2 * Copyright (c) 2007 Bobby Bingham
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
33 #include "scale_eval.h"
35 #include "libavutil/avstring.h"
36 #include "libavutil/eval.h"
37 #include "libavutil/internal.h"
38 #include "libavutil/mathematics.h"
39 #include "libavutil/opt.h"
40 #include "libavutil/parseutils.h"
41 #include "libavutil/pixdesc.h"
42 #include "libavutil/imgutils.h"
43 #include "libswscale/swscale.h"
45 static const char *const var_names[] = {
96 VAR_S2R_MAIN_DAR, VAR_S2R_MDAR,
111 typedef struct ScaleContext {
112 const AVClass *class;
113 struct SwsContext *sws; ///< software scaler context
114 struct SwsContext *isws[2]; ///< software scaler context for interlaced material
115 // context used for forwarding options to sws
116 struct SwsContext *sws_opts;
119 * New dimensions. Special values are:
120 * 0 = original width/height
121 * -1 = keep original aspect
122 * -N = try to keep aspect but make sure it is divisible by N
126 double param[2]; // sws params
128 int hsub, vsub; ///< chroma subsampling
129 int slice_y; ///< top of current output slice
130 int input_is_pal; ///< set to 1 if the input format is paletted
131 int output_is_pal; ///< set to 1 if the output format is paletted
134 char *w_expr; ///< width expression string
135 char *h_expr; ///< height expression string
138 double var_values[VARS_NB];
142 char *in_color_matrix;
143 char *out_color_matrix;
154 int force_original_aspect_ratio;
155 int force_divisible_by;
157 int eval_mode; ///< expression evaluation mode
161 const AVFilter ff_vf_scale2ref;
163 static int config_props(AVFilterLink *outlink);
165 static int check_exprs(AVFilterContext *ctx)
167 ScaleContext *scale = ctx->priv;
168 unsigned vars_w[VARS_NB] = { 0 }, vars_h[VARS_NB] = { 0 };
170 if (!scale->w_pexpr && !scale->h_pexpr)
171 return AVERROR(EINVAL);
174 av_expr_count_vars(scale->w_pexpr, vars_w, VARS_NB);
176 av_expr_count_vars(scale->h_pexpr, vars_h, VARS_NB);
178 if (vars_w[VAR_OUT_W] || vars_w[VAR_OW]) {
179 av_log(ctx, AV_LOG_ERROR, "Width expression cannot be self-referencing: '%s'.\n", scale->w_expr);
180 return AVERROR(EINVAL);
183 if (vars_h[VAR_OUT_H] || vars_h[VAR_OH]) {
184 av_log(ctx, AV_LOG_ERROR, "Height expression cannot be self-referencing: '%s'.\n", scale->h_expr);
185 return AVERROR(EINVAL);
188 if ((vars_w[VAR_OUT_H] || vars_w[VAR_OH]) &&
189 (vars_h[VAR_OUT_W] || vars_h[VAR_OW])) {
190 av_log(ctx, AV_LOG_WARNING, "Circular references detected for width '%s' and height '%s' - possibly invalid.\n", scale->w_expr, scale->h_expr);
193 if (ctx->filter != &ff_vf_scale2ref &&
194 (vars_w[VAR_S2R_MAIN_W] || vars_h[VAR_S2R_MAIN_W] ||
195 vars_w[VAR_S2R_MAIN_H] || vars_h[VAR_S2R_MAIN_H] ||
196 vars_w[VAR_S2R_MAIN_A] || vars_h[VAR_S2R_MAIN_A] ||
197 vars_w[VAR_S2R_MAIN_SAR] || vars_h[VAR_S2R_MAIN_SAR] ||
198 vars_w[VAR_S2R_MAIN_DAR] || vars_h[VAR_S2R_MAIN_DAR] ||
199 vars_w[VAR_S2R_MDAR] || vars_h[VAR_S2R_MDAR] ||
200 vars_w[VAR_S2R_MAIN_HSUB] || vars_h[VAR_S2R_MAIN_HSUB] ||
201 vars_w[VAR_S2R_MAIN_VSUB] || vars_h[VAR_S2R_MAIN_VSUB] ||
202 vars_w[VAR_S2R_MAIN_N] || vars_h[VAR_S2R_MAIN_N] ||
203 vars_w[VAR_S2R_MAIN_T] || vars_h[VAR_S2R_MAIN_T] ||
204 vars_w[VAR_S2R_MAIN_POS] || vars_h[VAR_S2R_MAIN_POS]) ) {
205 av_log(ctx, AV_LOG_ERROR, "Expressions with scale2ref variables are not valid in scale filter.\n");
206 return AVERROR(EINVAL);
209 if (scale->eval_mode == EVAL_MODE_INIT &&
