• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * Copyright (c) 2017 Paul B Mahol
3  *
4  * This file is part of FFmpeg.
5  *
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.
10  *
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.
15  *
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
19  */
20 
21 #include "config_components.h"
22 
23 #include "libavutil/avstring.h"
24 #include "libavutil/imgutils.h"
25 #include "libavutil/intreadwrite.h"
26 #include "libavutil/opt.h"
27 #include "libavutil/pixdesc.h"
28 
29 #include "avfilter.h"
30 #include "formats.h"
31 #include "internal.h"
32 #include "framesync.h"
33 #include "video.h"
34 
35 typedef struct MixContext {
36     const AVClass *class;
37     const AVPixFmtDescriptor *desc;
38     char *weights_str;
39     int nb_inputs;
40     int nb_threads;
41     int duration;
42     float *weights;
43     float scale;
44     float wfactor;
45 
46     int tmix;
47     int nb_frames;
48 
49     int depth;
50     int max;
51     int planes;
52     int nb_planes;
53     int linesizes[4];
54     int height[4];
55 
56     uint8_t **data;
57     int *linesize;
58 
59     AVFrame **frames;
60     FFFrameSync fs;
61 } MixContext;
62 
query_formats(AVFilterContext * ctx)63 static int query_formats(AVFilterContext *ctx)
64 {
65     unsigned reject_flags = AV_PIX_FMT_FLAG_BITSTREAM |
66                             AV_PIX_FMT_FLAG_HWACCEL   |
67                             AV_PIX_FMT_FLAG_PAL;
68     unsigned accept_flags = 0;
69 
70     if (!HAVE_BIGENDIAN)
71         reject_flags |= AV_PIX_FMT_FLAG_BE;
72     else
73         accept_flags |= AV_PIX_FMT_FLAG_BE;
74 
75     return ff_set_common_formats(ctx, ff_formats_pixdesc_filter(accept_flags, reject_flags));
76 }
77 
parse_weights(AVFilterContext * ctx)78 static int parse_weights(AVFilterContext *ctx)
79 {
80     MixContext *s = ctx->priv;
81     char *p, *arg, *saveptr = NULL;
82     int i, last = 0;
83 
84     s->wfactor = 0.f;
85     p = s->weights_str;
86     for (i = 0; i < s->nb_inputs; i++) {
87         if (!(arg = av_strtok(p, " |", &saveptr)))
88             break;
89 
90         p = NULL;
91         if (av_sscanf(arg, "%f", &s->weights[i]) != 1) {
92             av_log(ctx, AV_LOG_ERROR, "Invalid syntax for weights[%d].\n", i);
93             return AVERROR(EINVAL);
94         }
95         s->wfactor += s->weights[i];
96         last = i;
97     }
98 
99     for (; i < s->nb_inputs; i++) {
100         s->weights[i] = s->weights[last];
101         s->wfactor += s->weights[i];
102     }
103     if (s->scale == 0) {
104         s->wfactor = 1 / s->wfactor;
105     } else {
106         s->wfactor = s->scale;
107     }
108 
109     return 0;
110 }
111 
init(AVFilterContext * ctx)112 static av_cold int init(AVFilterContext *ctx)
113 {
114     MixContext *s = ctx->priv;
115     int ret;
116 
117     s->tmix = !strcmp(ctx->filter->name, "tmix");
118 
119     s->frames = av_calloc(s->nb_inputs, sizeof(*s->frames));
120     if (!s->frames)
121         return AVERROR(ENOMEM);
122 
123     s->weights = av_calloc(s->nb_inputs, sizeof(*s->weights));
124     if (!s->weights)
125         return AVERROR(ENOMEM);
126 
127     if (!s->tmix) {
128         for (int i = 0; i < s->nb_inputs; i++) {
129             AVFilterPad pad = { 0 };
130 
131             pad.type = AVMEDIA_TYPE_VIDEO;
132             pad.name = av_asprintf("input%d", i);
133             if (!pad.name)
134                 return AVERROR(ENOMEM);
135 
136             if ((ret = ff_append_inpad_free_name(ctx, &pad)) < 0)
137                 return ret;
138         }
139     }
140 
141     return parse_weights(ctx);
142 }
143 
144 typedef struct ThreadData {
145     AVFrame **in, *out;
146 } ThreadData;
147 
