2 * Copyright (c) 2010 Stefano Sabatini
3 * Copyright (c) 2010 Baptiste Coudurier
4 * Copyright (c) 2007 Bobby Bingham
6 * This file is part of FFmpeg.
8 * FFmpeg is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
13 * FFmpeg is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with FFmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
25 * overlay one video on top of another
30 #include "libavutil/common.h"
31 #include "libavutil/eval.h"
32 #include "libavutil/avstring.h"
33 #include "libavutil/pixdesc.h"
34 #include "libavutil/imgutils.h"
35 #include "libavutil/mathematics.h"
36 #include "libavutil/opt.h"
37 #include "libavutil/timestamp.h"
39 #include "drawutils.h"
40 #include "framesync.h"
42 #include "vf_overlay.h"
44 typedef struct ThreadData {
48 static const char *const var_names[] = {
49 "main_w", "W", ///< width of the main video
50 "main_h", "H", ///< height of the main video
51 "overlay_w", "w", ///< width of the overlay video
52 "overlay_h", "h", ///< height of the overlay video
57 "n", ///< number of frame
58 "pos", ///< position in the file
59 "t", ///< timestamp expressed in seconds
81 static av_cold void uninit(AVFilterContext *ctx)
83 OverlayContext *s = ctx->priv;
85 ff_framesync_uninit(&s->fs);
86 av_expr_free(s->x_pexpr); s->x_pexpr = NULL;
87 av_expr_free(s->y_pexpr); s->y_pexpr = NULL;
90 static inline int normalize_xy(double d, int chroma_sub)
94 return (int)d & ~((1 << chroma_sub) - 1);
97 static void eval_expr(AVFilterContext *ctx)
99 OverlayContext *s = ctx->priv;
101 s->var_values[VAR_X] = av_expr_eval(s->x_pexpr, s->var_values, NULL);
102 s->var_values[VAR_Y] = av_expr_eval(s->y_pexpr, s->var_values, NULL);
103 /* It is necessary if x is expressed from y */
104 s->var_values[VAR_X] = av_expr_eval(s->x_pexpr, s->var_values, NULL);
105 s->x = normalize_xy(s->var_values[VAR_X], s->hsub);
106 s->y = normalize_xy(s->var_values[VAR_Y], s->vsub);
109 static int set_expr(AVExpr **pexpr, const char *expr, const char *option, void *log_ctx)
116 ret = av_expr_parse(pexpr, expr, var_names,
117 NULL, NULL, NULL, NULL, 0, log_ctx);
119 av_log(log_ctx, AV_LOG_ERROR,
120 "Error when evaluating the expression '%s' for %s\n",
130 static int process_command(AVFilterContext *ctx, const char *cmd, const char *args,
131 char *res, int res_len, int flags)
133 OverlayContext *s = ctx->priv;
136 if (!strcmp(cmd, "x"))
137 ret = set_expr(&s->x_pexpr, args, cmd, ctx);
138 else if (!strcmp(cmd, "y"))
139 ret = set_expr(&s->y_pexpr, args, cmd, ctx);
141 ret = AVERROR(ENOSYS);
146 if (s->eval_mode == EVAL_MODE_INIT) {
148 av_log(ctx, AV_LOG_VERBOSE, "x:%f xi:%d y:%f yi:%d\n",
149 s->var_values[VAR_X], s->x,
150 s->var_values[VAR_Y], s->y);
155 static const enum AVPixelFormat alpha_pix_fmts[] = {
156 AV_PIX_FMT_YUVA420P, AV_PIX_FMT_YUVA422P, AV_PIX_FMT_YUVA444P,
157 AV_PIX_FMT_YUVA420P10, AV_PIX_FMT_YUVA422P10,
158 AV_PIX_FMT_ARGB, AV_PIX_FMT_ABGR, AV_PIX_FMT_RGBA,
159 AV_PIX_FMT_BGRA, AV_PIX_FMT_GBRAP, AV_PIX_FMT_NONE
162 static int query_formats(AVFilterContext *ctx)
164 OverlayContext *s = ctx->priv;
166 /* overlay formats contains alpha, for avoiding conversion with alpha information loss */
167 static const enum AVPixelFormat main_pix_fmts_yuv420[] = {
168 AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUVJ420P, AV_PIX_FMT_YUVA420P,
169 AV_PIX_FMT_NV12, AV_PIX_FMT_NV21,
172 static const enum AVPixelFormat overlay_pix_fmts_yuv420[] = {
173 AV_PIX_FMT_YUVA420P, AV_PIX_FMT_NONE
176 static const enum AVPixelFormat main_pix_fmts_yuv420p10[] = {
177 AV_PIX_FMT_YUV420P10, AV_PIX_FMT_YUVA420P10,
180 static const enum AVPixelFormat overlay_pix_fmts_yuv420p10[] = {
181 AV_PIX_FMT_YUVA420P10, AV_PIX_FMT_NONE
184 static const enum AVPixelFormat main_pix_fmts_yuv422[] = {
185 AV_PIX_FMT_YUV422P, AV_PIX_FMT_YUVJ422P, AV_PIX_FMT_YUVA422P, AV_PIX_FMT_NONE
187 static const enum AVPixelFormat overlay_pix_fmts_yuv422[] = {
188 AV_PIX_FMT_YUVA422P, AV_PIX_FMT_NONE
191 static const enum AVPixelFormat main_pix_fmts_yuv422p10[] = {
192 AV_PIX_FMT_YUV422P10, AV_PIX_FMT_YUVA422P10, AV_PIX_FMT_NONE
194 static const enum AVPixelFormat overlay_pix_fmts_yuv422p10[] = {
195 AV_PIX_FMT_YUVA422P10, AV_PIX_FMT_NONE
