On Wed, Nov 11, 2020 at 12:18:02AM +0000, Brian Matherly wrote:
>  > > Am I (at all) on the right track here?:
> > > 
> > > Yes!
> > > 
> > > This code seems to say the avformat producer supports rgba, rgb24, and 
> > > yuv422p? 
> > > https://github.com/mltframework/mlt/blob/master/src/modules/avformat/producer_avformat.c#L614-L630
> > > 
> > > This is a better place to look:
> > > https://github.com/mltframework/mlt/blob/master/src/modules/avformat/producer_avformat.c#L1167rgba,
> > >  rgb24, yuv420p, yuyv422
> > 
> > Got it, thanks! So AVPixelFormat is the ffmpeg(?) pixel format, and 
> > mlt_image_format is the type actually used by MLT services (or, at least, 
> > by filters and transitions)?
> 
> Yes.
> 
> > The definition of mlt_image_format 
> >(https://github.com/mltframework/mlt/blob/master/src/framework/mlt_types.h#L45-L58)
> > seems to have only mlt_image_yuv422p16 as a candidate capable of supporting 
> >10-bit color; is > that right? All the rest (excluding glsl which I haven't 
> >explored) seem to only be 8-bit?
> 
> Yes.
> 
> > And also is my math right that it is in fact able to support 10(+) bits of 
> >color? - i.e. 2 chroma-subsampled pixels have (32 bpp*2==64) bits of color 
> >information, divided by 4 (2 Y + 1 Cr + 1 Cb) == 16 bits per component?
> > A simpler way of saying this might be: the ffmpeg naming scheme for e.g. 
> >"yuv422p10le" is the same as MLT's naming scheme for "yuv422p16", thus since 
> >16 is greater than 10, we can definitely losslessly use 10-bit > colors (in 
> >services that support 16-bit colors)?
> No. yuv422p16 is 16 bits per sample (32bits per pixel) because we prefer to 
> operate on byte boundaries in MLT. It maps to AV_PIX_FMT_YUV422P16LE.
> https://github.com/mltframework/mlt/blob/master/src/modules/avformat/consumer_avformat.c#L591
> 

Thanks, all of this has been really helpful!

Would adding 10-bit color support to the avformat producer "just" mainly 
involve adding code like the below to producer_avformat.c, or is there a lot 
more to it?:

----------

Adding cases to the switch statement in `pick_pix_fmt`:

        case AV_PIX_FMT_YUV420P16LE:
        case AV_PIX_FMT_YUV422P16LE: 
        case AV_PIX_FMT_YUV420P10LE: 
        case AV_PIX_FMT_YUV422P10LE:
                return AV_PIX_FMT_YUV422P16LE;
     
----------

Adding cases to `pick_image_format`:

        case AV_PIX_FMT_YUV420P16LE:
        case AV_PIX_FMT_YUV422P16LE: 
        case AV_PIX_FMT_YUV420P10LE: 
        case AV_PIX_FMT_YUV422P10LE:
                return mlt_image_yuv422p16;

----------

Adding a case to `convert_image`

----------

Thanks again,
Tom


_______________________________________________
Mlt-devel mailing list
Mlt-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/mlt-devel

Reply via email to