2 * Copyright (c) 2011 Stefano Sabatini
4 * This file is part of FFmpeg.
6 * FFmpeg is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23 * filter for selecting which frame passes in the filterchain
26 #include "libavutil/eval.h"
27 #include "libavutil/fifo.h"
28 #include "libavutil/internal.h"
29 #include "libavutil/opt.h"
37 #include "libavcodec/dsputil.h"
40 static const char *const var_names[] = {
43 "pts", ///< original pts in the file of the frame
44 "start_pts", ///< first PTS in the stream, expressed in TB units
45 "prev_pts", ///< previous frame PTS
46 "prev_selected_pts", ///< previous selected frame PTS
48 "t", ///< first PTS in seconds
49 "start_t", ///< first PTS in the stream, expressed in seconds
50 "prev_t", ///< previous frame time
51 "prev_selected_t", ///< previously selected time
53 "pict_type", ///< the type of picture in the movie
62 "interlace_type", ///< the frame interlace type
67 "consumed_samples_n",///< number of samples consumed by the filter (only audio)
68 "samples_n", ///< number of samples in the current frame (only audio)
69 "sample_rate", ///< sample rate (only audio)
71 "n", ///< frame number (starting from zero)
72 "selected_n", ///< selected frame number (starting from zero)
73 "prev_selected_n", ///< number of the last selected frame
75 "key", ///< tell if the frame is a key frame
76 "pos", ///< original position in the file of the frame
89 VAR_PREV_SELECTED_PTS,
106 VAR_INTERLACE_TYPE_P,
107 VAR_INTERLACE_TYPE_T,
108 VAR_INTERLACE_TYPE_B,
110 VAR_CONSUMED_SAMPLES_N,
127 const AVClass *class;
130 double var_values[VAR_VARS_NB];
131 int do_scene_detect; ///< 1 if the expression requires scene detection variables, 0 otherwise
133 AVCodecContext *avctx; ///< codec context required for the DSPContext (scene detect only)
134 DSPContext c; ///< context providing optimized SAD methods (scene detect only)
135 double prev_mafd; ///< previous MAFD (scene detect only)
137 AVFilterBufferRef *prev_picref; ///< previous frame (scene detect only)
141 #define OFFSET(x) offsetof(SelectContext, x)
142 #define FLAGS AV_OPT_FLAG_FILTERING_PARAM
143 static const AVOption options[] = {
144 { "expr", "set selection expression", OFFSET(expr_str), AV_OPT_TYPE_STRING, {.str = "1"}, 0, 0, FLAGS },
145 { "e", "set selection expression", OFFSET(expr_str), AV_OPT_TYPE_STRING, {.str = "1"}, 0, 0, FLAGS },
149 static av_cold int init(AVFilterContext *ctx, const char *args, const AVClass *class)
151 SelectContext *select = ctx->priv;
152 const char *shorthand[] = { "expr", NULL };
155 select->class = class;
156 av_opt_set_defaults(select);
158 if ((ret = av_opt_set_from_string(select, args, shorthand, "=", ":")) < 0)
161 if ((ret = av_expr_parse(&select->expr, select->expr_str,
162 var_names, NULL, NULL, NULL, NULL, 0, ctx)) < 0) {
163 av_log(ctx, AV_LOG_ERROR, "Error while parsing expression '%s'\n", select->expr_str);
166 select->do_scene_detect = !!strstr(select->expr_str, "scene");
171 #define INTERLACE_TYPE_P 0
172 #define INTERLACE_TYPE_T 1
173 #define INTERLACE_TYPE_B 2
175 static int config_input(AVFilterLink *inlink)
177 SelectContext *select = inlink->dst->priv;
179 select->var_values[VAR_N] = 0.0;
180 select->var_values[VAR_SELECTED_N] = 0.0;
182 select->var_values[VAR_TB] = av_q2d(inlink->time_base);
184 select->var_values[VAR_PREV_PTS] = NAN;
185 select->var_values[VAR_PREV_SELECTED_PTS] = NAN;
186 select->var_values[VAR_PREV_SELECTED_T] = NAN;
187 select->var_values[VAR_START_PTS] = NAN;
188 select->var_values[VAR_START_T] = NAN;
190 select->var_values[VAR_PICT_TYPE_I] = AV_PICTURE_TYPE_I;
191 select->var_values[VAR_PICT_TYPE_P] = AV_PICTURE_TYPE_P;
192 select->var_values[VAR_PICT_TYPE_B] = AV_PICTURE_TYPE_B;
193 select->var_values[VAR_PICT_TYPE_SI] = AV_PICTURE_TYPE_SI;
194 select->var_values[VAR_PICT_TYPE_SP] = AV_PICTURE_TYPE_SP;
196 select->var_values[VAR_INTERLACE_TYPE_P] = INTERLACE_TYPE_P;