148 #define MIX_SLICE(type, fun, clip)                                                              \
149     for (int p = 0; p < s->nb_planes; p++) {                                                    \
150         const int slice_start = (s->height[p] * jobnr) / nb_jobs;                               \
151         const int slice_end = (s->height[p] * (jobnr+1)) / nb_jobs;                             \
152         const int width = s->linesizes[p] / sizeof(type);                                       \
153         type *dst = (type *)(out->data[p] + slice_start * out->linesize[p]);                    \
154         ptrdiff_t dst_linesize = out->linesize[p] / sizeof(type);                               \
155                                                                                                 \
156         if (!((1 << p) & s->planes)) {                                                          \
157             av_image_copy_plane((uint8_t *)dst, out->linesize[p],                               \
158                                 in[0]->data[p] + slice_start * in[0]->linesize[p],              \
159                                 in[0]->linesize[p],                                             \
160                                 s->linesizes[p], slice_end - slice_start);                      \
161             continue;                                                                           \
162         }                                                                                       \
163                                                                                                 \
164         for (int i = 0; i < s->nb_inputs; i++)                                                  \
165             linesize[i] = in[i]->linesize[p];                                                   \
166                                                                                                 \
167         for (int i = 0; i < s->nb_inputs; i++)                                                  \
168             srcf[i] = in[i]->data[p] + slice_start * linesize[i];                               \
169                                                                                                 \
170         for (int y = slice_start; y < slice_end; y++) {                                         \
171             for (int x = 0; x < width; x++) {                                                   \
172                 float val = 0.f;                                                                \
173                                                                                                 \
174                 for (int i = 0; i < s->nb_inputs; i++) {                                        \
175                     float src = *(type *)(srcf[i] + x * sizeof(type));                          \
176                                                                                                 \
177                     val += src * weights[i];                                                    \
178                 }                                                                               \
179                                                                                                 \
180                 dst[x] = clip(fun(val * s->wfactor), 0, s->max);                                \
181             }                                                                                   \
182                                                                                                 \
183             dst += dst_linesize;                                                                \
184             for (int i = 0; i < s->nb_inputs; i++)                                              \
