avfilter/vf_scale: add optional "ref" input

This is automatically enabled if the width/height expressions reference any ref_* variable. This will ultimately serve as a more principled replacement for the fundamentally broken scale2ref.

Niklas Haas committed 2 years ago

Commit bb804451 parent e82a3993 files · +156 −3Browse files

Unified · Split

3 files

Changelog +1 −0
doc/filters.texi +26 −0
libavfilter/vf_scale.c +129 −3

Changelog+1−0

@@ -7,6 +7,7 @@ version <next>: · 6 unmodified lines
7	7	- ffmpeg CLI filtergraph chaining
8	8	- LC3/LC3plus demuxer and muxer
9	9	- pad_vaapi, drawbox_vaapi filters
	10	+- vf_scale supports secondary ref input and framesync options
10	11
11	12
12	13	version 7.0:

doc/filters.texi+26−0

@@ -21562,8 +21562,34 @@ The position (byte offset) of the frame in the input stream, or NaN if · 21561 unmodified lines
21562	21562	this information is unavailable and/or meaningless (for example in case of synthetic video).
21563	21563	Only available with @code{eval=frame}.
21564	21564	Deprecated, do not use.
	21565	+
	21566	+@item ref_w, rw
	21567	+@item ref_h, rh
	21568	+@item ref_a
	21569	+@item ref_dar, rdar
	21570	+@item ref_n
	21571	+@item ref_t
	21572	+@item ref_pos
	21573	+Eqvuialent to the above, but for a second reference input. If any of these
	21574	+variables are present, this filter accepts two inputs.
21565	21575	@end table
21566	21576
	21577	+@subsection Examples
	21578	+
	21579	+@itemize
	21580	+@item
	21581	+Scale a subtitle stream (sub) to match the main video (main) in size before overlaying
	21582	+@example
	21583	+'[main]split[a][b]; [ref][a]scale=rw:rh[c]; [b][c]overlay'
	21584	+@end example
	21585	+
	21586	+@item
	21587	+Scale a logo to 1/10th the height of a video, while preserving its display aspect ratio.
	21588	+@example
	21589	+[logo-in][video-in]scale=w=oh*dar:h=rh/10[logo-out]
	21590	+@end example
	21591	+@end itemize
	21592	+
21567	21593	@section scale2ref
21568	21594
21569	21595	Scale (resize) the input video, based on a reference video.