210 (vars_w[VAR_N] || vars_h[VAR_N] ||
211 vars_w[VAR_T] || vars_h[VAR_T] ||
213 vars_w[VAR_POS] || vars_h[VAR_POS] ||
215 vars_w[VAR_S2R_MAIN_N] || vars_h[VAR_S2R_MAIN_N] ||
216 vars_w[VAR_S2R_MAIN_T] || vars_h[VAR_S2R_MAIN_T] ||
217 vars_w[VAR_S2R_MAIN_POS] || vars_h[VAR_S2R_MAIN_POS]) ) {
218 av_log(ctx, AV_LOG_ERROR, "Expressions with frame variables 'n', 't', 'pos' are not valid in init eval_mode.\n");
219 return AVERROR(EINVAL);
225 static int scale_parse_expr(AVFilterContext *ctx, char *str_expr, AVExpr **pexpr_ptr, const char *var, const char *args)
227 ScaleContext *scale = ctx->priv;
228 int ret, is_inited = 0;
229 char *old_str_expr = NULL;
230 AVExpr *old_pexpr = NULL;
233 old_str_expr = av_strdup(str_expr);
235 return AVERROR(ENOMEM);
236 av_opt_set(scale, var, args, 0);
240 old_pexpr = *pexpr_ptr;
245 ret = av_expr_parse(pexpr_ptr, args, var_names,
246 NULL, NULL, NULL, NULL, 0, ctx);
248 av_log(ctx, AV_LOG_ERROR, "Cannot parse expression for %s: '%s'\n", var, args);
252 ret = check_exprs(ctx);
256 if (is_inited && (ret = config_props(ctx->outputs[0])) < 0)
259 av_expr_free(old_pexpr);
261 av_freep(&old_str_expr);
266 av_expr_free(*pexpr_ptr);
269 av_opt_set(scale, var, old_str_expr, 0);
270 av_free(old_str_expr);
273 *pexpr_ptr = old_pexpr;
278 static av_cold int preinit(AVFilterContext *ctx)
280 ScaleContext *scale = ctx->priv;
283 scale->sws_opts = sws_alloc_context();
284 if (!scale->sws_opts)
285 return AVERROR(ENOMEM);
287 // set threads=0, so we can later check whether the user modified it
288 ret = av_opt_set_int(scale->sws_opts, "threads", 0, 0);
295 static av_cold int init(AVFilterContext *ctx)
297 ScaleContext *scale = ctx->priv;
301 if (scale->size_str && (scale->w_expr || scale->h_expr)) {
302 av_log(ctx, AV_LOG_ERROR,
303 "Size and width/height expressions cannot be set at the same time.\n");
304 return AVERROR(EINVAL);
307 if (scale->w_expr && !scale->h_expr)
308 FFSWAP(char *, scale->w_expr, scale->size_str);
310 if (scale->size_str) {
312 if ((ret = av_parse_video_size(&scale->w, &scale->h, scale->size_str)) < 0) {
313 av_log(ctx, AV_LOG_ERROR,
314 "Invalid size '%s'\n", scale->size_str);
317 snprintf(buf, sizeof(buf)-1, "%d", scale->w);
318 av_opt_set(scale, "w", buf, 0);
319 snprintf(buf, sizeof(buf)-1, "%d", scale->h);
320 av_opt_set(scale, "h", buf, 0);
323 av_opt_set(scale, "w", "iw", 0);
325 av_opt_set(scale, "h", "ih", 0);
327 ret = scale_parse_expr(ctx, NULL, &scale->w_pexpr, "width", scale->w_expr);
331 ret = scale_parse_expr(ctx, NULL, &scale->h_pexpr, "height", scale->h_expr);
335 av_log(ctx, AV_LOG_VERBOSE, "w:%s h:%s flags:'%s' interl:%d\n",
336 scale->w_expr, scale->h_expr, (char *)av_x_if_null(scale->flags_str, ""), scale->interlaced);
338 if (scale->flags_str && *scale->flags_str) {
339 ret = av_opt_set(scale->sws_opts, "sws_flags", scale->flags_str, 0);
344 for (int i = 0; i < FF_ARRAY_ELEMS(scale->param); i++)
345 if (scale->param[i] != DBL_MAX) {
346 ret = av_opt_set_double(scale->sws_opts, i ? "param1" : "param0",
352 // use generic thread-count if the user did not set it explicitly
353 ret = av_opt_get_int(scale->sws_opts, "threads", 0, &threads);
357 av_opt_set_int(scale->sws_opts, "threads", ff_filter_get_nb_threads(ctx), 0);
359 scale->in_frame_range = AVCOL_RANGE_UNSPECIFIED;
364 static av_cold void uninit(AVFilterContext *ctx)
366 ScaleContext *scale = ctx->priv;
367 av_expr_free(scale->w_pexpr);
