stereo3d: detect layout signalled in the video stream (frame packing SEI, st3d)

DOCS/interface-changes/vf-format-stereo-in.txt

@@ -0,0 +1 @@
+change `--vf=format:stereo-in=no` to keep the source stereo 3D mode, `--vf=format:stereo-in=mono` now overrides it with 2D

DOCS/man/vf.rst

@@ -361,7 +361,8 @@ Available mpv-only filters are:
     ``<stereo-in>``
         Set the stereo mode the video is assumed to be encoded in. Use
         ``--vf=format:stereo-in=help`` to list all available modes. Check with
-        the ``stereo3d`` filter documentation to see what the names mean.
+        the ``stereo3d`` filter documentation to see what the names mean. The
+        default ``no`` keeps the mode signaled by the source.
 
     ``<rotate>``
         Set the rotation the video is assumed to be encoded with in degrees.

filters/f_decoder_wrapper.c

@@ -678,7 +678,8 @@ static void fix_image_params(struct priv *p,
     if (m.p_w <= 0 || m.p_h <= 0)
         m.p_w = m.p_h = 1;
 
-    m.stereo3d = p->codec->stereo_mode;
+    if (!m.stereo3d)
+        m.stereo3d = p->codec->stereo_mode;
 
     if (!mp_rect_equals(&p->codec->crop, &(struct mp_rect){0})) {
         struct mp_rect crop = p->codec->crop;

video/csputils.c

@@ -130,12 +130,12 @@ const struct m_opt_choice_alternatives pl_alpha_names[] = {
     {0}
 };
 
-// The short name _must_ match with what vf_stereo3d accepts (if supported).
+// The short names follow the input format names of the vf_stereo3d.
 // The long name in comments is closer to the Matroska spec (StereoMode element).
 // The numeric index matches the Matroska StereoMode value. If you add entries
 // that don't match Matroska, make sure demux_mkv.c rejects them properly.
 const struct m_opt_choice_alternatives mp_stereo3d_names[] = {
-    {"no",     -1}, // disable/invalid
+    {"no",     -1}, // unset, vf_format keeps the source mode
     {"mono",    0},
     {"sbs2l",   1}, // "side_by_side_left"
     {"ab2r",    2}, // "top_bottom_right"
@@ -144,8 +144,8 @@ const struct m_opt_choice_alternatives mp_stereo3d_names[] = {
     {"checkl",  5}, // "checkboard_left"  (unsupported by vf_stereo3d)
     {"irr",     6}, // "row_interleaved_right"
     {"irl",     7}, // "row_interleaved_left"
-    {"icr",     8}, // "column_interleaved_right" (unsupported by vf_stereo3d)
-    {"icl",     9}, // "column_interleaved_left" (unsupported by vf_stereo3d)
+    {"icr",     8}, // "column_interleaved_right"
+    {"icl",     9}, // "column_interleaved_left"
     {"arcc",   10}, // "anaglyph_cyan_red" (Matroska: unclear which mode)
     {"sbs2r",  11}, // "side_by_side_right"
     {"agmc",   12}, // "anaglyph_green_magenta" (Matroska: unclear which mode)

video/csputils.h

@@ -49,13 +49,21 @@ extern const struct m_opt_choice_alternatives mp_csp_light_names[];
 // The numeric values (except -1) match the Matroska StereoMode element value.
 enum mp_stereo3d_mode {
     MP_STEREO3D_INVALID = -1,
-    /* only modes explicitly referenced in the code are listed */
     MP_STEREO3D_MONO = 0,
     MP_STEREO3D_SBS2L = 1,
     MP_STEREO3D_AB2R = 2,
     MP_STEREO3D_AB2L = 3,
+    MP_STEREO3D_CHECKR = 4,
+    MP_STEREO3D_CHECKL = 5,
+    MP_STEREO3D_IRR = 6,
+    MP_STEREO3D_IRL = 7,
+    MP_STEREO3D_ICR = 8,
+    MP_STEREO3D_ICL = 9,
+    MP_STEREO3D_ARCC = 10,
     MP_STEREO3D_SBS2R = 11,
-    /* no explicit enum entries for most valid values */
+    MP_STEREO3D_AGMC = 12,
+    MP_STEREO3D_AL = 13,
+    MP_STEREO3D_AR = 14,
     MP_STEREO3D_COUNT = 15, // 14 is last valid mode
 };
 