libavfilter/vf_scale.c+129−3

@@ -59,6 +59,17 @@ static const char *const var_names[] = { · 58 unmodified lines
59	59	#if FF_API_FRAME_PKT
60	60	"pos",
61	61	#endif
	62	+ "ref_w", "rw",
	63	+ "ref_h", "rh",
	64	+ "ref_a",
	65	+ "ref_sar",
	66	+ "ref_dar", "rdar",
	67	+ "ref_hsub",
	68	+ "ref_vsub",
	69	+ "ref_n",
	70	+ "ref_t",
	71	+ "ref_pos",
	72	+ /* Legacy variables for scale2ref */
62	73	"main_w",
63	74	"main_h",
64	75	"main_a",
@@ -89,6 +100,16 @@ enum var_name { · 24 unmodified lines
89	100	#if FF_API_FRAME_PKT
90	101	VAR_POS,
91	102	#endif
	103	+ VAR_REF_W, VAR_RW,
	104	+ VAR_REF_H, VAR_RH,
	105	+ VAR_REF_A,
	106	+ VAR_REF_SAR,
	107	+ VAR_REF_DAR, VAR_RDAR,
	108	+ VAR_REF_HSUB,
	109	+ VAR_REF_VSUB,
	110	+ VAR_REF_N,
	111	+ VAR_REF_T,
	112	+ VAR_REF_POS,
92	113	VAR_S2R_MAIN_W,
93	114	VAR_S2R_MAIN_H,
94	115	VAR_S2R_MAIN_A,
@@ -131,6 +152,7 @@ typedef struct ScaleContext { · 36 unmodified lines
131	152	int input_is_pal; ///< set to 1 if the input format is paletted
132	153	int output_is_pal; ///< set to 1 if the output format is paletted
133	154	int interlaced;
	155	+ int uses_ref;
134	156
135	157	char *w_expr; ///< width expression string
136	158	char *h_expr; ///< height expression string
@@ -190,6 +212,38 @@ static int check_exprs(AVFilterContext *ctx) · 53 unmodified lines
190	212	av_log(ctx, AV_LOG_WARNING, "Circular references detected for width '%s' and height '%s' - possibly invalid.\n", scale->w_expr, scale->h_expr);
191	213	}
192	214
	215	+ if (vars_w[VAR_REF_W] \|\| vars_h[VAR_REF_W] \|\|
	216	+ vars_w[VAR_RW] \|\| vars_h[VAR_RW] \|\|
	217	+ vars_w[VAR_REF_H] \|\| vars_h[VAR_REF_H] \|\|
	218	+ vars_w[VAR_RH] \|\| vars_h[VAR_RH] \|\|
	219	+ vars_w[VAR_REF_A] \|\| vars_h[VAR_REF_A] \|\|
	220	+ vars_w[VAR_REF_SAR] \|\| vars_h[VAR_REF_SAR] \|\|
	221	+ vars_w[VAR_REF_DAR] \|\| vars_h[VAR_REF_DAR] \|\|
	222	+ vars_w[VAR_RDAR] \|\| vars_h[VAR_RDAR] \|\|
	223	+ vars_w[VAR_REF_HSUB] \|\| vars_h[VAR_REF_HSUB] \|\|
	224	+ vars_w[VAR_REF_VSUB] \|\| vars_h[VAR_REF_VSUB] \|\|
	225	+ vars_w[VAR_REF_N] \|\| vars_h[VAR_REF_N] \|\|
	226	+ vars_w[VAR_REF_T] \|\| vars_h[VAR_REF_T] \|\|
	227	+ vars_w[VAR_REF_POS] \|\| vars_h[VAR_REF_POS]) {
	228	+ scale->uses_ref = 1;
	229	+ }
	230	+
	231	+ if (ctx->filter != &ff_vf_scale2ref &&
	232	+ (vars_w[VAR_S2R_MAIN_W] \|\| vars_h[VAR_S2R_MAIN_W] \|\|
	233	+ vars_w[VAR_S2R_MAIN_H] \|\| vars_h[VAR_S2R_MAIN_H] \|\|
	234	+ vars_w[VAR_S2R_MAIN_A] \|\| vars_h[VAR_S2R_MAIN_A] \|\|
	235	+ vars_w[VAR_S2R_MAIN_SAR] \|\| vars_h[VAR_S2R_MAIN_SAR] \|\|
	236	+ vars_w[VAR_S2R_MAIN_DAR] \|\| vars_h[VAR_S2R_MAIN_DAR] \|\|
	237	+ vars_w[VAR_S2R_MDAR] \|\| vars_h[VAR_S2R_MDAR] \|\|
	238	+ vars_w[VAR_S2R_MAIN_HSUB] \|\| vars_h[VAR_S2R_MAIN_HSUB] \|\|
	239	+ vars_w[VAR_S2R_MAIN_VSUB] \|\| vars_h[VAR_S2R_MAIN_VSUB] \|\|
	240	+ vars_w[VAR_S2R_MAIN_N] \|\| vars_h[VAR_S2R_MAIN_N] \|\|
	241	+ vars_w[VAR_S2R_MAIN_T] \|\| vars_h[VAR_S2R_MAIN_T] \|\|
	242	+ vars_w[VAR_S2R_MAIN_POS] \|\| vars_h[VAR_S2R_MAIN_POS]) ) {
	243	+ av_log(ctx, AV_LOG_ERROR, "Expressions with scale2ref variables are not valid in scale filter.\n");
	244	+ return AVERROR(EINVAL);
	245	+ }
	246	+
193	247	if (ctx->filter != &ff_vf_scale2ref &&
194	248	(vars_w[VAR_S2R_MAIN_W] \|\| vars_h[VAR_S2R_MAIN_W] \|\|
195	249	vars_w[VAR_S2R_MAIN_H] \|\| vars_h[VAR_S2R_MAIN_H] \|\|
@@ -385,6 +439,9 @@ static av_cold int init(AVFilterContext *ctx) · 189 unmodified lines
385	439	if (!threads)
386	440	av_opt_set_int(scale->sws_opts, "threads", ff_filter_get_nb_threads(ctx), 0);
387	441
	442	+ if (ctx->filter != &ff_vf_scale2ref)
	443	+ ctx->nb_inputs = scale->uses_ref ? 2 : 1;
	444	+
388	445	return 0;
389	446	}
390	447
@@ -506,6 +563,20 @@ static int scale_eval_dimensions(AVFilterContext *ctx) · 115 unmodified lines
506	563	scale->var_values[VAR_S2R_MAIN_VSUB] = 1 << main_desc->log2_chroma_h;
507	564	}
508	565
	566	+ if (scale->uses_ref) {
	567	+ const AVFilterLink *reflink = ctx->inputs[1];
	568	+ const AVPixFmtDescriptor *ref_desc = av_pix_fmt_desc_get(reflink->format);
	569	+ scale->var_values[VAR_REF_W] = scale->var_values[VAR_RW] = reflink->w;