368 av_expr_free(scale->h_pexpr);
369 scale->w_pexpr = scale->h_pexpr = NULL;
370 sws_freeContext(scale->sws_opts);
371 sws_freeContext(scale->sws);
372 sws_freeContext(scale->isws[0]);
373 sws_freeContext(scale->isws[1]);
377 static int query_formats(AVFilterContext *ctx)
379 AVFilterFormats *formats;
380 const AVPixFmtDescriptor *desc;
381 enum AVPixelFormat pix_fmt;
386 while ((desc = av_pix_fmt_desc_next(desc))) {
387 pix_fmt = av_pix_fmt_desc_get_id(desc);
388 if ((sws_isSupportedInput(pix_fmt) ||
389 sws_isSupportedEndiannessConversion(pix_fmt))
390 && (ret = ff_add_format(&formats, pix_fmt)) < 0) {
394 if ((ret = ff_formats_ref(formats, &ctx->inputs[0]->outcfg.formats)) < 0)
399 while ((desc = av_pix_fmt_desc_next(desc))) {
400 pix_fmt = av_pix_fmt_desc_get_id(desc);
401 if ((sws_isSupportedOutput(pix_fmt) || pix_fmt == AV_PIX_FMT_PAL8 ||
402 sws_isSupportedEndiannessConversion(pix_fmt))
403 && (ret = ff_add_format(&formats, pix_fmt)) < 0) {
407 if ((ret = ff_formats_ref(formats, &ctx->outputs[0]->incfg.formats)) < 0)
413 static const int *parse_yuv_type(const char *s, enum AVColorSpace colorspace)
418 if (s && strstr(s, "bt709")) {
419 colorspace = AVCOL_SPC_BT709;
420 } else if (s && strstr(s, "fcc")) {
421 colorspace = AVCOL_SPC_FCC;
422 } else if (s && strstr(s, "smpte240m")) {
423 colorspace = AVCOL_SPC_SMPTE240M;
424 } else if (s && (strstr(s, "bt601") || strstr(s, "bt470") || strstr(s, "smpte170m"))) {
425 colorspace = AVCOL_SPC_BT470BG;
426 } else if (s && strstr(s, "bt2020")) {
427 colorspace = AVCOL_SPC_BT2020_NCL;
430 if (colorspace < 1 || colorspace > 10 || colorspace == 8) {
431 colorspace = AVCOL_SPC_BT470BG;
434 return sws_getCoefficients(colorspace);
437 static int scale_eval_dimensions(AVFilterContext *ctx)
439 ScaleContext *scale = ctx->priv;
440 const char scale2ref = ctx->filter == &ff_vf_scale2ref;
441 const AVFilterLink *inlink = scale2ref ? ctx->inputs[1] : ctx->inputs[0];
442 const AVFilterLink *outlink = ctx->outputs[0];
443 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(inlink->format);
444 const AVPixFmtDescriptor *out_desc = av_pix_fmt_desc_get(outlink->format);
449 const AVPixFmtDescriptor *main_desc;
450 const AVFilterLink *main_link;
453 main_link = ctx->inputs[0];
454 main_desc = av_pix_fmt_desc_get(main_link->format);
457 scale->var_values[VAR_IN_W] = scale->var_values[VAR_IW] = inlink->w;
458 scale->var_values[VAR_IN_H] = scale->var_values[VAR_IH] = inlink->h;
459 scale->var_values[VAR_OUT_W] = scale->var_values[VAR_OW] = NAN;
460 scale->var_values[VAR_OUT_H] = scale->var_values[VAR_OH] = NAN;
461 scale->var_values[VAR_A] = (double) inlink->w / inlink->h;
462 scale->var_values[VAR_SAR] = inlink->sample_aspect_ratio.num ?
463 (double) inlink->sample_aspect_ratio.num / inlink->sample_aspect_ratio.den : 1;
464 scale->var_values[VAR_DAR] = scale->var_values[VAR_A] * scale->var_values[VAR_SAR];
465 scale->var_values[VAR_HSUB] = 1 << desc->log2_chroma_w;
466 scale->var_values[VAR_VSUB] = 1 << desc->log2_chroma_h;
467 scale->var_values[VAR_OHSUB] = 1 << out_desc->log2_chroma_w;
468 scale->var_values[VAR_OVSUB] = 1 << out_desc->log2_chroma_h;
471 scale->var_values[VAR_S2R_MAIN_W] = main_link->w;
472 scale->var_values[VAR_S2R_MAIN_H] = main_link->h;
473 scale->var_values[VAR_S2R_MAIN_A] = (double) main_link->w / main_link->h;
474 scale->var_values[VAR_S2R_MAIN_SAR] = main_link->sample_aspect_ratio.num ?