198 static const enum AVPixelFormat main_pix_fmts_yuv444[] = {
199 AV_PIX_FMT_YUV444P, AV_PIX_FMT_YUVJ444P, AV_PIX_FMT_YUVA444P, AV_PIX_FMT_NONE
201 static const enum AVPixelFormat overlay_pix_fmts_yuv444[] = {
202 AV_PIX_FMT_YUVA444P, AV_PIX_FMT_NONE
205 static const enum AVPixelFormat main_pix_fmts_gbrp[] = {
206 AV_PIX_FMT_GBRP, AV_PIX_FMT_GBRAP, AV_PIX_FMT_NONE
208 static const enum AVPixelFormat overlay_pix_fmts_gbrp[] = {
209 AV_PIX_FMT_GBRAP, AV_PIX_FMT_NONE
212 static const enum AVPixelFormat main_pix_fmts_rgb[] = {
213 AV_PIX_FMT_ARGB, AV_PIX_FMT_RGBA,
214 AV_PIX_FMT_ABGR, AV_PIX_FMT_BGRA,
215 AV_PIX_FMT_RGB24, AV_PIX_FMT_BGR24,
218 static const enum AVPixelFormat overlay_pix_fmts_rgb[] = {
219 AV_PIX_FMT_ARGB, AV_PIX_FMT_RGBA,
220 AV_PIX_FMT_ABGR, AV_PIX_FMT_BGRA,
224 AVFilterFormats *main_formats = NULL;
225 AVFilterFormats *overlay_formats = NULL;
229 case OVERLAY_FORMAT_YUV420:
230 if (!(main_formats = ff_make_format_list(main_pix_fmts_yuv420)) ||
231 !(overlay_formats = ff_make_format_list(overlay_pix_fmts_yuv420))) {
232 ret = AVERROR(ENOMEM);
236 case OVERLAY_FORMAT_YUV420P10:
237 if (!(main_formats = ff_make_format_list(main_pix_fmts_yuv420p10)) ||
238 !(overlay_formats = ff_make_format_list(overlay_pix_fmts_yuv420p10))) {
239 ret = AVERROR(ENOMEM);
243 case OVERLAY_FORMAT_YUV422:
244 if (!(main_formats = ff_make_format_list(main_pix_fmts_yuv422)) ||
245 !(overlay_formats = ff_make_format_list(overlay_pix_fmts_yuv422))) {
246 ret = AVERROR(ENOMEM);
250 case OVERLAY_FORMAT_YUV422P10:
251 if (!(main_formats = ff_make_format_list(main_pix_fmts_yuv422p10)) ||
252 !(overlay_formats = ff_make_format_list(overlay_pix_fmts_yuv422p10))) {
253 ret = AVERROR(ENOMEM);
257 case OVERLAY_FORMAT_YUV444:
258 if (!(main_formats = ff_make_format_list(main_pix_fmts_yuv444)) ||
259 !(overlay_formats = ff_make_format_list(overlay_pix_fmts_yuv444))) {
260 ret = AVERROR(ENOMEM);
264 case OVERLAY_FORMAT_RGB:
265 if (!(main_formats = ff_make_format_list(main_pix_fmts_rgb)) ||
266 !(overlay_formats = ff_make_format_list(overlay_pix_fmts_rgb))) {
267 ret = AVERROR(ENOMEM);
271 case OVERLAY_FORMAT_GBRP:
272 if (!(main_formats = ff_make_format_list(main_pix_fmts_gbrp)) ||
273 !(overlay_formats = ff_make_format_list(overlay_pix_fmts_gbrp))) {
274 ret = AVERROR(ENOMEM);
278 case OVERLAY_FORMAT_AUTO:
279 if (!(main_formats = ff_make_format_list(alpha_pix_fmts))) {
280 ret = AVERROR(ENOMEM);
288 if (s->format == OVERLAY_FORMAT_AUTO) {
289 ret = ff_set_common_formats(ctx, main_formats);
293 if ((ret = ff_formats_ref(main_formats , &ctx->inputs[MAIN]->out_formats )) < 0 ||
294 (ret = ff_formats_ref(overlay_formats, &ctx->inputs[OVERLAY]->out_formats)) < 0 ||
295 (ret = ff_formats_ref(main_formats , &ctx->outputs[MAIN]->in_formats )) < 0)
302 av_freep(&main_formats->formats);
303 av_freep(&main_formats);
305 av_freep(&overlay_formats->formats);
306 av_freep(&overlay_formats);
310 static int config_input_overlay(AVFilterLink *inlink)
312 AVFilterContext *ctx = inlink->dst;
313 OverlayContext *s = inlink->dst->priv;
315 const AVPixFmtDescriptor *pix_desc = av_pix_fmt_desc_get(inlink->format);
317 av_image_fill_max_pixsteps(s->overlay_pix_step, NULL, pix_desc);
319 /* Finish the configuration by evaluating the expressions
320 now when both inputs are configured. */
321 s->var_values[VAR_MAIN_W ] = s->var_values[VAR_MW] = ctx->inputs[MAIN ]->w;
322 s->var_values[VAR_MAIN_H ] = s->var_values[VAR_MH] = ctx->inputs[MAIN ]->h;
323 s->var_values[VAR_OVERLAY_W] = s->var_values[VAR_OW] = ctx->inputs[OVERLAY]->w;
324 s->var_values[VAR_OVERLAY_H] = s->var_values[VAR_OH] = ctx->inputs[OVERLAY]->h;
325 s->var_values[VAR_HSUB] = 1<<pix_desc->log2_chroma_w;
326 s->var_values[VAR_VSUB] = 1<<pix_desc->log2_chroma_h;
327 s->var_values[VAR_X] = NAN;
328 s->var_values[VAR_Y] = NAN;
329 s->var_values[VAR_N] = 0;
330 s->var_values[VAR_T] = NAN;
331 s->var_values[VAR_POS] = NAN;
333 if ((ret = set_expr(&s->x_pexpr, s->x_expr, "x", ctx)) < 0 ||
334 (ret = set_expr(&s->y_pexpr, s->y_expr, "y", ctx)) < 0)
337 s->overlay_is_packed_rgb =
338 ff_fill_rgba_map(s->overlay_rgba_map, inlink->format) >= 0;
339 s->overlay_has_alpha = ff_fmt_is_in(inlink->format, alpha_pix_fmts);
