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 "dualinput.h"
40 #include "drawutils.h"
43 static const char *const var_names[] = {
44 "main_w", "W", ///< width of the main video
45 "main_h", "H", ///< height of the main video
46 "overlay_w", "w", ///< width of the overlay video
47 "overlay_h", "h", ///< height of the overlay video
52 "n", ///< number of frame
53 "pos", ///< position in the file
54 "t", ///< timestamp expressed in seconds
61 VAR_OVERLAY_W, VAR_OW,
62 VAR_OVERLAY_H, VAR_OH,
79 static const char * const eof_action_str[] = {
80 "repeat", "endall", "pass"
102 OVERLAY_FORMAT_YUV420,
103 OVERLAY_FORMAT_YUV422,
104 OVERLAY_FORMAT_YUV444,
111 typedef struct OverlayContext {
112 const AVClass *class;
113 int x, y; ///< position of overlaid picture
115 uint8_t main_is_packed_rgb;
116 uint8_t main_rgba_map[4];
117 uint8_t main_has_alpha;
118 uint8_t overlay_is_packed_rgb;
119 uint8_t overlay_rgba_map[4];
120 uint8_t overlay_has_alpha;
121 int format; ///< OverlayFormat
122 int eval_mode; ///< EvalMode
124 FFDualInputContext dinput;
126 int main_pix_step[4]; ///< steps per pixel for each plane of the main output
127 int overlay_pix_step[4]; ///< steps per pixel for each plane of the overlay
128 int hsub, vsub; ///< chroma subsampling values
129 const AVPixFmtDescriptor *main_desc; ///< format descriptor for main input
131 double var_values[VAR_VARS_NB];
132 char *x_expr, *y_expr;
134 int eof_action; ///< action to take on EOF from source
136 AVExpr *x_pexpr, *y_pexpr;
138 void (*blend_image)(AVFilterContext *ctx, AVFrame *dst, const AVFrame *src, int x, int y);
141 static av_cold void uninit(AVFilterContext *ctx)
143 OverlayContext *s = ctx->priv;
145 ff_dualinput_uninit(&s->dinput);
146 av_expr_free(s->x_pexpr); s->x_pexpr = NULL;
147 av_expr_free(s->y_pexpr); s->y_pexpr = NULL;
150 static inline int normalize_xy(double d, int chroma_sub)
154 return (int)d & ~((1 << chroma_sub) - 1);
157 static void eval_expr(AVFilterContext *ctx)
159 OverlayContext *s = ctx->priv;
161 s->var_values[VAR_X] = av_expr_eval(s->x_pexpr, s->var_values, NULL);
162 s->var_values[VAR_Y] = av_expr_eval(s->y_pexpr, s->var_values, NULL);
163 s->var_values[VAR_X] = av_expr_eval(s->x_pexpr, s->var_values, NULL);
164 s->x = normalize_xy(s->var_values[VAR_X], s->hsub);
165 s->y = normalize_xy(s->var_values[VAR_Y], s->vsub);
168 static int set_expr(AVExpr **pexpr, const char *expr, const char *option, void *log_ctx)
175 ret = av_expr_parse(pexpr, expr, var_names,
176 NULL, NULL, NULL, NULL, 0, log_ctx);
178 av_log(log_ctx, AV_LOG_ERROR,
179 "Error when evaluating the expression '%s' for %s\n",
189 static int process_command(AVFilterContext *ctx, const char *cmd, const char *args,
190 char *res, int res_len, int flags)
192 OverlayContext *s = ctx->priv;
195 if (!strcmp(cmd, "x"))
196 ret = set_expr(&s->x_pexpr, args, cmd, ctx);
197 else if (!strcmp(cmd, "y"))
198 ret = set_expr(&s->y_pexpr, args, cmd, ctx);
200 ret = AVERROR(ENOSYS);
205 if (s->eval_mode == EVAL_MODE_INIT) {
207 av_log(ctx, AV_LOG_VERBOSE, "x:%f xi:%d y:%f yi:%d\n",
208 s->var_values[VAR_X], s->x,
209 s->var_values[VAR_Y], s->y);