video/filter/vf_format.c

@@ -100,7 +100,7 @@ static void set_params(struct vf_format_opts *p, struct mp_image_params *out,
         out->light = p->light;
     if (p->chroma_location)
         out->chroma_location = p->chroma_location;
-    if (p->stereo_in)
+    if (p->stereo_in >= 0)
         out->stereo3d = p->stereo_in;
     if (p->rotate >= 0)
         out->rotate = p->rotate;
@@ -304,6 +304,7 @@ const struct mp_user_filter_entry vf_format = {
         .name = "format",
         .priv_size = sizeof(OPT_BASE_STRUCT),
         .priv_defaults = &(const OPT_BASE_STRUCT){
+            .stereo_in = -1,
             .rotate = -1,
             .dovi = true,
             .enhancement_layer = true,

video/mp_image.c

@@ -26,6 +26,7 @@
 #include <libavutil/hwcontext.h>
 #include <libavutil/intreadwrite.h>
 #include <libavutil/rational.h>
+#include <libavutil/stereo3d.h>
 #include <libavcodec/avcodec.h>
 #include <libavutil/mastering_display_metadata.h>
 #include <libplacebo/utils/libav.h>
@@ -1079,6 +1080,29 @@ void mp_image_params_guess_csp(struct mp_image_params *params)
     }
 }
 
+static enum mp_stereo3d_mode stereo3d_from_av(const AVStereo3D *s3d)
+{
+    bool inv = s3d->flags & AV_STEREO3D_FLAG_INVERT;
+    switch (s3d->type) {
+    case AV_STEREO3D_SIDEBYSIDE:
+    case AV_STEREO3D_SIDEBYSIDE_QUINCUNX:
+        return inv ? MP_STEREO3D_SBS2R : MP_STEREO3D_SBS2L;
+    case AV_STEREO3D_TOPBOTTOM:
+        return inv ? MP_STEREO3D_AB2R : MP_STEREO3D_AB2L;
+    case AV_STEREO3D_CHECKERBOARD:
+        return inv ? MP_STEREO3D_CHECKR : MP_STEREO3D_CHECKL;
+    case AV_STEREO3D_LINES:
+        return inv ? MP_STEREO3D_IRR : MP_STEREO3D_IRL;
+    case AV_STEREO3D_COLUMNS:
+        return inv ? MP_STEREO3D_ICR : MP_STEREO3D_ICL;
+    case AV_STEREO3D_FRAMESEQUENCE:
+        return inv ? MP_STEREO3D_AR : MP_STEREO3D_AL;
+    default:
+        // AV_STEREO3D_2D and types mpv has no mode for.
+        return MP_STEREO3D_MONO;
+    }
+}
+
 // Create a new mp_image reference to av_frame.
 struct mp_image *mp_image_from_av_frame(struct AVFrame *src)
 {
@@ -1131,6 +1155,10 @@ struct mp_image *mp_image_from_av_frame(struct AVFrame *src)
 
     dst->params.chroma_location = pl_chroma_from_av(src->chroma_location);
 
+    sd = av_frame_get_side_data(src, AV_FRAME_DATA_STEREO3D);
+    if (sd)
+        dst->params.stereo3d = stereo3d_from_av((const AVStereo3D *)sd->data);
+
     if (src->opaque_ref) {
         struct mp_image_params *p = (void *)src->opaque_ref->data;
         dst->params.stereo3d = p->stereo3d;