341 if (s->eval_mode == EVAL_MODE_INIT) {
343 av_log(ctx, AV_LOG_VERBOSE, "x:%f xi:%d y:%f yi:%d\n",
344 s->var_values[VAR_X], s->x,
345 s->var_values[VAR_Y], s->y);
348 av_log(ctx, AV_LOG_VERBOSE,
349 "main w:%d h:%d fmt:%s overlay w:%d h:%d fmt:%s\n",
350 ctx->inputs[MAIN]->w, ctx->inputs[MAIN]->h,
351 av_get_pix_fmt_name(ctx->inputs[MAIN]->format),
352 ctx->inputs[OVERLAY]->w, ctx->inputs[OVERLAY]->h,
353 av_get_pix_fmt_name(ctx->inputs[OVERLAY]->format));
357 static int config_output(AVFilterLink *outlink)
359 AVFilterContext *ctx = outlink->src;
360 OverlayContext *s = ctx->priv;
363 if ((ret = ff_framesync_init_dualinput(&s->fs, ctx)) < 0)
366 outlink->w = ctx->inputs[MAIN]->w;
367 outlink->h = ctx->inputs[MAIN]->h;
368 outlink->time_base = ctx->inputs[MAIN]->time_base;
370 return ff_framesync_configure(&s->fs);
373 // divide by 255 and round to nearest
374 // apply a fast variant: (X+127)/255 = ((X+127)*257+257)>>16 = ((X+128)*257)>>16
375 #define FAST_DIV255(x) ((((x) + 128) * 257) >> 16)
377 // calculate the unpremultiplied alpha, applying the general equation:
378 // alpha = alpha_overlay / ( (alpha_main + alpha_overlay) - (alpha_main * alpha_overlay) )
379 // (((x) << 16) - ((x) << 9) + (x)) is a faster version of: 255 * 255 * x
380 // ((((x) + (y)) << 8) - ((x) + (y)) - (y) * (x)) is a faster version of: 255 * (x + y)
381 #define UNPREMULTIPLY_ALPHA(x, y) ((((x) << 16) - ((x) << 9) + (x)) / ((((x) + (y)) << 8) - ((x) + (y)) - (y) * (x)))
384 * Blend image in src to destination buffer dst at position (x, y).
387 static av_always_inline void blend_slice_packed_rgb(AVFilterContext *ctx,
388 AVFrame *dst, const AVFrame *src,
389 int main_has_alpha, int x, int y,
390 int is_straight, int jobnr, int nb_jobs)
392 OverlayContext *s = ctx->priv;
393 int i, imax, j, jmax;
394 const int src_w = src->width;
395 const int src_h = src->height;
396 const int dst_w = dst->width;
397 const int dst_h = dst->height;
398 uint8_t alpha; ///< the amount of overlay to blend on to main
399 const int dr = s->main_rgba_map[R];
400 const int dg = s->main_rgba_map[G];
401 const int db = s->main_rgba_map[B];
402 const int da = s->main_rgba_map[A];
403 const int dstep = s->main_pix_step[0];
404 const int sr = s->overlay_rgba_map[R];
405 const int sg = s->overlay_rgba_map[G];
406 const int sb = s->overlay_rgba_map[B];
407 const int sa = s->overlay_rgba_map[A];
408 const int sstep = s->overlay_pix_step[0];
409 int slice_start, slice_end;
410 uint8_t *S, *sp, *d, *dp;
413 imax = FFMIN3(-y + dst_h, FFMIN(src_h, dst_h), y + src_h);
415 slice_start = i + (imax * jobnr) / nb_jobs;
416 slice_end = i + (imax * (jobnr+1)) / nb_jobs;
418 sp = src->data[0] + (slice_start) * src->linesize[0];
419 dp = dst->data[0] + (y + slice_start) * dst->linesize[0];
421 for (i = slice_start; i < slice_end; i++) {
424 d = dp + (x+j) * dstep;
426 for (jmax = FFMIN(-x + dst_w, src_w); j < jmax; j++) {
429 // if the main channel has an alpha channel, alpha has to be calculated
430 // to create an un-premultiplied (straight) alpha value
431 if (main_has_alpha && alpha != 0 && alpha != 255) {
432 uint8_t alpha_d = d[da];
433 alpha = UNPREMULTIPLY_ALPHA(alpha, alpha_d);
445 // main_value = main_value * (1 - alpha) + overlay_value * alpha
446 // since alpha is in the range 0-255, the result must divided by 255
447 d[dr] = is_straight ? FAST_DIV255(d[dr] * (255 - alpha) + S[sr] * alpha) :
448 FFMIN(FAST_DIV255(d[dr] * (255 - alpha)) + S[sr], 255);
449 d[dg] = is_straight ? FAST_DIV255(d[dg] * (255 - alpha) + S[sg] * alpha) :
450 FFMIN(FAST_DIV255(d[dg] * (255 - alpha)) + S[sg], 255);
451 d[db] = is_straight ? FAST_DIV255(d[db] * (255 - alpha) + S[sb] * alpha) :
452 FFMIN(FAST_DIV255(d[db] * (255 - alpha)) + S[sb], 255);
454 if (main_has_alpha) {
462 // apply alpha compositing: main_alpha += (1-main_alpha) * overlay_alpha
463 d[da] += FAST_DIV255((255 - d[da]) * S[sa]);
469 dp += dst->linesize[0];
470 sp += src->linesize[0];
474 #define DEFINE_BLEND_PLANE(depth, nbits) \
475 static av_always_inline void blend_plane_##depth##_##nbits##bits(AVFilterContext *ctx, \
476 AVFrame *dst, const AVFrame *src, \
477 int src_w, int src_h, \