185                 srcf[i] += linesize[i];                                                         \
186         }                                                                                       \
187     }
188 
189 #define CLIP8(x, min, max) av_clip_uint8(x)
190 #define CLIP16(x, min, max) av_clip(x, min, max)
191 #define CLIPF(x, min, max) (x)
192 #define NOP(x) (x)
193 
mix_frames(AVFilterContext * ctx,void * arg,int jobnr,int nb_jobs)194 static int mix_frames(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
195 {
196     MixContext *s = ctx->priv;
197     ThreadData *td = arg;
198     AVFrame **in = td->in;
199     AVFrame *out = td->out;
200     const float *weights = s->weights;
201     uint8_t **srcf = s->data + jobnr * s->nb_inputs;
202     int *linesize = s->linesize + jobnr * s->nb_inputs;
203 
204     if (s->depth <= 8) {
205         MIX_SLICE(uint8_t, lrintf, CLIP8)
206     } else if (s->depth <= 16) {
207         MIX_SLICE(uint16_t, lrintf, CLIP16)
208     } else {
209         MIX_SLICE(float, NOP, CLIPF)
210     }
211 
212     return 0;
213 }
214 
process_frame(FFFrameSync * fs)215 static int process_frame(FFFrameSync *fs)
216 {
217     AVFilterContext *ctx = fs->parent;
218     AVFilterLink *outlink = ctx->outputs[0];
219     MixContext *s = fs->opaque;
220     AVFrame **in = s->frames;
221     AVFrame *out;
222     ThreadData td;
223     int i, ret;
224 
225     for (i = 0; i < s->nb_inputs; i++) {
226         if ((ret = ff_framesync_get_frame(&s->fs, i, &in[i], 0)) < 0)
227             return ret;
228     }
229 
230     if (ctx->is_disabled) {
231         out = av_frame_clone(s->frames[0]);
232         if (!out)
233             return AVERROR(ENOMEM);
234         out->pts = av_rescale_q(s->fs.pts, s->fs.time_base, outlink->time_base);
235         return ff_filter_frame(outlink, out);
236     }
237 
238     out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
239     if (!out)
240         return AVERROR(ENOMEM);
241     out->pts = av_rescale_q(s->fs.pts, s->fs.time_base, outlink->time_base);
242 
243     td.in = in;
244     td.out = out;
245     ff_filter_execute(ctx, mix_frames, &td, NULL,
246                       FFMIN(s->height[1], s->nb_threads));
247 
248     return ff_filter_frame(outlink, out);
249 }
250 
config_output(AVFilterLink * outlink)251 static int config_output(AVFilterLink *outlink)
252 {
253     AVFilterContext *ctx = outlink->src;
254     MixContext *s = ctx->priv;
255     AVRational frame_rate = ctx->inputs[0]->frame_rate;
256     AVRational sar = ctx->inputs[0]->sample_aspect_ratio;
257     AVFilterLink *inlink = ctx->inputs[0];
258     int height = ctx->inputs[0]->h;
259     int width = ctx->inputs[0]->w;
260     FFFrameSyncIn *in;
261     int i, ret;
262 
263     if (!s->tmix) {
264         for (i = 1; i < s->nb_inputs; i++) {
265             if (ctx->inputs[i]->h != height || ctx->inputs[i]->w != width) {
266                 av_log(ctx, AV_LOG_ERROR, "Input %d size (%dx%d) does not match input %d size (%dx%d).\n", i, ctx->inputs[i]->w, ctx->inputs[i]->h, 0, width, height);
267                 return AVERROR(EINVAL);
268             }
269         }
270     }
271 
272     s->nb_threads = ff_filter_get_nb_threads(ctx);
273     s->desc = av_pix_fmt_desc_get(outlink->format);
274     if (!s->desc)
275         return AVERROR_BUG;
276     s->nb_planes = av_pix_fmt_count_planes(outlink->format);
277     s->depth = s->desc->comp[0].depth;
278     s->max = (1 << s->depth) - 1;
279 
280     if ((ret = av_image_fill_linesizes(s->linesizes, inlink->format, inlink->w)) < 0)
281         return ret;
282 
283     s->height[1] = s->height[2] = AV_CEIL_RSHIFT(inlink->h, s->desc->log2_chroma_h);
284     s->height[0] = s->height[3] = inlink->h;
285 