214 static const enum AVPixelFormat alpha_pix_fmts[] = {
215 AV_PIX_FMT_YUVA420P, AV_PIX_FMT_YUVA422P, AV_PIX_FMT_YUVA444P,
216 AV_PIX_FMT_ARGB, AV_PIX_FMT_ABGR, AV_PIX_FMT_RGBA,
217 AV_PIX_FMT_BGRA, AV_PIX_FMT_GBRAP, AV_PIX_FMT_NONE
220 static int query_formats(AVFilterContext *ctx)
222 OverlayContext *s = ctx->priv;
224 /* overlay formats contains alpha, for avoiding conversion with alpha information loss */
225 static const enum AVPixelFormat main_pix_fmts_yuv420[] = {
226 AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUVJ420P, AV_PIX_FMT_YUVA420P,
227 AV_PIX_FMT_NV12, AV_PIX_FMT_NV21,
230 static const enum AVPixelFormat overlay_pix_fmts_yuv420[] = {
231 AV_PIX_FMT_YUVA420P, AV_PIX_FMT_NONE
234 static const enum AVPixelFormat main_pix_fmts_yuv422[] = {
235 AV_PIX_FMT_YUV422P, AV_PIX_FMT_YUVJ422P, AV_PIX_FMT_YUVA422P, AV_PIX_FMT_NONE
237 static const enum AVPixelFormat overlay_pix_fmts_yuv422[] = {
238 AV_PIX_FMT_YUVA422P, AV_PIX_FMT_NONE
241 static const enum AVPixelFormat main_pix_fmts_yuv444[] = {
242 AV_PIX_FMT_YUV444P, AV_PIX_FMT_YUVJ444P, AV_PIX_FMT_YUVA444P, AV_PIX_FMT_NONE
244 static const enum AVPixelFormat overlay_pix_fmts_yuv444[] = {
245 AV_PIX_FMT_YUVA444P, AV_PIX_FMT_NONE
248 static const enum AVPixelFormat main_pix_fmts_gbrp[] = {
249 AV_PIX_FMT_GBRP, AV_PIX_FMT_GBRAP, AV_PIX_FMT_NONE
251 static const enum AVPixelFormat overlay_pix_fmts_gbrp[] = {
252 AV_PIX_FMT_GBRAP, AV_PIX_FMT_NONE
255 static const enum AVPixelFormat main_pix_fmts_rgb[] = {
256 AV_PIX_FMT_ARGB, AV_PIX_FMT_RGBA,
257 AV_PIX_FMT_ABGR, AV_PIX_FMT_BGRA,
258 AV_PIX_FMT_RGB24, AV_PIX_FMT_BGR24,
261 static const enum AVPixelFormat overlay_pix_fmts_rgb[] = {
262 AV_PIX_FMT_ARGB, AV_PIX_FMT_RGBA,
263 AV_PIX_FMT_ABGR, AV_PIX_FMT_BGRA,
267 AVFilterFormats *main_formats = NULL;
268 AVFilterFormats *overlay_formats = NULL;
272 case OVERLAY_FORMAT_YUV420:
273 if (!(main_formats = ff_make_format_list(main_pix_fmts_yuv420)) ||
274 !(overlay_formats = ff_make_format_list(overlay_pix_fmts_yuv420))) {
275 ret = AVERROR(ENOMEM);
279 case OVERLAY_FORMAT_YUV422:
280 if (!(main_formats = ff_make_format_list(main_pix_fmts_yuv422)) ||
281 !(overlay_formats = ff_make_format_list(overlay_pix_fmts_yuv422))) {
282 ret = AVERROR(ENOMEM);
286 case OVERLAY_FORMAT_YUV444:
287 if (!(main_formats = ff_make_format_list(main_pix_fmts_yuv444)) ||
288 !(overlay_formats = ff_make_format_list(overlay_pix_fmts_yuv444))) {
289 ret = AVERROR(ENOMEM);
293 case OVERLAY_FORMAT_RGB:
294 if (!(main_formats = ff_make_format_list(main_pix_fmts_rgb)) ||
295 !(overlay_formats = ff_make_format_list(overlay_pix_fmts_rgb))) {
296 ret = AVERROR(ENOMEM);
300 case OVERLAY_FORMAT_GBRP:
301 if (!(main_formats = ff_make_format_list(main_pix_fmts_gbrp)) ||
302 !(overlay_formats = ff_make_format_list(overlay_pix_fmts_gbrp))) {
303 ret = AVERROR(ENOMEM);
307 case OVERLAY_FORMAT_AUTO:
308 if (!(main_formats = ff_make_format_list(alpha_pix_fmts))) {
309 ret = AVERROR(ENOMEM);
317 if (s->format == OVERLAY_FORMAT_AUTO) {
318 ret = ff_set_common_formats(ctx, main_formats);
322 if ((ret = ff_formats_ref(main_formats , &ctx->inputs[MAIN]->out_formats )) < 0 ||