478 int dst_w, int dst_h, \
479 int i, int hsub, int vsub, \
481 int main_has_alpha, \
490 OverlayContext *octx = ctx->priv; \
491 int src_wp = AV_CEIL_RSHIFT(src_w, hsub); \
492 int src_hp = AV_CEIL_RSHIFT(src_h, vsub); \
493 int dst_wp = AV_CEIL_RSHIFT(dst_w, hsub); \
494 int dst_hp = AV_CEIL_RSHIFT(dst_h, vsub); \
497 uint##depth##_t *s, *sp, *d, *dp, *dap, *a, *da, *ap; \
498 int jmax, j, k, kmax; \
499 int slice_start, slice_end; \
500 const uint##depth##_t max = (1 << nbits) - 1; \
501 const uint##depth##_t mid = (1 << (nbits -1)) ; \
502 int bytes = depth / 8; \
506 jmax = FFMIN3(-yp + dst_hp, FFMIN(src_hp, dst_hp), yp + src_hp); \
508 slice_start = j + (jmax * jobnr) / nb_jobs; \
509 slice_end = j + (jmax * (jobnr+1)) / nb_jobs; \
511 sp = (uint##depth##_t *)(src->data[i] + (slice_start) * src->linesize[i]); \
512 dp = (uint##depth##_t *)(dst->data[dst_plane] \
513 + (yp + slice_start) * dst->linesize[dst_plane] \
515 ap = (uint##depth##_t *)(src->data[3] + (slice_start << vsub) * src->linesize[3]); \
516 dap = (uint##depth##_t *)(dst->data[3] + ((yp + slice_start) << vsub) * dst->linesize[3]); \
518 for (j = slice_start; j < slice_end; j++) { \
520 d = dp + (xp+k) * dst_step; \
522 a = ap + (k<<hsub); \
523 da = dap + ((xp+k) << hsub); \
524 kmax = FFMIN(-xp + dst_wp, src_wp); \
526 if (nbits == 8 && ((vsub && j+1 < src_hp) || !vsub) && octx->blend_row[i]) { \
527 int c = octx->blend_row[i]((uint8_t*)d, (uint8_t*)da, (uint8_t*)s, \
528 (uint8_t*)a, kmax - k, src->linesize[3]); \
532 da += (1 << hsub) * c; \
533 a += (1 << hsub) * c; \
536 for (; k < kmax; k++) { \
537 int alpha_v, alpha_h, alpha; \
539 /* average alpha for color components, improve quality */ \
540 if (hsub && vsub && j+1 < src_hp && k+1 < src_wp) { \
541 alpha = (a[0] + a[src->linesize[3]] + \
542 a[1] + a[src->linesize[3]+1]) >> 2; \
543 } else if (hsub || vsub) { \
544 alpha_h = hsub && k+1 < src_wp ? \
545 (a[0] + a[1]) >> 1 : a[0]; \
546 alpha_v = vsub && j+1 < src_hp ? \
547 (a[0] + a[src->linesize[3]]) >> 1 : a[0]; \
548 alpha = (alpha_v + alpha_h) >> 1; \
551 /* if the main channel has an alpha channel, alpha has to be calculated */ \
552 /* to create an un-premultiplied (straight) alpha value */ \
553 if (main_has_alpha && alpha != 0 && alpha != max) { \
554 /* average alpha for color components, improve quality */ \
556 if (hsub && vsub && j+1 < src_hp && k+1 < src_wp) { \
557 alpha_d = (da[0] + da[dst->linesize[3]] + \
558 da[1] + da[dst->linesize[3]+1]) >> 2; \
559 } else if (hsub || vsub) { \
560 alpha_h = hsub && k+1 < src_wp ? \
561 (da[0] + da[1]) >> 1 : da[0]; \
562 alpha_v = vsub && j+1 < src_hp ? \
563 (da[0] + da[dst->linesize[3]]) >> 1 : da[0]; \
564 alpha_d = (alpha_v + alpha_h) >> 1; \
567 alpha = UNPREMULTIPLY_ALPHA(alpha, alpha_d); \
571 *d = (*d * (max - alpha) + *s * alpha) / max; \
573 *d = FAST_DIV255(*d * (255 - alpha) + *s * alpha); \
577 *d = av_clip((*d * (max - alpha) + *s * alpha) / max + *s - mid, -mid, mid) + mid; \
579 *d = FFMIN((*d * (max - alpha) + *s * alpha) / max + *s, max); \
582 *d = av_clip(FAST_DIV255((*d - mid) * (max - alpha)) + *s - mid, -mid, mid) + mid; \
584 *d = FFMIN(FAST_DIV255(*d * (max - alpha)) + *s, max); \
592 dp += dst->linesize[dst_plane] / bytes; \
593 sp += src->linesize[i] / bytes; \
594 ap += (1 << vsub) * src->linesize[3] / bytes; \
595 dap += (1 << vsub) * dst->linesize[3] / bytes; \
598 DEFINE_BLEND_PLANE(8, 8);
599 DEFINE_BLEND_PLANE(16, 10);
601 #define DEFINE_ALPHA_COMPOSITE(depth, nbits) \
602 static inline void alpha_composite_##depth##_##nbits##bits(const AVFrame *src, const AVFrame *dst, \
603 int src_w, int src_h, \
604 int dst_w, int dst_h, \
606 int jobnr, int nb_jobs) \
608 uint##depth##_t alpha; /* the amount of overlay to blend on to main */ \
609 uint##depth##_t *s, *sa, *d, *da; \
610 int i, imax, j, jmax; \
611 int slice_start, slice_end; \
612 const uint##depth##_t max = (1 << nbits) - 1; \
613 int bytes = depth / 8; \
615 imax = FFMIN(-y + dst_h, src_h); \
616 slice_start = (imax * jobnr) / nb_jobs; \
617 slice_end = ((imax * (jobnr+1)) / nb_jobs); \
620 sa = (uint##depth##_t *)(src->data[3] + (i + slice_start) * src->linesize[3]); \
