[vlc-devel] pixel_bits and pixel_size ?

Juha Jeronen juha.jeronen at jyu.fi
Mon Mar 19 08:12:13 CET 2012


Hi all,

I think this info should be wikified. Suggestions for a good page to put
it on?

 -J


On 17/03/12 20:54, Laurent Aimar wrote:
> On Sat, Mar 17, 2012 at 08:33:22PM +0200, Rémi Denis-Courmont wrote:
>> 	Hello,
>>
>> I don't understand what pixel_bits and pixel_size are supposed to represent. 
>> Is it the number of bits per pixel on the first plane? Shouldn't it be a per-
>> plane rather than per-chroma value?
> The names are not perfect, suggestions are welcome :)
>
> In vlc_chroma_description_t:
> pixel_bits is the number of bits per pixel (on a line) for a given plane
> pixels_size is the number of bytes per pixel (on a line) for a given plane
> given that the number of pixels on a line is the same as the width (scaled
> by the sampling ratio for the plane).
>
> Ex:
> For standard I420: pixel_bits = 8, pixels_size 1
> For 10 bit I420: pixel_bits = 10, pixels_size 2
> For YUY2: pixel_bits = 16, pixels_size = 2
> Thoses are pretty logical in my opinion (given the definition).
>
>
> For NV12: pixel_bits = 8, pixels_size = 1
> This one is buggy, but the picture allocation works as the UV plane
> is declared as not sub sampled. I think it should be changed to:
>
> Y: pixel_bits = 8, pixels_size = 1
> UV: pixel_bits = 16, pixels_size = 2, w_num = 1, w_den = 2
> but it would force to move the pixel_bits/pixels_size per plane just for this
> one, which make use of them more complicated.
>
> So, the current definitions/values work for the picture allocation.
> For a user (eg: when implementing a filter), thoses values are only usefull
> for planar formats IMHO as packed/semi-packed format usually need dedicated
> code :(
>




More information about the vlc-devel mailing list