323 (ret = ff_formats_ref(overlay_formats, &ctx->inputs[OVERLAY]->out_formats)) < 0 ||
324 (ret = ff_formats_ref(main_formats , &ctx->outputs[MAIN]->in_formats )) < 0)
331 av_freep(&main_formats->formats);
332 av_freep(&main_formats);
334 av_freep(&overlay_formats->formats);
335 av_freep(&overlay_formats);
339 static int config_input_overlay(AVFilterLink *inlink)
341 AVFilterContext *ctx = inlink->dst;
342 OverlayContext *s = inlink->dst->priv;
344 const AVPixFmtDescriptor *pix_desc = av_pix_fmt_desc_get(inlink->format);
346 av_image_fill_max_pixsteps(s->overlay_pix_step, NULL, pix_desc);
348 /* Finish the configuration by evaluating the expressions
349 now when both inputs are configured. */
350 s->var_values[VAR_MAIN_W ] = s->var_values[VAR_MW] = ctx->inputs[MAIN ]->w;
351 s->var_values[VAR_MAIN_H ] = s->var_values[VAR_MH] = ctx->inputs[MAIN ]->h;
352 s->var_values[VAR_OVERLAY_W] = s->var_values[VAR_OW] = ctx->inputs[OVERLAY]->w;
353 s->var_values[VAR_OVERLAY_H] = s->var_values[VAR_OH] = ctx->inputs[OVERLAY]->h;
354 s->var_values[VAR_HSUB] = 1<<pix_desc->log2_chroma_w;
355 s->var_values[VAR_VSUB] = 1<<pix_desc->log2_chroma_h;
356 s->var_values[VAR_X] = NAN;
357 s->var_values[VAR_Y] = NAN;
358 s->var_values[VAR_N] = 0;
359 s->var_values[VAR_T] = NAN;
360 s->var_values[VAR_POS] = NAN;
362 if ((ret = set_expr(&s->x_pexpr, s->x_expr, "x", ctx)) < 0 ||
363 (ret = set_expr(&s->y_pexpr, s->y_expr, "y", ctx)) < 0)
366 s->overlay_is_packed_rgb =
367 ff_fill_rgba_map(s->overlay_rgba_map, inlink->format) >= 0;
368 s->overlay_has_alpha = ff_fmt_is_in(inlink->format, alpha_pix_fmts);
370 if (s->eval_mode == EVAL_MODE_INIT) {
372 av_log(ctx, AV_LOG_VERBOSE, "x:%f xi:%d y:%f yi:%d\n",
373 s->var_values[VAR_X], s->x,
374 s->var_values[VAR_Y], s->y);
377 av_log(ctx, AV_LOG_VERBOSE,
378 "main w:%d h:%d fmt:%s overlay w:%d h:%d fmt:%s eof_action:%s\n",
379 ctx->inputs[MAIN]->w, ctx->inputs[MAIN]->h,
380 av_get_pix_fmt_name(ctx->inputs[MAIN]->format),
381 ctx->inputs[OVERLAY]->w, ctx->inputs[OVERLAY]->h,
382 av_get_pix_fmt_name(ctx->inputs[OVERLAY]->format),
383 eof_action_str[s->eof_action]);
387 static int config_output(AVFilterLink *outlink)
389 AVFilterContext *ctx = outlink->src;
390 OverlayContext *s = ctx->priv;
393 if ((ret = ff_dualinput_init(ctx, &s->dinput)) < 0)
396 outlink->w = ctx->inputs[MAIN]->w;
397 outlink->h = ctx->inputs[MAIN]->h;
398 outlink->time_base = ctx->inputs[MAIN]->time_base;
403 // divide by 255 and round to nearest
404 // apply a fast variant: (X+127)/255 = ((X+127)*257+257)>>16 = ((X+128)*257)>>16
405 #define FAST_DIV255(x) ((((x) + 128) * 257) >> 16)
407 // calculate the unpremultiplied alpha, applying the general equation:
408 // alpha = alpha_overlay / ( (alpha_main + alpha_overlay) - (alpha_main * alpha_overlay) )
409 // (((x) << 16) - ((x) << 9) + (x)) is a faster version of: 255 * 255 * x
410 // ((((x) + (y)) << 8) - ((x) + (y)) - (y) * (x)) is a faster version of: 255 * (x + y)
411 #define UNPREMULTIPLY_ALPHA(x, y) ((((x) << 16) - ((x) << 9) + (x)) / ((((x) + (y)) << 8) - ((x) + (y)) - (y) * (x)))
414 * Blend image in src to destination buffer dst at position (x, y).