197 select->var_values[VAR_INTERLACE_TYPE_T] = INTERLACE_TYPE_T;
198 select->var_values[VAR_INTERLACE_TYPE_B] = INTERLACE_TYPE_B;
200 select->var_values[VAR_PICT_TYPE] = NAN;
201 select->var_values[VAR_INTERLACE_TYPE] = NAN;
202 select->var_values[VAR_SCENE] = NAN;
203 select->var_values[VAR_CONSUMED_SAMPLES_N] = NAN;
204 select->var_values[VAR_SAMPLES_N] = NAN;
206 select->var_values[VAR_SAMPLE_RATE] =
207 inlink->type == AVMEDIA_TYPE_AUDIO ? inlink->sample_rate : NAN;
210 if (select->do_scene_detect) {
211 select->avctx = avcodec_alloc_context3(NULL);
213 return AVERROR(ENOMEM);
214 dsputil_init(&select->c, select->avctx);
221 static double get_scene_score(AVFilterContext *ctx, AVFilterBufferRef *picref)
224 SelectContext *select = ctx->priv;
225 AVFilterBufferRef *prev_picref = select->prev_picref;
228 picref->video->h == prev_picref->video->h &&
229 picref->video->w == prev_picref->video->w &&
230 picref->linesize[0] == prev_picref->linesize[0]) {
231 int x, y, nb_sad = 0;
234 uint8_t *p1 = picref->data[0];
235 uint8_t *p2 = prev_picref->data[0];
236 const int linesize = picref->linesize[0];
238 for (y = 0; y < picref->video->h - 8; y += 8) {
239 for (x = 0; x < picref->video->w*3 - 8; x += 8) {
240 sad += select->c.sad[1](select, p1 + x, p2 + x,
248 mafd = nb_sad ? sad / nb_sad : 0;
249 diff = fabs(mafd - select->prev_mafd);
250 ret = av_clipf(FFMIN(mafd, diff) / 100., 0, 1);
251 select->prev_mafd = mafd;
252 avfilter_unref_buffer(prev_picref);
254 select->prev_picref = avfilter_ref_buffer(picref, ~0);
259 #define D2TS(d) (isnan(d) ? AV_NOPTS_VALUE : (int64_t)(d))
260 #define TS2D(ts) ((ts) == AV_NOPTS_VALUE ? NAN : (double)(ts))
262 static int select_frame(AVFilterContext *ctx, AVFilterBufferRef *ref)
264 SelectContext *select = ctx->priv;
265 AVFilterLink *inlink = ctx->inputs[0];
268 if (isnan(select->var_values[VAR_START_PTS]))
269 select->var_values[VAR_START_PTS] = TS2D(ref->pts);
270 if (isnan(select->var_values[VAR_START_T]))
271 select->var_values[VAR_START_T] = TS2D(ref->pts) * av_q2d(inlink->time_base);
273 select->var_values[VAR_PTS] = TS2D(ref->pts);
274 select->var_values[VAR_T ] = TS2D(ref->pts) * av_q2d(inlink->time_base);
275 select->var_values[VAR_POS] = ref->pos == -1 ? NAN : ref->pos;
276 select->var_values[VAR_PREV_PTS] = TS2D(ref ->pts);
278 switch (inlink->type) {
279 case AVMEDIA_TYPE_AUDIO:
280 select->var_values[VAR_SAMPLES_N] = ref->audio->nb_samples;
283 case AVMEDIA_TYPE_VIDEO:
284 select->var_values[VAR_INTERLACE_TYPE] =
285 !ref->video->interlaced ? INTERLACE_TYPE_P :
286 ref->video->top_field_first ? INTERLACE_TYPE_T : INTERLACE_TYPE_B;
287 select->var_values[VAR_PICT_TYPE] = ref->video->pict_type;
289 if (select->do_scene_detect) {
291 select->var_values[VAR_SCENE] = get_scene_score(ctx, ref);
292 // TODO: document metadata
293 snprintf(buf, sizeof(buf), "%f", select->var_values[VAR_SCENE]);
294 av_dict_set(&ref->metadata, "lavfi.scene_score", buf, 0);
300 res = av_expr_eval(select->expr, select->var_values, NULL);
301 av_log(inlink->dst, AV_LOG_DEBUG,
302 "n:%d pts:%d t:%f pos:%d key:%d",
303 (int)select->var_values[VAR_N],
304 (int)select->var_values[VAR_PTS],
305 select->var_values[VAR_T],
306 (int)select->var_values[VAR_POS],
307 (int)select->var_values[VAR_KEY]);
309 switch (inlink->type) {
310 case AVMEDIA_TYPE_VIDEO:
311 av_log(inlink->dst, AV_LOG_DEBUG, " interlace_type:%c pict_type:%c scene:%f",
312 select->var_values[VAR_INTERLACE_TYPE] == INTERLACE_TYPE_P ? 'P' :
313 select->var_values[VAR_INTERLACE_TYPE] == INTERLACE_TYPE_T ? 'T' :
314 select->var_values[VAR_INTERLACE_TYPE] == INTERLACE_TYPE_B ? 'B' : '?',
315 av_get_picture_type_char(select->var_values[VAR_PICT_TYPE]),
316 select->var_values[VAR_SCENE]);
318 case AVMEDIA_TYPE_AUDIO:
319 av_log(inlink->dst, AV_LOG_DEBUG, " samples_n:%d consumed_samples_n:%d",
320 (int)select->var_values[VAR_SAMPLES_N],
