On Mon, Feb 23, 2015 at 10:24 PM, Vittorio Giovara
<[email protected]> wrote:
> These value represent the original bitstream size, before any extra
> operation, such as rotation or container cropping, may be applied.
>
> Container cropping is one of the ways to encode an odd-size h264 video
> in yuv420 color space. Right now, when it is applied, the original frame size
> is lost, and there is no way to understand where the cropping took place
> (eg. at the bitstream level or at the container level).
>
> Not only analysis might want to preserve that information, but users
> may carry out additional padding when it is actually not necessary.
>
> This change allows to preserve all sizes from container and bitstream in
> the AVCodecContext. The values are to be considered as follows:
>  * coded_width/coded_height: size of the full coded video surface;
>  * orig_width/orig_height: size of the video frame to be displayed;
>  * width/height: size of the actual video to be displayed.
>
> When container cropping is not present, orig_width/orig_height match
> width/height; under no circumstance, they are allowed to be greater than
> coded_width/coded_height or smaller than width/height.
>

IMHO, this seems rather ugly. Especially with Anton's plans to
liberate demuxing from the codec context, there ought to be a better
way to store this.

- Hendrik
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to