417 static void blend_image_packed_rgb(AVFilterContext *ctx,
418 AVFrame *dst, const AVFrame *src,
419 int main_has_alpha, int x, int y)
421 OverlayContext *s = ctx->priv;
422 int i, imax, j, jmax;
423 const int src_w = src->width;
424 const int src_h = src->height;
425 const int dst_w = dst->width;
426 const int dst_h = dst->height;
427 uint8_t alpha; ///< the amount of overlay to blend on to main
428 const int dr = s->main_rgba_map[R];
429 const int dg = s->main_rgba_map[G];
430 const int db = s->main_rgba_map[B];
431 const int da = s->main_rgba_map[A];
432 const int dstep = s->main_pix_step[0];
433 const int sr = s->overlay_rgba_map[R];
434 const int sg = s->overlay_rgba_map[G];
435 const int sb = s->overlay_rgba_map[B];
436 const int sa = s->overlay_rgba_map[A];
437 const int sstep = s->overlay_pix_step[0];
438 uint8_t *S, *sp, *d, *dp;
441 sp = src->data[0] + i * src->linesize[0];
442 dp = dst->data[0] + (y+i) * dst->linesize[0];
444 for (imax = FFMIN(-y + dst_h, src_h); i < imax; i++) {
447 d = dp + (x+j) * dstep;
449 for (jmax = FFMIN(-x + dst_w, src_w); j < jmax; j++) {
452 // if the main channel has an alpha channel, alpha has to be calculated
453 // to create an un-premultiplied (straight) alpha value
454 if (main_has_alpha && alpha != 0 && alpha != 255) {
455 uint8_t alpha_d = d[da];
456 alpha = UNPREMULTIPLY_ALPHA(alpha, alpha_d);
468 // main_value = main_value * (1 - alpha) + overlay_value * alpha
469 // since alpha is in the range 0-255, the result must divided by 255
470 d[dr] = FAST_DIV255(d[dr] * (255 - alpha) + S[sr] * alpha);
471 d[dg] = FAST_DIV255(d[dg] * (255 - alpha) + S[sg] * alpha);
472 d[db] = FAST_DIV255(d[db] * (255 - alpha) + S[sb] * alpha);
474 if (main_has_alpha) {
482 // apply alpha compositing: main_alpha += (1-main_alpha) * overlay_alpha
483 d[da] += FAST_DIV255((255 - d[da]) * S[sa]);
489 dp += dst->linesize[0];
490 sp += src->linesize[0];
494 static av_always_inline void blend_plane(AVFilterContext *ctx,
495 AVFrame *dst, const AVFrame *src,
496 int src_w, int src_h,
497 int dst_w, int dst_h,
498 int i, int hsub, int vsub,
505 int src_wp = AV_CEIL_RSHIFT(src_w, hsub);
506 int src_hp = AV_CEIL_RSHIFT(src_h, vsub);
507 int dst_wp = AV_CEIL_RSHIFT(dst_w, hsub);
508 int dst_hp = AV_CEIL_RSHIFT(dst_h, vsub);
511 uint8_t *s, *sp, *d, *dp, *a, *ap;
512 int jmax, j, k, kmax;
515 sp = src->data[i] + j * src->linesize[i];
516 dp = dst->data[dst_plane]
517 + (yp+j) * dst->linesize[dst_plane]
519 ap = src->data[3] + (j<<vsub) * src->linesize[3];
521 for (jmax = FFMIN(-yp + dst_hp, src_hp); j < jmax; j++) {
523 d = dp + (xp+k) * dst_step;
527 for (kmax = FFMIN(-xp + dst_wp, src_wp); k < kmax; k++) {
528 int alpha_v, alpha_h, alpha;
530 // average alpha for color components, improve quality
531 if (hsub && vsub && j+1 < src_hp && k+1 < src_wp) {
532 alpha = (a[0] + a[src->linesize[3]] +
533 a[1] + a[src->linesize[3]+1]) >> 2;
534 } else if (hsub || vsub) {
535 alpha_h = hsub && k+1 < src_wp ?