286     s->data = av_calloc(s->nb_threads * s->nb_inputs, sizeof(*s->data));
287     if (!s->data)
288         return AVERROR(ENOMEM);
289 
290     s->linesize = av_calloc(s->nb_threads * s->nb_inputs, sizeof(*s->linesize));
291     if (!s->linesize)
292         return AVERROR(ENOMEM);
293 
294     if (s->tmix)
295         return 0;
296 
297     outlink->w          = width;
298     outlink->h          = height;
299     outlink->frame_rate = frame_rate;
300     outlink->sample_aspect_ratio = sar;
301 
302     if ((ret = ff_framesync_init(&s->fs, ctx, s->nb_inputs)) < 0)
303         return ret;
304 
305     in = s->fs.in;
306     s->fs.opaque = s;
307     s->fs.on_event = process_frame;
308 
309     for (i = 0; i < s->nb_inputs; i++) {
310         AVFilterLink *inlink = ctx->inputs[i];
311 
312         in[i].time_base = inlink->time_base;
313         in[i].sync   = 1;
314         in[i].before = EXT_STOP;
315         in[i].after  = (s->duration == 1 || (s->duration == 2 && i == 0)) ? EXT_STOP : EXT_INFINITY;
316     }
317 
318     ret = ff_framesync_configure(&s->fs);
319     outlink->time_base = s->fs.time_base;
320 
321     return ret;
322 }
323 
uninit(AVFilterContext * ctx)324 static av_cold void uninit(AVFilterContext *ctx)
325 {
326     MixContext *s = ctx->priv;
327     int i;
328 
329     ff_framesync_uninit(&s->fs);
330     av_freep(&s->weights);
331     av_freep(&s->data);
332     av_freep(&s->linesize);
333 
334     if (s->tmix) {
335         for (i = 0; i < s->nb_frames && s->frames; i++)
336             av_frame_free(&s->frames[i]);
337     }
338     av_freep(&s->frames);
339 }
340 
process_command(AVFilterContext * ctx,const char * cmd,const char * args,char * res,int res_len,int flags)341 static int process_command(AVFilterContext *ctx, const char *cmd, const char *args,
342                            char *res, int res_len, int flags)
343 {
344     int ret;
345 
346     ret = ff_filter_process_command(ctx, cmd, args, res, res_len, flags);
347     if (ret < 0)
348         return ret;
349 
350     return parse_weights(ctx);
351 }
352 
activate(AVFilterContext * ctx)353 static int activate(AVFilterContext *ctx)
354 {
355     MixContext *s = ctx->priv;
356     return ff_framesync_activate(&s->fs);
357 }
358 
359 #define OFFSET(x) offsetof(MixContext, x)
360 #define FLAGS AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_FILTERING_PARAM
361 #define TFLAGS AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_FILTERING_PARAM | AV_OPT_FLAG_RUNTIME_PARAM
362 
363 static const AVOption mix_options[] = {
364     { "inputs", "set number of inputs", OFFSET(nb_inputs), AV_OPT_TYPE_INT, {.i64=2}, 2, INT16_MAX, .flags = FLAGS },
365     { "weights", "set weight for each input", OFFSET(weights_str), AV_OPT_TYPE_STRING, {.str="1 1"}, 0, 0, .flags = TFLAGS },
366     { "scale", "set scale", OFFSET(scale), AV_OPT_TYPE_FLOAT, {.dbl=0}, 0, INT16_MAX, .flags = TFLAGS },
367     { "planes", "set what planes to filter", OFFSET(planes),   AV_OPT_TYPE_FLAGS, {.i64=15}, 0, 15,  .flags = TFLAGS },
368     { "duration", "how to determine end of stream", OFFSET(duration), AV_OPT_TYPE_INT, {.i64=0}, 0, 2, .flags = FLAGS, "duration" },
369         { "longest",  "Duration of longest input",  0, AV_OPT_TYPE_CONST, {.i64=0}, 0, 0, FLAGS, "duration" },
370         { "shortest", "Duration of shortest input", 0, AV_OPT_TYPE_CONST, {.i64=1}, 0, 0, FLAGS, "duration" },
371         { "first",    "Duration of first input",    0, AV_OPT_TYPE_CONST, {.i64=2}, 0, 0, FLAGS, "duration" },
372     { NULL },
373 };
374 
375 static const AVFilterPad outputs[] = {
376     {
377         .name          = "default",
378         .type          = AVMEDIA_TYPE_VIDEO,
379         .config_props  = config_output,
380     },
