSC-DV use different forms of YV12, while MPEG2-YV12 is the 3rd one. Please refer to the following website for details:

http://www.mir.com/DMG/chroma.html

Wilbert
7th September 2005, 18:13
NTSC-DV doesn't use YV12. It uses a 4:1:1 format.

henryho_hk
8th September 2005, 00:27
NTSC-DV doesn't use YV12. It uses a 4:1:1 format.

Sorry for the wrong information provided. :stupid: :stupid: :stupid:

I am now wondering whether the YV12 of AVISynth, PAL-DV and XViD have the same definition, esp. on the contribution of chroma pixels on the neighbouring real pixels.

------------

Added on 8 Sep. 2005 ~

:stupid: :stupid: :stupid:
I should have read the documentation in:
http://www.avisynth.org/docs/english/advancedtopics/sampling.htm

The next question is how the FFDShow YV12 decoder places the chroma information of PAL-DV, considering the big difference of PAL-DV YV12 and MPEG2 (and AVISynth) YV12.