536 (a[0] + a[1]) >> 1 : a[0];
537 alpha_v = vsub && j+1 < src_hp ?
538 (a[0] + a[src->linesize[3]]) >> 1 : a[0];
539 alpha = (alpha_v + alpha_h) >> 1;
542 // if the main channel has an alpha channel, alpha has to be calculated
543 // to create an un-premultiplied (straight) alpha value
544 if (main_has_alpha && alpha != 0 && alpha != 255) {
545 // average alpha for color components, improve quality
547 if (hsub && vsub && j+1 < src_hp && k+1 < src_wp) {
548 alpha_d = (d[0] + d[src->linesize[3]] +
549 d[1] + d[src->linesize[3]+1]) >> 2;
550 } else if (hsub || vsub) {
551 alpha_h = hsub && k+1 < src_wp ?
552 (d[0] + d[1]) >> 1 : d[0];
553 alpha_v = vsub && j+1 < src_hp ?
554 (d[0] + d[src->linesize[3]]) >> 1 : d[0];
555 alpha_d = (alpha_v + alpha_h) >> 1;
558 alpha = UNPREMULTIPLY_ALPHA(alpha, alpha_d);
560 *d = FAST_DIV255(*d * (255 - alpha) + *s * alpha);
565 dp += dst->linesize[dst_plane];
566 sp += src->linesize[i];
567 ap += (1 << vsub) * src->linesize[3];
571 static inline void alpha_composite(const AVFrame *src, const AVFrame *dst,
572 int src_w, int src_h,
573 int dst_w, int dst_h,
576 uint8_t alpha; ///< the amount of overlay to blend on to main
577 uint8_t *s, *sa, *d, *da;
578 int i, imax, j, jmax;
581 sa = src->data[3] + i * src->linesize[3];
582 da = dst->data[3] + (y+i) * dst->linesize[3];
584 for (imax = FFMIN(-y + dst_h, src_h); i < imax; i++) {
589 for (jmax = FFMIN(-x + dst_w, src_w); j < jmax; j++) {
591 if (alpha != 0 && alpha != 255) {
592 uint8_t alpha_d = *d;
593 alpha = UNPREMULTIPLY_ALPHA(alpha, alpha_d);
602 // apply alpha compositing: main_alpha += (1-main_alpha) * overlay_alpha
603 *d += FAST_DIV255((255 - *d) * *s);
608 da += dst->linesize[3];
609 sa += src->linesize[3];
613 static av_always_inline void blend_image_yuv(AVFilterContext *ctx,
614 AVFrame *dst, const AVFrame *src,
619 OverlayContext *s = ctx->priv;
620 const int src_w = src->width;
621 const int src_h = src->height;
622 const int dst_w = dst->width;
623 const int dst_h = dst->height;
626 alpha_composite(src, dst, src_w, src_h, dst_w, dst_h, x, y);
628 blend_plane(ctx, dst, src, src_w, src_h, dst_w, dst_h, 0, 0, 0, x, y, main_has_alpha,
629 s->main_desc->comp[0].plane, s->main_desc->comp[0].offset, s->main_desc->comp[0].step);
630 blend_plane(ctx, dst, src, src_w, src_h, dst_w, dst_h, 1, hsub, vsub, x, y, main_has_alpha,
631 s->main_desc->comp[1].plane, s->main_desc->comp[1].offset, s->main_desc->comp[1].step);
632 blend_plane(ctx, dst, src, src_w, src_h, dst_w, dst_h, 2, hsub, vsub, x, y, main_has_alpha,
633 s->main_desc->comp[2].plane, s->main_desc->comp[2].offset, s->main_desc->comp[2].step);
636 static av_always_inline void blend_image_planar_rgb(AVFilterContext *ctx,
637 AVFrame *dst, const AVFrame *src,
642 OverlayContext *s = ctx->priv;
643 const int src_w = src->width;
644 const int src_h = src->height;
645 const int dst_w = dst->width;
646 const int dst_h = dst->height;
649 alpha_composite(src, dst, src_w, src_h, dst_w, dst_h, x, y);