381 };
382 
383 #if CONFIG_MIX_FILTER
384 AVFILTER_DEFINE_CLASS(mix);
385 
386 const AVFilter ff_vf_mix = {
387     .name          = "mix",
388     .description   = NULL_IF_CONFIG_SMALL("Mix video inputs."),
389     .priv_size     = sizeof(MixContext),
390     .priv_class    = &mix_class,
391     FILTER_OUTPUTS(outputs),
392     FILTER_QUERY_FUNC(query_formats),
393     .init          = init,
394     .uninit        = uninit,
395     .activate      = activate,
396     .flags         = AVFILTER_FLAG_DYNAMIC_INPUTS | AVFILTER_FLAG_SLICE_THREADS |
397                      AVFILTER_FLAG_SUPPORT_TIMELINE_INTERNAL,
398     .process_command = process_command,
399 };
400 
401 #endif /* CONFIG_MIX_FILTER */
402 
403 #if CONFIG_TMIX_FILTER
tmix_filter_frame(AVFilterLink * inlink,AVFrame * in)404 static int tmix_filter_frame(AVFilterLink *inlink, AVFrame *in)
405 {
406     AVFilterContext *ctx = inlink->dst;
407     AVFilterLink *outlink = ctx->outputs[0];
408     MixContext *s = ctx->priv;
409     ThreadData td;
410     AVFrame *out;
411 
412     if (s->nb_inputs == 1)
413         return ff_filter_frame(outlink, in);
414 
415     if (s->nb_frames < s->nb_inputs) {
416         s->frames[s->nb_frames] = in;
417         s->nb_frames++;
418         while (s->nb_frames < s->nb_inputs) {
419             s->frames[s->nb_frames] = av_frame_clone(s->frames[s->nb_frames - 1]);
420             if (!s->frames[s->nb_frames])
421                 return AVERROR(ENOMEM);
422             s->nb_frames++;
423         }
424     } else {
425         av_frame_free(&s->frames[0]);
426         memmove(&s->frames[0], &s->frames[1], sizeof(*s->frames) * (s->nb_inputs - 1));
427         s->frames[s->nb_inputs - 1] = in;
428     }
429 
430     if (ctx->is_disabled) {
431         out = av_frame_clone(s->frames[0]);
432         if (!out)
433             return AVERROR(ENOMEM);
434         return ff_filter_frame(outlink, out);
435     }
436 
437     out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
438     if (!out)
439         return AVERROR(ENOMEM);
440     out->pts = s->frames[s->nb_frames - 1]->pts;
441 
442     td.out = out;
443     td.in = s->frames;
444     ff_filter_execute(ctx, mix_frames, &td, NULL,
445                       FFMIN(s->height[1], s->nb_threads));
446 
447     return ff_filter_frame(outlink, out);
448 }
449 
450 static const AVOption tmix_options[] = {
451     { "frames", "set number of successive frames to mix", OFFSET(nb_inputs), AV_OPT_TYPE_INT, {.i64=3}, 1, 1024, .flags = FLAGS },
452     { "weights", "set weight for each frame", OFFSET(weights_str), AV_OPT_TYPE_STRING, {.str="1 1 1"}, 0, 0, .flags = TFLAGS },
453     { "scale", "set scale", OFFSET(scale), AV_OPT_TYPE_FLOAT, {.dbl=0}, 0, INT16_MAX, .flags = TFLAGS },
454     { "planes", "set what planes to filter", OFFSET(planes),   AV_OPT_TYPE_FLAGS, {.i64=15}, 0, 15,  .flags = TFLAGS },
455     { NULL },
456 };
457 
458 static const AVFilterPad inputs[] = {
459     {
460         .name          = "default",
461         .type          = AVMEDIA_TYPE_VIDEO,
462         .filter_frame  = tmix_filter_frame,
463     },
464 };
465 
466 AVFILTER_DEFINE_CLASS(tmix);
467 
468 const AVFilter ff_vf_tmix = {
469     .name          = "tmix",
470     .description   = NULL_IF_CONFIG_SMALL("Mix successive video frames."),
471     .priv_size     = sizeof(MixContext),
472     .priv_class    = &tmix_class,
473     FILTER_OUTPUTS(outputs),
474     FILTER_INPUTS(inputs),
475     FILTER_QUERY_FUNC(query_formats),
476     .init          = init,
477     .uninit        = uninit,
478     .flags         = AVFILTER_FLAG_SUPPORT_TIMELINE_INTERNAL | AVFILTER_FLAG_SLICE_THREADS,
479     .process_command = process_command,
480 };
481 
482 #endif /* CONFIG_TMIX_FILTER */
483