[theora-dev] Encoding - CPU usage ...

[Mat]

David Kuehling wrote:
>> Can't you grab directly in YUV ?  ( like V4L2_PIX_FMT_YUV420 )
>>     
> In my experience the least-common denominator for V4L2 capture would be
> V4L2_PIX_FMT_YUYV.  Which is twice as many chroma data as YUV420.  

You can choose any colorspace you like but if you convert it to YUV I 
don't think the extra chroma data will count.

I use  V4L2_PIX_FMT_YUV420  because in this way I can use  ffmpeg's  
PIX_FMT_YUV420P  or  ogg's  OC_PF_420  directly ;)
( and also SDL overlays almost directly )

> But reducing that shouldn't take much cpu power.
>   

In this way you can skip a post processing function on every frame you 
grab... I think it can help the CPU.
And you were asking about RGB to YUV conversion in these posts...

BTW, if I remember correctly in  ffmpeg  there should be a  
img_convert()  function to convert the frames between different colorspaces

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.xiph.org/pipermail/theora-dev/attachments/20061009/f9f2674f/attachment.html