651 blend_plane(ctx, dst, src, src_w, src_h, dst_w, dst_h, 0, 0, 0, x, y, main_has_alpha,
652 s->main_desc->comp[1].plane, s->main_desc->comp[1].offset, s->main_desc->comp[1].step);
653 blend_plane(ctx, dst, src, src_w, src_h, dst_w, dst_h, 1, hsub, vsub, x, y, main_has_alpha,
654 s->main_desc->comp[2].plane, s->main_desc->comp[2].offset, s->main_desc->comp[2].step);
655 blend_plane(ctx, dst, src, src_w, src_h, dst_w, dst_h, 2, hsub, vsub, x, y, main_has_alpha,
656 s->main_desc->comp[0].plane, s->main_desc->comp[0].offset, s->main_desc->comp[0].step);
659 static void blend_image_yuv420(AVFilterContext *ctx, AVFrame *dst, const AVFrame *src, int x, int y)
661 blend_image_yuv(ctx, dst, src, 1, 1, 0, x, y);
664 static void blend_image_yuva420(AVFilterContext *ctx, AVFrame *dst, const AVFrame *src, int x, int y)
666 blend_image_yuv(ctx, dst, src, 1, 1, 1, x, y);
669 static void blend_image_yuv422(AVFilterContext *ctx, AVFrame *dst, const AVFrame *src, int x, int y)
671 blend_image_yuv(ctx, dst, src, 1, 0, 0, x, y);
674 static void blend_image_yuva422(AVFilterContext *ctx, AVFrame *dst, const AVFrame *src, int x, int y)
676 blend_image_yuv(ctx, dst, src, 1, 0, 1, x, y);
679 static void blend_image_yuv444(AVFilterContext *ctx, AVFrame *dst, const AVFrame *src, int x, int y)
681 blend_image_yuv(ctx, dst, src, 0, 0, 0, x, y);
684 static void blend_image_yuva444(AVFilterContext *ctx, AVFrame *dst, const AVFrame *src, int x, int y)
686 blend_image_yuv(ctx, dst, src, 0, 0, 1, x, y);
689 static void blend_image_gbrp(AVFilterContext *ctx, AVFrame *dst, const AVFrame *src, int x, int y)
691 blend_image_planar_rgb(ctx, dst, src, 0, 0, 0, x, y);
694 static void blend_image_gbrap(AVFilterContext *ctx, AVFrame *dst, const AVFrame *src, int x, int y)
696 blend_image_planar_rgb(ctx, dst, src, 0, 0, 1, x, y);
699 static void blend_image_rgb(AVFilterContext *ctx, AVFrame *dst, const AVFrame *src, int x, int y)
701 blend_image_packed_rgb(ctx, dst, src, 0, x, y);
704 static void blend_image_rgba(AVFilterContext *ctx, AVFrame *dst, const AVFrame *src, int x, int y)
706 blend_image_packed_rgb(ctx, dst, src, 1, x, y);
709 static int config_input_main(AVFilterLink *inlink)
711 OverlayContext *s = inlink->dst->priv;
712 const AVPixFmtDescriptor *pix_desc = av_pix_fmt_desc_get(inlink->format);
714 av_image_fill_max_pixsteps(s->main_pix_step, NULL, pix_desc);
716 s->hsub = pix_desc->log2_chroma_w;
717 s->vsub = pix_desc->log2_chroma_h;
719 s->main_desc = pix_desc;
721 s->main_is_packed_rgb =
722 ff_fill_rgba_map(s->main_rgba_map, inlink->format) >= 0;
723 s->main_has_alpha = ff_fmt_is_in(inlink->format, alpha_pix_fmts);
725 case OVERLAY_FORMAT_YUV420:
726 s->blend_image = s->main_has_alpha ? blend_image_yuva420 : blend_image_yuv420;
728 case OVERLAY_FORMAT_YUV422:
729 s->blend_image = s->main_has_alpha ? blend_image_yuva422 : blend_image_yuv422;
731 case OVERLAY_FORMAT_YUV444:
732 s->blend_image = s->main_has_alpha ? blend_image_yuva444 : blend_image_yuv444;
734 case OVERLAY_FORMAT_RGB:
735 s->blend_image = s->main_has_alpha ? blend_image_rgba : blend_image_rgb;