621 da = (uint##depth##_t *)(dst->data[3] + (y + i + slice_start) * dst->linesize[3]); \
623 for (i = i + slice_start; i < slice_end; i++) { \
628 for (jmax = FFMIN(-x + dst_w, src_w); j < jmax; j++) { \
630 if (alpha != 0 && alpha != max) { \
631 uint8_t alpha_d = *d; \
632 alpha = UNPREMULTIPLY_ALPHA(alpha, alpha_d); \
636 else if (alpha > 0) { \
637 /* apply alpha compositing: main_alpha += (1-main_alpha) * overlay_alpha */ \
639 *d += (max - *d) * *s / max; \
641 *d += FAST_DIV255((max - *d) * *s); \
646 da += dst->linesize[3] / bytes; \
647 sa += src->linesize[3] / bytes; \
650 DEFINE_ALPHA_COMPOSITE(8, 8);
651 DEFINE_ALPHA_COMPOSITE(16, 10);
653 #define DEFINE_BLEND_SLICE_YUV(depth, nbits) \
654 static av_always_inline void blend_slice_yuv_##depth##_##nbits##bits(AVFilterContext *ctx, \
655 AVFrame *dst, const AVFrame *src, \
656 int hsub, int vsub, \
657 int main_has_alpha, \
660 int jobnr, int nb_jobs) \
662 OverlayContext *s = ctx->priv; \
663 const int src_w = src->width; \
664 const int src_h = src->height; \
665 const int dst_w = dst->width; \
666 const int dst_h = dst->height; \
668 blend_plane_##depth##_##nbits##bits(ctx, dst, src, src_w, src_h, dst_w, dst_h, 0, 0, 0, \
669 x, y, main_has_alpha, s->main_desc->comp[0].plane, s->main_desc->comp[0].offset, \
670 s->main_desc->comp[0].step, is_straight, 1, jobnr, nb_jobs); \
671 blend_plane_##depth##_##nbits##bits(ctx, dst, src, src_w, src_h, dst_w, dst_h, 1, hsub, vsub, \
672 x, y, main_has_alpha, s->main_desc->comp[1].plane, s->main_desc->comp[1].offset, \
673 s->main_desc->comp[1].step, is_straight, 1, jobnr, nb_jobs); \
674 blend_plane_##depth##_##nbits##bits(ctx, dst, src, src_w, src_h, dst_w, dst_h, 2, hsub, vsub, \
675 x, y, main_has_alpha, s->main_desc->comp[2].plane, s->main_desc->comp[2].offset, \
676 s->main_desc->comp[2].step, is_straight, 1, jobnr, nb_jobs); \
678 if (main_has_alpha) \
679 alpha_composite_##depth##_##nbits##bits(src, dst, src_w, src_h, dst_w, dst_h, x, y, \
682 DEFINE_BLEND_SLICE_YUV(8, 8);
683 DEFINE_BLEND_SLICE_YUV(16, 10);
685 static av_always_inline void blend_slice_planar_rgb(AVFilterContext *ctx,
686 AVFrame *dst, const AVFrame *src,
694 OverlayContext *s = ctx->priv;
695 const int src_w = src->width;
696 const int src_h = src->height;
697 const int dst_w = dst->width;
698 const int dst_h = dst->height;
700 blend_plane_8_8bits(ctx, dst, src, src_w, src_h, dst_w, dst_h, 0, 0, 0, x, y, main_has_alpha,
701 s->main_desc->comp[1].plane, s->main_desc->comp[1].offset, s->main_desc->comp[1].step, is_straight, 0,
703 blend_plane_8_8bits(ctx, dst, src, src_w, src_h, dst_w, dst_h, 1, hsub, vsub, x, y, main_has_alpha,
704 s->main_desc->comp[2].plane, s->main_desc->comp[2].offset, s->main_desc->comp[2].step, is_straight, 0,
706 blend_plane_8_8bits(ctx, dst, src, src_w, src_h, dst_w, dst_h, 2, hsub, vsub, x, y, main_has_alpha,
707 s->main_desc->comp[0].plane, s->main_desc->comp[0].offset, s->main_desc->comp[0].step, is_straight, 0,
711 alpha_composite_8_8bits(src, dst, src_w, src_h, dst_w, dst_h, x, y, jobnr, nb_jobs);
714 static int blend_slice_yuv420(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
716 OverlayContext *s = ctx->priv;
717 ThreadData *td = arg;
718 blend_slice_yuv_8_8bits(ctx, td->dst, td->src, 1, 1, 0, s->x, s->y, 1, jobnr, nb_jobs);
722 static int blend_slice_yuva420(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
724 OverlayContext *s = ctx->priv;
725 ThreadData *td = arg;
726 blend_slice_yuv_8_8bits(ctx, td->dst, td->src, 1, 1, 1, s->x, s->y, 1, jobnr, nb_jobs);
730 static int blend_slice_yuv420p10(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
732 OverlayContext *s = ctx->priv;
733 ThreadData *td = arg;
734 blend_slice_yuv_16_10bits(ctx, td->dst, td->src, 1, 1, 0, s->x, s->y, 1, jobnr, nb_jobs);
738 static int blend_slice_yuva420p10(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
740 OverlayContext *s = ctx->priv;
741 ThreadData *td = arg;
742 blend_slice_yuv_16_10bits(ctx, td->dst, td->src, 1, 1, 1, s->x, s->y, 1, jobnr, nb_jobs);
746 static int blend_slice_yuv422p10(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
748 OverlayContext *s = ctx->priv;
749 ThreadData *td = arg;
