On Thu, Sep 21, 2017 at 5:40 PM, Steve Lhomme <rob...@gmail.com> wrote:

> On Thu, Sep 21, 2017 at 4:58 PM, Vittorio Giovara
> <vittorio.giov...@gmail.com> wrote:
> > On Thu, Sep 21, 2017 at 4:03 PM, Steve Lhomme <rob...@gmail.com> wrote:
> >
> >> From: "Mohammed (Shaan) Huzaifa Danish" <sha...@gmail.com>
> >>
> >> ---
> >> fix previous patch skipping the wrong amount of bits
> >> ---
> >>  libavcodec/h264_sei.c   | 6 ++++--
> >>  libavcodec/h264_sei.h   | 1 +
> >>  libavcodec/h264_slice.c | 2 ++
> >>  libavutil/stereo3d.h    | 4 ++++
> >>  4 files changed, 11 insertions(+), 2 deletions(-)
> >>
> >> diff --git a/libavcodec/h264_sei.c b/libavcodec/h264_sei.c
> >> index 03fca9017f..96ac427931 100644
> >> --- a/libavcodec/h264_sei.c
> >> +++ b/libavcodec/h264_sei.c
> >> @@ -315,9 +315,11 @@ static int decode_frame_packing_arrangeme
> nt(H264SEIFramePacking
> >> *h,
> >>          h->content_interpretation_type    = get_bits(gb, 6);
> >>
> >>          // the following skips: spatial_flipping_flag,
> >> frame0_flipped_flag,
> >> -        // field_views_flag, current_frame_is_frame0_flag,
> >> +        // field_views_flag
> >> +        skip_bits(gb, 3);
> >> +        h->current_frame_is_frame0_flag   = get_bits1(gb);
> >>          // frame0_self_contained_flag, frame1_self_contained_flag
> >> -        skip_bits(gb, 6);
> >> +        skip_bits(gb, 2);
> >>
> >>          if (!h->quincunx_subsampling && h->arrangement_type != 5)
> >>              skip_bits(gb, 16);      // frame[01]_grid_position_[xy]
> >> diff --git a/libavcodec/h264_sei.h b/libavcodec/h264_sei.h
> >> index f6ac6034da..c3a19dd831 100644
> >> --- a/libavcodec/h264_sei.h
> >> +++ b/libavcodec/h264_sei.h
> >> @@ -108,6 +108,7 @@ typedef struct H264SEIFramePacking {
> >>      int arrangement_type;
> >>      int content_interpretation_type;
> >>      int quincunx_subsampling;
> >> +    int current_frame_is_frame0_flag;
> >>  } H264SEIFramePacking;
> >>
> >>  typedef struct H264SEIDisplayOrientation {
> >> diff --git a/libavcodec/h264_slice.c b/libavcodec/h264_slice.c
> >> index 5dd01d836e..f5b78bfe29 100644
> >> --- a/libavcodec/h264_slice.c
> >> +++ b/libavcodec/h264_slice.c
> >> @@ -1112,6 +1112,8 @@ static int h264_export_frame_props(H264Context
> *h)
> >>
> >>          if (fp->content_interpretation_type == 2)
> >>              stereo->flags = AV_STEREO3D_FLAG_INVERT;
> >> +        if (fp->current_frame_is_frame0_flag)
> >> +            stereo->flags |= AV_STEREO3D_FLAG_FRAME0;
> >>      }
> >>
> >>      if (h->sei.display_orientation.present &&
> >> diff --git a/libavutil/stereo3d.h b/libavutil/stereo3d.h
> >> index 0fa9f63a2c..4c8ab5c40a 100644
> >> --- a/libavutil/stereo3d.h
> >> +++ b/libavutil/stereo3d.h
> >> @@ -146,6 +146,10 @@ enum AVStereo3DType {
> >>   * Inverted views, Right/Bottom represents the left view.
> >>   */
> >>  #define AV_STEREO3D_FLAG_INVERT     (1 << 0)
> >> +/**
> >> + * This frame is frame0 (left view), otherwise it's the right view.
> >> + */
> >> +#define AV_STEREO3D_FLAG_FRAME0     (1 << 1)
> >>
> >
> > Hey Steve,
> > I would rather not expose a flag for something this specific (as it
> applies
> > to a single type only).
> > The name is also peculiar, and confusing, the left view has always
> priority
> > so that should be the default variant (ie no flag).
>
> If left is the default, how do you signal that it's the right eye ?
>

You can keep track of the decoded frames, if frame_num%2 == 1 it means
you're on a right view.


> Given in all other cases where this SEI is not present you don't know
> if you can assume it's there or not ?


if you want, given that the left view has priority, you could do something
like
if packing_type == 5
  if flag0 == 0
    flags |= invert
  else
    flags = 0


> The flag signals that this frame is the left eye and the one coming
> next is the right eye.
>

I'm not sure that's the best approach because this flag would be for a very
specific packing type only, and not generic (like the inverted one).

Do you have a sample available?
-- 
Vittorio
_______________________________________________
libav-devel mailing list
libav-devel@libav.org
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to