737 case OVERLAY_FORMAT_GBRP:
738 s->blend_image = s->main_has_alpha ? blend_image_gbrap : blend_image_gbrp;
740 case OVERLAY_FORMAT_AUTO:
741 switch (inlink->format) {
742 case AV_PIX_FMT_YUVA420P:
743 s->blend_image = blend_image_yuva420;
745 case AV_PIX_FMT_YUVA422P:
746 s->blend_image = blend_image_yuva422;
748 case AV_PIX_FMT_YUVA444P:
749 s->blend_image = blend_image_yuva444;
751 case AV_PIX_FMT_ARGB:
752 case AV_PIX_FMT_RGBA:
753 case AV_PIX_FMT_BGRA:
754 case AV_PIX_FMT_ABGR:
755 s->blend_image = blend_image_rgba;
757 case AV_PIX_FMT_GBRAP:
758 s->blend_image = blend_image_gbrap;
769 static AVFrame *do_blend(AVFilterContext *ctx, AVFrame *mainpic,
770 const AVFrame *second)
772 OverlayContext *s = ctx->priv;
773 AVFilterLink *inlink = ctx->inputs[0];
775 if (s->eval_mode == EVAL_MODE_FRAME) {
776 int64_t pos = mainpic->pkt_pos;
778 s->var_values[VAR_N] = inlink->frame_count_out;
779 s->var_values[VAR_T] = mainpic->pts == AV_NOPTS_VALUE ?
780 NAN : mainpic->pts * av_q2d(inlink->time_base);
781 s->var_values[VAR_POS] = pos == -1 ? NAN : pos;
783 s->var_values[VAR_OVERLAY_W] = s->var_values[VAR_OW] = second->width;
784 s->var_values[VAR_OVERLAY_H] = s->var_values[VAR_OH] = second->height;
785 s->var_values[VAR_MAIN_W ] = s->var_values[VAR_MW] = mainpic->width;
786 s->var_values[VAR_MAIN_H ] = s->var_values[VAR_MH] = mainpic->height;
789 av_log(ctx, AV_LOG_DEBUG, "n:%f t:%f pos:%f x:%f xi:%d y:%f yi:%d\n",
790 s->var_values[VAR_N], s->var_values[VAR_T], s->var_values[VAR_POS],
791 s->var_values[VAR_X], s->x,
792 s->var_values[VAR_Y], s->y);
795 if (s->x < mainpic->width && s->x + second->width >= 0 ||
796 s->y < mainpic->height && s->y + second->height >= 0)
797 s->blend_image(ctx, mainpic, second, s->x, s->y);
801 static int filter_frame(AVFilterLink *inlink, AVFrame *inpicref)
803 OverlayContext *s = inlink->dst->priv;
804 av_log(inlink->dst, AV_LOG_DEBUG, "Incoming frame (time:%s) from link #%d\n", av_ts2timestr(inpicref->pts, &inlink->time_base), FF_INLINK_IDX(inlink));
805 return ff_dualinput_filter_frame(&s->dinput, inlink, inpicref);
808 static int request_frame(AVFilterLink *outlink)
810 OverlayContext *s = outlink->src->priv;
811 return ff_dualinput_request_frame(&s->dinput, outlink);
814 static av_cold int init(AVFilterContext *ctx)
816 OverlayContext *s = ctx->priv;
818 if (!s->dinput.repeatlast || s->eof_action == EOF_ACTION_PASS) {
819 s->dinput.repeatlast = 0;
820 s->eof_action = EOF_ACTION_PASS;
822 if (s->dinput.shortest || s->eof_action == EOF_ACTION_ENDALL) {
823 s->dinput.shortest = 1;
824 s->eof_action = EOF_ACTION_ENDALL;
827 s->dinput.process = do_blend;
831 #define OFFSET(x) offsetof(OverlayContext, x)
832 #define FLAGS AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
834 static const AVOption overlay_options[] = {
835 { "x", "set the x expression", OFFSET(x_expr), AV_OPT_TYPE_STRING, {.str = "0"}, CHAR_MIN, CHAR_MAX, FLAGS },
836 { "y", "set the y expression", OFFSET(y_expr), AV_OPT_TYPE_STRING, {.str = "0"}, CHAR_MIN, CHAR_MAX, FLAGS },