750 blend_slice_yuv_16_10bits(ctx, td->dst, td->src, 1, 0, 0, s->x, s->y, 1, jobnr, nb_jobs);
754 static int blend_slice_yuva422p10(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
756 OverlayContext *s = ctx->priv;
757 ThreadData *td = arg;
758 blend_slice_yuv_16_10bits(ctx, td->dst, td->src, 1, 0, 1, s->x, s->y, 1, jobnr, nb_jobs);
762 static int blend_slice_yuv422(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
764 OverlayContext *s = ctx->priv;
765 ThreadData *td = arg;
766 blend_slice_yuv_8_8bits(ctx, td->dst, td->src, 1, 0, 0, s->x, s->y, 1, jobnr, nb_jobs);
770 static int blend_slice_yuva422(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
772 OverlayContext *s = ctx->priv;
773 ThreadData *td = arg;
774 blend_slice_yuv_8_8bits(ctx, td->dst, td->src, 1, 0, 1, s->x, s->y, 1, jobnr, nb_jobs);
778 static int blend_slice_yuv444(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
780 OverlayContext *s = ctx->priv;
781 ThreadData *td = arg;
782 blend_slice_yuv_8_8bits(ctx, td->dst, td->src, 0, 0, 0, s->x, s->y, 1, jobnr, nb_jobs);
786 static int blend_slice_yuva444(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
788 OverlayContext *s = ctx->priv;
789 ThreadData *td = arg;
790 blend_slice_yuv_8_8bits(ctx, td->dst, td->src, 0, 0, 1, s->x, s->y, 1, jobnr, nb_jobs);
794 static int blend_slice_gbrp(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
796 OverlayContext *s = ctx->priv;
797 ThreadData *td = arg;
798 blend_slice_planar_rgb(ctx, td->dst, td->src, 0, 0, 0, s->x, s->y, 1, jobnr, nb_jobs);
802 static int blend_slice_gbrap(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
804 OverlayContext *s = ctx->priv;
805 ThreadData *td = arg;
806 blend_slice_planar_rgb(ctx, td->dst, td->src, 0, 0, 1, s->x, s->y, 1, jobnr, nb_jobs);
810 static int blend_slice_yuv420_pm(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
812 OverlayContext *s = ctx->priv;
813 ThreadData *td = arg;
814 blend_slice_yuv_8_8bits(ctx, td->dst, td->src, 1, 1, 0, s->x, s->y, 0, jobnr, nb_jobs);
818 static int blend_slice_yuva420_pm(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
820 OverlayContext *s = ctx->priv;
821 ThreadData *td = arg;
822 blend_slice_yuv_8_8bits(ctx, td->dst, td->src, 1, 1, 1, s->x, s->y, 0, jobnr, nb_jobs);
826 static int blend_slice_yuv422_pm(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
828 OverlayContext *s = ctx->priv;
829 ThreadData *td = arg;
830 blend_slice_yuv_8_8bits(ctx, td->dst, td->src, 1, 0, 0, s->x, s->y, 0, jobnr, nb_jobs);
834 static int blend_slice_yuva422_pm(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
836 OverlayContext *s = ctx->priv;
837 ThreadData *td = arg;
838 blend_slice_yuv_8_8bits(ctx, td->dst, td->src, 1, 0, 1, s->x, s->y, 0, jobnr, nb_jobs);
842 static int blend_slice_yuv444_pm(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
844 OverlayContext *s = ctx->priv;
845 ThreadData *td = arg;
846 blend_slice_yuv_8_8bits(ctx, td->dst, td->src, 0, 0, 0, s->x, s->y, 0, jobnr, nb_jobs);
850 static int blend_slice_yuva444_pm(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
852 OverlayContext *s = ctx->priv;
853 ThreadData *td = arg;
854 blend_slice_yuv_8_8bits(ctx, td->dst, td->src, 0, 0, 1, s->x, s->y, 0, jobnr, nb_jobs);
858 static int blend_slice_gbrp_pm(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
860 OverlayContext *s = ctx->priv;
861 ThreadData *td = arg;
862 blend_slice_planar_rgb(ctx, td->dst, td->src, 0, 0, 0, s->x, s->y, 0, jobnr, nb_jobs);
866 static int blend_slice_gbrap_pm(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
868 OverlayContext *s = ctx->priv;
869 ThreadData *td = arg;
870 blend_slice_planar_rgb(ctx, td->dst, td->src, 0, 0, 1, s->x, s->y, 0, jobnr, nb_jobs);
874 static int blend_slice_rgb(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
876 OverlayContext *s = ctx->priv;
877 ThreadData *td = arg;
878 blend_slice_packed_rgb(ctx, td->dst, td->src, 0, s->x, s->y, 1, jobnr, nb_jobs);
882 static int blend_slice_rgba(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
884 OverlayContext *s = ctx->priv;
885 ThreadData *td = arg;
886 blend_slice_packed_rgb(ctx, td->dst, td->src, 1, s->x, s->y, 1, jobnr, nb_jobs);