	570	+ scale->var_values[VAR_REF_H] = scale->var_values[VAR_RH] = reflink->h;
	571	+ scale->var_values[VAR_REF_A] = (double) reflink->w / reflink->h;
	572	+ scale->var_values[VAR_REF_SAR] = reflink->sample_aspect_ratio.num ?
	573	+ (double) reflink->sample_aspect_ratio.num / reflink->sample_aspect_ratio.den : 1;
	574	+ scale->var_values[VAR_REF_DAR] = scale->var_values[VAR_RDAR] =
	575	+ scale->var_values[VAR_REF_A] * scale->var_values[VAR_REF_SAR];
	576	+ scale->var_values[VAR_REF_HSUB] = 1 << ref_desc->log2_chroma_w;
	577	+ scale->var_values[VAR_REF_VSUB] = 1 << ref_desc->log2_chroma_h;
	578	+ }
	579	+
509	580	res = av_expr_eval(scale->w_pexpr, scale->var_values, NULL);
510	581	eval_w = scale->var_values[VAR_OUT_W] = scale->var_values[VAR_OW] = (int) res == 0 ? inlink->w : (int) res;
511	582
@@ -693,6 +764,13 @@ static int config_props(AVFilterLink *outlink) · 181 unmodified lines
693	764	scale->fs.in[0].sync = 1;
694	765	scale->fs.in[0].before = EXT_STOP;
695	766	scale->fs.in[0].after = EXT_STOP;
	767	+ if (scale->uses_ref) {
	768	+ av_assert0(ctx->nb_inputs == 2);
	769	+ scale->fs.in[1].time_base = ctx->inputs[1]->time_base;
	770	+ scale->fs.in[1].sync = 0;
	771	+ scale->fs.in[1].before = EXT_NULL;
	772	+ scale->fs.in[1].after = EXT_INFINITY;
	773	+ }
696	774
697	775	ret = ff_framesync_configure(&scale->fs);
698	776	if (ret < 0)
@@ -919,13 +997,55 @@ scale: · 220 unmodified lines
919	997	static int do_scale(FFFrameSync *fs)
920	998	{
921	999	AVFilterContext *ctx = fs->parent;
	1000	+ ScaleContext *scale = ctx->priv;
922	1001	AVFilterLink *outlink = ctx->outputs[0];
923		− AVFrame in, out;
924		− int ret;
	1002	+ AVFrame out, in = NULL, *ref = NULL;
	1003	+ int ret = 0, frame_changed;
925	1004
926	1005	ret = ff_framesync_get_frame(fs, 0, &in, 1);
927	1006	if (ret < 0)
928		− return ret;
	1007	+ goto err;
	1008	+
	1009	+ if (scale->uses_ref) {
	1010	+ ret = ff_framesync_get_frame(fs, 1, &ref, 0);
	1011	+ if (ret < 0)
	1012	+ goto err;
	1013	+ }
	1014	+
	1015	+ if (ref) {
	1016	+ AVFilterLink *reflink = ctx->inputs[1];
	1017	+ frame_changed = ref->width != reflink->w \|\|
	1018	+ ref->height != reflink->h \|\|
	1019	+ ref->format != reflink->format \|\|
	1020	+ ref->sample_aspect_ratio.den != reflink->sample_aspect_ratio.den \|\|
	1021	+ ref->sample_aspect_ratio.num != reflink->sample_aspect_ratio.num \|\|
	1022	+ ref->colorspace != reflink->colorspace \|\|
	1023	+ ref->color_range != reflink->color_range;
	1024	+
	1025	+ if (frame_changed) {
	1026	+ reflink->format = ref->format;
	1027	+ reflink->w = ref->width;
	1028	+ reflink->h = ref->height;
	1029	+ reflink->sample_aspect_ratio.num = ref->sample_aspect_ratio.num;
	1030	+ reflink->sample_aspect_ratio.den = ref->sample_aspect_ratio.den;
	1031	+ reflink->colorspace = ref->colorspace;
	1032	+ reflink->color_range = ref->color_range;
	1033	+
	1034	+ ret = config_props(outlink);
	1035	+ if (ret < 0)
	1036	+ goto err;
	1037	+ }
	1038	+
	1039	+ if (scale->eval_mode == EVAL_MODE_FRAME) {
	1040	+ scale->var_values[VAR_REF_N] = reflink->frame_count_out;
	1041	+ scale->var_values[VAR_REF_T] = TS2T(ref->pts, reflink->time_base);
	1042	+#if FF_API_FRAME_PKT
	1043	+FF_DISABLE_DEPRECATION_WARNINGS
	1044	+ scale->var_values[VAR_REF_POS] = ref->pkt_pos == -1 ? NAN : ref->pkt_pos;
	1045	+FF_ENABLE_DEPRECATION_WARNINGS
	1046	+#endif
	1047	+ }
	1048	+ }
929	1049
930	1050	ret = scale_frame(ctx->inputs[0], in, &out);
931	1051	if (out) {
@@ -933,6 +1053,9 @@ static int do_scale(FFFrameSync *fs) · 1 unmodified line
933	1053	return ff_filter_frame(outlink, out);
934	1054	}
935	1055
	1056	+err:
	1057	+ if (ret < 0)
	1058	+ av_frame_free(&in);
936	1059	return ret;
937	1060	}
938	1061
@@ -1108,6 +1231,9 @@ static const AVFilterPad avfilter_vf_scale_inputs[] = { · 169 unmodified lines
1108	1231	{
1109	1232	.name = "default",
1110	1233	.type = AVMEDIA_TYPE_VIDEO,
	1234	+ }, {
	1235	+ .name = "ref",
	1236	+ .type = AVMEDIA_TYPE_VIDEO,
1111	1237	},
1112	1238	};
1113	1239

avfilter/vf_scale: add optional "ref" input

Keyboard shortcuts

General

Issues and pull requests

Issues, pull requests and Inbox

Repository code