video sample size is 12 bit

[noobsaibot]

Hello folks, in my first attempt to encode a video using xvid I got a video which has a "sample size" of 12 bit as seen under file->properties in windows.

I am encoding a fullscreen video, and I have this doubt, 'cause when encoding the same video using divx 5 I get a 24 bit sample size.

I am following the guide on doom9.org, and my question is: why does this happen, and Am I losing quality due to this?

[Selur]

"why does this happen, and Am I losing quality due to this?"
use the forum search this has been answered some times before,.. And no you don't lose quality you stay the whole time in the Yv12 colorspace,..

Cu Selur

Ps.: you should find posts like this:
http://forum.doom9.org/showthread.php?s=&threadid=13431

[unmei]

I'm currently using pure YV12 chain nd little to no avisynth filtering... but still this question bugs me when i read this:

when applying heavy filtering (like default convolution3d et all) wouldnt produce less errors to go into rgb 24 or 32 (hiest possible) in the beginning of the chain and only downsamle to YV12 in the end ?

My reasoning comes from audio processing, where its better to have the entire filter chain at may bitdepth (64 ie) to reduce math errors like rounding...

i mean in a YV12 chain a filter has 2 possibilites:
a) process few-bit samples with probably little precision or
b) upsample internally and downsample ("dither") before output.
In a) you can get quite a bunch or math errors accumulated and in b) you have many consequtive colorspace conversions - both are not that great :(

well, like i said the reasoning is from audio and could be completely wrong for video (especially if there are only integer operations, but i dare to doubt there is no single float in any of the filters).

mods please feel free to move if you think its a AviSynth or whatever post, it just happened to come to my mind when reading the above question :)

typos VY->YV

[temporance]

My reasoning comes from audio processing, where its better to have the entire filter chain at may bitdepthBitdepth is always exactly 8-bits for PC video whether you're working with RGB24, RGB32, YV12 or YUY2. What changes is chroma subsampling - which is more akin to reducing audio sample rate. In terms of quality, it's just as good to apply filtering with YV12.

[unmei]

so you have "half the samples at full bitdepth", not "all samples at half the bitdepth" for chroma ...hell i should have known this :)

thanks , i must have been sleeping while writing this :p

[geoffwa]

Your audio processing analogy should also include the fact that upsampling and downsampling the signal is a lossy process (in most instances anyway).;)