890 static int blend_slice_rgb_pm(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
892 OverlayContext *s = ctx->priv;
893 ThreadData *td = arg;
894 blend_slice_packed_rgb(ctx, td->dst, td->src, 0, s->x, s->y, 0, jobnr, nb_jobs);
898 static int blend_slice_rgba_pm(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
900 OverlayContext *s = ctx->priv;
901 ThreadData *td = arg;
902 blend_slice_packed_rgb(ctx, td->dst, td->src, 1, s->x, s->y, 0, jobnr, nb_jobs);
906 static int config_input_main(AVFilterLink *inlink)
908 OverlayContext *s = inlink->dst->priv;
909 const AVPixFmtDescriptor *pix_desc = av_pix_fmt_desc_get(inlink->format);
911 av_image_fill_max_pixsteps(s->main_pix_step, NULL, pix_desc);
913 s->hsub = pix_desc->log2_chroma_w;
914 s->vsub = pix_desc->log2_chroma_h;
916 s->main_desc = pix_desc;
918 s->main_is_packed_rgb =
919 ff_fill_rgba_map(s->main_rgba_map, inlink->format) >= 0;
920 s->main_has_alpha = ff_fmt_is_in(inlink->format, alpha_pix_fmts);
922 case OVERLAY_FORMAT_YUV420:
923 s->blend_slice = s->main_has_alpha ? blend_slice_yuva420 : blend_slice_yuv420;
925 case OVERLAY_FORMAT_YUV420P10:
926 s->blend_slice = s->main_has_alpha ? blend_slice_yuva420p10 : blend_slice_yuv420p10;
928 case OVERLAY_FORMAT_YUV422:
929 s->blend_slice = s->main_has_alpha ? blend_slice_yuva422 : blend_slice_yuv422;
931 case OVERLAY_FORMAT_YUV422P10:
932 s->blend_slice = s->main_has_alpha ? blend_slice_yuva422p10 : blend_slice_yuv422p10;
934 case OVERLAY_FORMAT_YUV444:
935 s->blend_slice = s->main_has_alpha ? blend_slice_yuva444 : blend_slice_yuv444;
937 case OVERLAY_FORMAT_RGB:
938 s->blend_slice = s->main_has_alpha ? blend_slice_rgba : blend_slice_rgb;
940 case OVERLAY_FORMAT_GBRP:
941 s->blend_slice = s->main_has_alpha ? blend_slice_gbrap : blend_slice_gbrp;
943 case OVERLAY_FORMAT_AUTO:
944 switch (inlink->format) {
945 case AV_PIX_FMT_YUVA420P:
946 s->blend_slice = blend_slice_yuva420;
948 case AV_PIX_FMT_YUVA420P10:
949 s->blend_slice = blend_slice_yuva420p10;
951 case AV_PIX_FMT_YUVA422P:
952 s->blend_slice = blend_slice_yuva422;
954 case AV_PIX_FMT_YUVA422P10:
955 s->blend_slice = blend_slice_yuva422p10;
957 case AV_PIX_FMT_YUVA444P:
958 s->blend_slice = blend_slice_yuva444;
960 case AV_PIX_FMT_ARGB:
961 case AV_PIX_FMT_RGBA:
962 case AV_PIX_FMT_BGRA:
963 case AV_PIX_FMT_ABGR:
964 s->blend_slice = blend_slice_rgba;
966 case AV_PIX_FMT_GBRAP:
967 s->blend_slice = blend_slice_gbrap;
976 if (!s->alpha_format)
980 case OVERLAY_FORMAT_YUV420:
981 s->blend_slice = s->main_has_alpha ? blend_slice_yuva420_pm : blend_slice_yuv420_pm;
983 case OVERLAY_FORMAT_YUV422:
984 s->blend_slice = s->main_has_alpha ? blend_slice_yuva422_pm : blend_slice_yuv422_pm;
986 case OVERLAY_FORMAT_YUV444:
987 s->blend_slice = s->main_has_alpha ? blend_slice_yuva444_pm : blend_slice_yuv444_pm;
989 case OVERLAY_FORMAT_RGB:
990 s->blend_slice = s->main_has_alpha ? blend_slice_rgba_pm : blend_slice_rgb_pm;
992 case OVERLAY_FORMAT_GBRP:
993 s->blend_slice = s->main_has_alpha ? blend_slice_gbrap_pm : blend_slice_gbrp_pm;
995 case OVERLAY_FORMAT_AUTO:
996 switch (inlink->format) {
997 case AV_PIX_FMT_YUVA420P:
998 s->blend_slice = blend_slice_yuva420_pm;
1000 case AV_PIX_FMT_YUVA422P:
1001 s->blend_slice = blend_slice_yuva422_pm;
1003 case AV_PIX_FMT_YUVA444P:
1004 s->blend_slice = blend_slice_yuva444_pm;
1006 case AV_PIX_FMT_ARGB:
1007 case AV_PIX_FMT_RGBA:
1008 case AV_PIX_FMT_BGRA:
1009 case AV_PIX_FMT_ABGR:
1010 s->blend_slice = blend_slice_rgba_pm;
1012 case AV_PIX_FMT_GBRAP:
1013 s->blend_slice = blend_slice_gbrap_pm;
1024 ff_overlay_init_x86(s, s->format, inlink->format,
1025 s->alpha_format, s->main_has_alpha);
1030 static int do_blend(FFFrameSync *fs)
1032 AVFilterContext *ctx = fs->parent;
1033 AVFrame *mainpic, *second;
1034 OverlayContext *s = ctx->priv;
1035 AVFilterLink *inlink = ctx->inputs[0];
1038 ret = ff_framesync_dualinput_get_writable(fs, &mainpic, &second);
1042 return ff_filter_frame(ctx->outputs[0], mainpic);
1044 if (s->eval_mode == EVAL_MODE_FRAME) {
1045 int64_t pos = mainpic->pkt_pos;
1047 s->var_values[VAR_N] = inlink->frame_count_out;
1048 s->var_values[VAR_T] = mainpic->pts == AV_NOPTS_VALUE ?