video/out/gpu/osd.c

@@ -53,7 +53,6 @@ struct mpgl_osd_part {
     struct ra_tex *texture;
     int w, h;
     int num_subparts;
-    int prev_num_subparts;
     struct sub_bitmap *subparts;
     int num_vertices;
     struct vertex *vertices;
@@ -66,7 +65,6 @@ struct mpgl_osd {
     struct mpgl_osd_part *parts[MAX_OSD_PARTS];
     const struct ra_format *fmt_table[SUBBITMAP_COUNT];
     bool formats[SUBBITMAP_COUNT];
-    bool change_flag; // for reporting to API user only
     // temporary
     int stereo_mode;
     struct mp_osd_res osd_res;
@@ -81,7 +79,6 @@ struct mpgl_osd *mpgl_osd_init(struct ra *ra, struct mp_log *log,
         .log = log,
         .osd = osd,
         .ra = ra,
-        .change_flag = true,
         .scratch = talloc_zero_size(ctx, 1),
     };
 
@@ -194,7 +191,6 @@ static void gen_osd_cb(void *pctx, struct sub_bitmaps *imgs)
             ok = false;
 
         osd->change_id = imgs->change_id;
-        ctx->change_flag = true;
     }
     osd->num_subparts = ok ? imgs->num_parts : 0;
 
@@ -325,15 +321,6 @@ void mpgl_osd_generate(struct mpgl_osd *ctx, struct mp_osd_res res, double pts,
 
     osd_draw(ctx->osd, ctx->osd_res, pts, draw_flags, ctx->formats, gen_osd_cb, ctx);
     ctx->stereo_mode = stereo_mode;
-
-    // Parts going away does not necessarily result in gen_osd_cb() being called
-    // (not even with num_parts==0), so check this separately.
-    for (int n = 0; n < MAX_OSD_PARTS; n++) {
-        struct mpgl_osd_part *part = ctx->parts[n];
-        if (part->num_subparts !=  part->prev_num_subparts)
-            ctx->change_flag = true;
-        part->prev_num_subparts = part->num_subparts;
-    }
 }
 
 // See osd_resize() for remarks. This function is an optional optimization too.
@@ -342,11 +329,3 @@ void mpgl_osd_resize(struct mpgl_osd *ctx, struct mp_osd_res res, int stereo_mod
     set_res(ctx, res, stereo_mode);
     osd_resize(ctx->osd, ctx->osd_res);
 }
-
-bool mpgl_osd_check_change(struct mpgl_osd *ctx, struct mp_osd_res *res,
-                           double pts)
-{
-    ctx->change_flag = false;
-    mpgl_osd_generate(ctx, *res, pts, 0, 0);
-    return ctx->change_flag;
-}

video/out/gpu/osd.h

@@ -19,7 +19,5 @@ bool mpgl_osd_draw_prepare(struct mpgl_osd *ctx, int index,
                            struct gl_shader_cache *sc);
 void mpgl_osd_draw_finish(struct mpgl_osd *ctx, int index,
                           struct gl_shader_cache *sc, const struct ra_fbo *fbo);
-bool mpgl_osd_check_change(struct mpgl_osd *ctx, struct mp_osd_res *res,
-                           double pts);
 
 #endif

video/out/gpu/video.c

@@ -3794,12 +3794,6 @@ void gl_video_set_osd_pts(struct gl_video *p, double pts)
     p->osd_pts = pts;
 }
 
-bool gl_video_check_osd_change(struct gl_video *p, struct mp_osd_res *res,
-                               double pts)
-{
-    return p->osd ? mpgl_osd_check_change(p->osd, res, pts) : false;
-}
-
 void gl_video_resize(struct gl_video *p,
                      struct mp_rect *src, struct mp_rect *dst,
                      struct mp_osd_res *osd)

video/out/gpu/video.h

@@ -212,8 +212,6 @@ void gl_video_set_fb_depth(struct gl_video *p, int fb_depth);
 void gl_video_perfdata(struct gl_video *p, struct voctrl_performance_data *out);
 void gl_video_set_clear_color(struct gl_video *p, struct m_color color);
 void gl_video_set_osd_pts(struct gl_video *p, double pts);
-bool gl_video_check_osd_change(struct gl_video *p, struct mp_osd_res *osd,
-                               double pts);
 
 void gl_video_screenshot(struct gl_video *p, struct vo_frame *frame,
                          struct voctrl_screenshot *args);