837 { "eof_action", "Action to take when encountering EOF from secondary input ",
838 OFFSET(eof_action), AV_OPT_TYPE_INT, { .i64 = EOF_ACTION_REPEAT },
839 EOF_ACTION_REPEAT, EOF_ACTION_PASS, .flags = FLAGS, "eof_action" },
840 { "repeat", "Repeat the previous frame.", 0, AV_OPT_TYPE_CONST, { .i64 = EOF_ACTION_REPEAT }, .flags = FLAGS, "eof_action" },
841 { "endall", "End both streams.", 0, AV_OPT_TYPE_CONST, { .i64 = EOF_ACTION_ENDALL }, .flags = FLAGS, "eof_action" },
842 { "pass", "Pass through the main input.", 0, AV_OPT_TYPE_CONST, { .i64 = EOF_ACTION_PASS }, .flags = FLAGS, "eof_action" },
843 { "eval", "specify when to evaluate expressions", OFFSET(eval_mode), AV_OPT_TYPE_INT, {.i64 = EVAL_MODE_FRAME}, 0, EVAL_MODE_NB-1, FLAGS, "eval" },
844 { "init", "eval expressions once during initialization", 0, AV_OPT_TYPE_CONST, {.i64=EVAL_MODE_INIT}, .flags = FLAGS, .unit = "eval" },
845 { "frame", "eval expressions per-frame", 0, AV_OPT_TYPE_CONST, {.i64=EVAL_MODE_FRAME}, .flags = FLAGS, .unit = "eval" },
846 { "shortest", "force termination when the shortest input terminates", OFFSET(dinput.shortest), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, FLAGS },
847 { "format", "set output format", OFFSET(format), AV_OPT_TYPE_INT, {.i64=OVERLAY_FORMAT_YUV420}, 0, OVERLAY_FORMAT_NB-1, FLAGS, "format" },
848 { "yuv420", "", 0, AV_OPT_TYPE_CONST, {.i64=OVERLAY_FORMAT_YUV420}, .flags = FLAGS, .unit = "format" },
849 { "yuv422", "", 0, AV_OPT_TYPE_CONST, {.i64=OVERLAY_FORMAT_YUV422}, .flags = FLAGS, .unit = "format" },
850 { "yuv444", "", 0, AV_OPT_TYPE_CONST, {.i64=OVERLAY_FORMAT_YUV444}, .flags = FLAGS, .unit = "format" },
851 { "rgb", "", 0, AV_OPT_TYPE_CONST, {.i64=OVERLAY_FORMAT_RGB}, .flags = FLAGS, .unit = "format" },
852 { "gbrp", "", 0, AV_OPT_TYPE_CONST, {.i64=OVERLAY_FORMAT_GBRP}, .flags = FLAGS, .unit = "format" },
853 { "auto", "", 0, AV_OPT_TYPE_CONST, {.i64=OVERLAY_FORMAT_AUTO}, .flags = FLAGS, .unit = "format" },
854 { "repeatlast", "repeat overlay of the last overlay frame", OFFSET(dinput.repeatlast), AV_OPT_TYPE_BOOL, {.i64=1}, 0, 1, FLAGS },
858 AVFILTER_DEFINE_CLASS(overlay);
860 static const AVFilterPad avfilter_vf_overlay_inputs[] = {
863 .type = AVMEDIA_TYPE_VIDEO,
864 .config_props = config_input_main,
865 .filter_frame = filter_frame,
870 .type = AVMEDIA_TYPE_VIDEO,
871 .config_props = config_input_overlay,
872 .filter_frame = filter_frame,
877 static const AVFilterPad avfilter_vf_overlay_outputs[] = {
880 .type = AVMEDIA_TYPE_VIDEO,
881 .config_props = config_output,
882 .request_frame = request_frame,
887 AVFilter ff_vf_overlay = {
889 .description = NULL_IF_CONFIG_SMALL("Overlay a video source on top of the input."),
892 .priv_size = sizeof(OverlayContext),
893 .priv_class = &overlay_class,
894 .query_formats = query_formats,
895 .process_command = process_command,
896 .inputs = avfilter_vf_overlay_inputs,
897 .outputs = avfilter_vf_overlay_outputs,
898 .flags = AVFILTER_FLAG_SUPPORT_TIMELINE_INTERNAL,