1049 NAN : mainpic->pts * av_q2d(inlink->time_base);
1050 s->var_values[VAR_POS] = pos == -1 ? NAN : pos;
1052 s->var_values[VAR_OVERLAY_W] = s->var_values[VAR_OW] = second->width;
1053 s->var_values[VAR_OVERLAY_H] = s->var_values[VAR_OH] = second->height;
1054 s->var_values[VAR_MAIN_W ] = s->var_values[VAR_MW] = mainpic->width;
1055 s->var_values[VAR_MAIN_H ] = s->var_values[VAR_MH] = mainpic->height;
1058 av_log(ctx, AV_LOG_DEBUG, "n:%f t:%f pos:%f x:%f xi:%d y:%f yi:%d\n",
1059 s->var_values[VAR_N], s->var_values[VAR_T], s->var_values[VAR_POS],
1060 s->var_values[VAR_X], s->x,
1061 s->var_values[VAR_Y], s->y);
1064 if (s->x < mainpic->width && s->x + second->width >= 0 &&
1065 s->y < mainpic->height && s->y + second->height >= 0) {
1070 ctx->internal->execute(ctx, s->blend_slice, &td, NULL, FFMIN(FFMAX(1, FFMIN3(s->y + second->height, FFMIN(second->height, mainpic->height), mainpic->height - s->y)),
1071 ff_filter_get_nb_threads(ctx)));
1073 return ff_filter_frame(ctx->outputs[0], mainpic);
1076 static av_cold int init(AVFilterContext *ctx)
1078 OverlayContext *s = ctx->priv;
1080 s->fs.on_event = do_blend;
1084 static int activate(AVFilterContext *ctx)
1086 OverlayContext *s = ctx->priv;
1087 return ff_framesync_activate(&s->fs);
1090 #define OFFSET(x) offsetof(OverlayContext, x)
1091 #define FLAGS AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
1093 static const AVOption overlay_options[] = {
1094 { "x", "set the x expression", OFFSET(x_expr), AV_OPT_TYPE_STRING, {.str = "0"}, 0, 0, FLAGS },
1095 { "y", "set the y expression", OFFSET(y_expr), AV_OPT_TYPE_STRING, {.str = "0"}, 0, 0, FLAGS },
1096 { "eof_action", "Action to take when encountering EOF from secondary input ",
1097 OFFSET(fs.opt_eof_action), AV_OPT_TYPE_INT, { .i64 = EOF_ACTION_REPEAT },
1098 EOF_ACTION_REPEAT, EOF_ACTION_PASS, .flags = FLAGS, "eof_action" },
1099 { "repeat", "Repeat the previous frame.", 0, AV_OPT_TYPE_CONST, { .i64 = EOF_ACTION_REPEAT }, .flags = FLAGS, "eof_action" },
1100 { "endall", "End both streams.", 0, AV_OPT_TYPE_CONST, { .i64 = EOF_ACTION_ENDALL }, .flags = FLAGS, "eof_action" },
1101 { "pass", "Pass through the main input.", 0, AV_OPT_TYPE_CONST, { .i64 = EOF_ACTION_PASS }, .flags = FLAGS, "eof_action" },
1102 { "eval", "specify when to evaluate expressions", OFFSET(eval_mode), AV_OPT_TYPE_INT, {.i64 = EVAL_MODE_FRAME}, 0, EVAL_MODE_NB-1, FLAGS, "eval" },
1103 { "init", "eval expressions once during initialization", 0, AV_OPT_TYPE_CONST, {.i64=EVAL_MODE_INIT}, .flags = FLAGS, .unit = "eval" },
1104 { "frame", "eval expressions per-frame", 0, AV_OPT_TYPE_CONST, {.i64=EVAL_MODE_FRAME}, .flags = FLAGS, .unit = "eval" },
1105 { "shortest", "force termination when the shortest input terminates", OFFSET(fs.opt_shortest), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, FLAGS },
1106 { "format", "set output format", OFFSET(format), AV_OPT_TYPE_INT, {.i64=OVERLAY_FORMAT_YUV420}, 0, OVERLAY_FORMAT_NB-1, FLAGS, "format" },
1107 { "yuv420", "", 0, AV_OPT_TYPE_CONST, {.i64=OVERLAY_FORMAT_YUV420}, .flags = FLAGS, .unit = "format" },
1108 { "yuv420p10", "", 0, AV_OPT_TYPE_CONST, {.i64=OVERLAY_FORMAT_YUV420P10}, .flags = FLAGS, .unit = "format" },
1109 { "yuv422", "", 0, AV_OPT_TYPE_CONST, {.i64=OVERLAY_FORMAT_YUV422}, .flags = FLAGS, .unit = "format" },
1110 { "yuv422p10", "", 0, AV_OPT_TYPE_CONST, {.i64=OVERLAY_FORMAT_YUV422P10}, .flags = FLAGS, .unit = "format" },
1111 { "yuv444", "", 0, AV_OPT_TYPE_CONST, {.i64=OVERLAY_FORMAT_YUV444}, .flags = FLAGS, .unit = "format" },
1112 { "rgb", "", 0, AV_OPT_TYPE_CONST, {.i64=OVERLAY_FORMAT_RGB}, .flags = FLAGS, .unit = "format" },
1113 { "gbrp", "", 0, AV_OPT_TYPE_CONST, {.i64=OVERLAY_FORMAT_GBRP}, .flags = FLAGS, .unit = "format" },
1114 { "auto", "", 0, AV_OPT_TYPE_CONST, {.i64=OVERLAY_FORMAT_AUTO}, .flags = FLAGS, .unit = "format" },
1115 { "repeatlast", "repeat overlay of the last overlay frame", OFFSET(fs.opt_repeatlast), AV_OPT_TYPE_BOOL, {.i64=1}, 0, 1, FLAGS },
1116 { "alpha", "alpha format", OFFSET(alpha_format), AV_OPT_TYPE_INT, {.i64=0}, 0, 1, FLAGS, "alpha_format" },
1117 { "straight", "", 0, AV_OPT_TYPE_CONST, {.i64=0}, .flags = FLAGS, .unit = "alpha_format" },
1118 { "premultiplied", "", 0, AV_OPT_TYPE_CONST, {.i64=1}, .flags = FLAGS, .unit = "alpha_format" },
1122 FRAMESYNC_DEFINE_CLASS(overlay, OverlayContext, fs);
1124 static const AVFilterPad avfilter_vf_overlay_inputs[] = {
1127 .type = AVMEDIA_TYPE_VIDEO,
1128 .config_props = config_input_main,
1132 .type = AVMEDIA_TYPE_VIDEO,
1133 .config_props = config_input_overlay,
1138 static const AVFilterPad avfilter_vf_overlay_outputs[] = {
1141 .type = AVMEDIA_TYPE_VIDEO,
1142 .config_props = config_output,
1147 AVFilter ff_vf_overlay = {
1149 .description = NULL_IF_CONFIG_SMALL("Overlay a video source on top of the input."),
1150 .preinit = overlay_framesync_preinit,
1153 .priv_size = sizeof(OverlayContext),
1154 .priv_class = &overlay_class,
1155 .query_formats = query_formats,
1156 .activate = activate,
1157 .process_command = process_command,
1158 .inputs = avfilter_vf_overlay_inputs,
1159 .outputs = avfilter_vf_overlay_outputs,
1160 .flags = AVFILTER_FLAG_SUPPORT_TIMELINE_INTERNAL |
1161 AVFILTER_FLAG_SLICE_THREADS,