321 (int)select->var_values[VAR_CONSUMED_SAMPLES_N]);
325 av_log(inlink->dst, AV_LOG_DEBUG, " -> select:%f\n", res);
328 select->var_values[VAR_PREV_SELECTED_N] = select->var_values[VAR_N];
329 select->var_values[VAR_PREV_SELECTED_PTS] = select->var_values[VAR_PTS];
330 select->var_values[VAR_PREV_SELECTED_T] = select->var_values[VAR_T];
331 select->var_values[VAR_SELECTED_N] += 1.0;
332 if (inlink->type == AVMEDIA_TYPE_AUDIO)
333 select->var_values[VAR_CONSUMED_SAMPLES_N] += ref->audio->nb_samples;
336 select->var_values[VAR_N] += 1.0;
341 static int filter_frame(AVFilterLink *inlink, AVFilterBufferRef *frame)
343 SelectContext *select = inlink->dst->priv;
345 select->select = select_frame(inlink->dst, frame);
347 return ff_filter_frame(inlink->dst->outputs[0], frame);
349 avfilter_unref_bufferp(&frame);
353 static int request_frame(AVFilterLink *outlink)
355 AVFilterContext *ctx = outlink->src;
356 SelectContext *select = ctx->priv;
357 AVFilterLink *inlink = outlink->src->inputs[0];
361 int ret = ff_request_frame(inlink);
364 } while (!select->select);
369 static av_cold void uninit(AVFilterContext *ctx)
371 SelectContext *select = ctx->priv;
373 av_expr_free(select->expr);
378 if (select->do_scene_detect) {
379 avfilter_unref_bufferp(&select->prev_picref);
381 avcodec_close(select->avctx);
382 av_freep(&select->avctx);
388 static int query_formats(AVFilterContext *ctx)
390 SelectContext *select = ctx->priv;
392 if (!select->do_scene_detect) {
393 return ff_default_query_formats(ctx);
395 static const enum AVPixelFormat pix_fmts[] = {
396 AV_PIX_FMT_RGB24, AV_PIX_FMT_BGR24,
399 ff_set_common_formats(ctx, ff_make_format_list(pix_fmts));
404 #if CONFIG_ASELECT_FILTER
406 #define aselect_options options
407 AVFILTER_DEFINE_CLASS(aselect);
409 static av_cold int aselect_init(AVFilterContext *ctx, const char *args)
411 SelectContext *select = ctx->priv;
414 if ((ret = init(ctx, args, &aselect_class)) < 0)
417 if (select->do_scene_detect) {
418 av_log(ctx, AV_LOG_ERROR, "Scene detection is ignored in aselect filter\n");
419 return AVERROR(EINVAL);
425 static const AVFilterPad avfilter_af_aselect_inputs[] = {
428 .type = AVMEDIA_TYPE_AUDIO,
429 .get_audio_buffer = ff_null_get_audio_buffer,
430 .config_props = config_input,
431 .filter_frame = filter_frame,
436 static const AVFilterPad avfilter_af_aselect_outputs[] = {
439 .type = AVMEDIA_TYPE_AUDIO,
444 AVFilter avfilter_af_aselect = {
446 .description = NULL_IF_CONFIG_SMALL("Select audio frames to pass in output."),
447 .init = aselect_init,
449 .priv_size = sizeof(SelectContext),
450 .inputs = avfilter_af_aselect_inputs,
451 .outputs = avfilter_af_aselect_outputs,
452 .priv_class = &aselect_class,
454 #endif /* CONFIG_ASELECT_FILTER */
456 #if CONFIG_SELECT_FILTER
458 #define select_options options
459 AVFILTER_DEFINE_CLASS(select);
461 static av_cold int select_init(AVFilterContext *ctx, const char *args)
463 SelectContext *select = ctx->priv;
466 if ((ret = init(ctx, args, &select_class)) < 0)
469 if (select->do_scene_detect && !CONFIG_AVCODEC) {
470 av_log(ctx, AV_LOG_ERROR, "Scene detection is not available without libavcodec.\n");
471 return AVERROR(EINVAL);
477 static const AVFilterPad avfilter_vf_select_inputs[] = {
480 .type = AVMEDIA_TYPE_VIDEO,
481 .get_video_buffer = ff_null_get_video_buffer,
482 .min_perms = AV_PERM_PRESERVE,
483 .config_props = config_input,
484 .filter_frame = filter_frame,
489 static const AVFilterPad avfilter_vf_select_outputs[] = {
492 .type = AVMEDIA_TYPE_VIDEO,
493 .request_frame = request_frame,
498 AVFilter avfilter_vf_select = {
500 .description = NULL_IF_CONFIG_SMALL("Select video frames to pass in output."),
503 .query_formats = query_formats,
505 .priv_size = sizeof(SelectContext),
507 .inputs = avfilter_vf_select_inputs,
508 .outputs = avfilter_vf_select_outputs,
509 .priv_class = &select_class,
511 #endif /* CONFIG_SELECT_FILTER */