475 (double) main_link->sample_aspect_ratio.num / main_link->sample_aspect_ratio.den : 1;
476 scale->var_values[VAR_S2R_MAIN_DAR] = scale->var_values[VAR_S2R_MDAR] =
477 scale->var_values[VAR_S2R_MAIN_A] * scale->var_values[VAR_S2R_MAIN_SAR];
478 scale->var_values[VAR_S2R_MAIN_HSUB] = 1 << main_desc->log2_chroma_w;
479 scale->var_values[VAR_S2R_MAIN_VSUB] = 1 << main_desc->log2_chroma_h;
482 res = av_expr_eval(scale->w_pexpr, scale->var_values, NULL);
483 eval_w = scale->var_values[VAR_OUT_W] = scale->var_values[VAR_OW] = (int) res == 0 ? inlink->w : (int) res;
485 res = av_expr_eval(scale->h_pexpr, scale->var_values, NULL);
487 expr = scale->h_expr;
488 ret = AVERROR(EINVAL);
491 eval_h = scale->var_values[VAR_OUT_H] = scale->var_values[VAR_OH] = (int) res == 0 ? inlink->h : (int) res;
493 res = av_expr_eval(scale->w_pexpr, scale->var_values, NULL);
495 expr = scale->w_expr;
496 ret = AVERROR(EINVAL);
499 eval_w = scale->var_values[VAR_OUT_W] = scale->var_values[VAR_OW] = (int) res == 0 ? inlink->w : (int) res;
507 av_log(ctx, AV_LOG_ERROR,
508 "Error when evaluating the expression '%s'.\n", expr);
512 static int config_props(AVFilterLink *outlink)
514 AVFilterContext *ctx = outlink->src;
515 AVFilterLink *inlink0 = outlink->src->inputs[0];
516 AVFilterLink *inlink = ctx->filter == &ff_vf_scale2ref ?
517 outlink->src->inputs[1] :
518 outlink->src->inputs[0];
519 enum AVPixelFormat outfmt = outlink->format;
520 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(inlink->format);
521 const AVPixFmtDescriptor *outdesc = av_pix_fmt_desc_get(outfmt);
522 ScaleContext *scale = ctx->priv;
523 uint8_t *flags_val = NULL;
526 if ((ret = scale_eval_dimensions(ctx)) < 0)
529 outlink->w = scale->w;
530 outlink->h = scale->h;
532 ff_scale_adjust_dimensions(inlink, &outlink->w, &outlink->h,
533 scale->force_original_aspect_ratio,
534 scale->force_divisible_by);
536 if (outlink->w > INT_MAX ||
537 outlink->h > INT_MAX ||
538 (outlink->h * inlink->w) > INT_MAX ||
539 (outlink->w * inlink->h) > INT_MAX)
540 av_log(ctx, AV_LOG_ERROR, "Rescaled value for width or height is too big.\n");
542 /* TODO: make algorithm configurable */
544 scale->input_is_pal = desc->flags & AV_PIX_FMT_FLAG_PAL;
545 if (outfmt == AV_PIX_FMT_PAL8) outfmt = AV_PIX_FMT_BGR8;
546 scale->output_is_pal = av_pix_fmt_desc_get(outfmt)->flags & AV_PIX_FMT_FLAG_PAL;
549 sws_freeContext(scale->sws);
551 sws_freeContext(scale->isws[0]);
553 sws_freeContext(scale->isws[1]);
554 scale->isws[0] = scale->isws[1] = scale->sws = NULL;
555 if (inlink0->w == outlink->w &&
556 inlink0->h == outlink->h &&
557 !scale->out_color_matrix &&
558 scale->in_range == scale->out_range &&
559 inlink0->format == outlink->format)
562 struct SwsContext **swscs[3] = {&scale->sws, &scale->isws[0], &scale->isws[1]};
565 for (i = 0; i < 3; i++) {
566 int in_v_chr_pos = scale->in_v_chr_pos, out_v_chr_pos = scale->out_v_chr_pos;
567 struct SwsContext *const s = sws_alloc_context();
569 return AVERROR(ENOMEM);
572 ret = av_opt_copy(s, scale->sws_opts);
576 av_opt_set_int(s, "srcw", inlink0 ->w, 0);
577 av_opt_set_int(s, "srch", inlink0 ->h >> !!i, 0);
578 av_opt_set_int(s, "src_format", inlink0->format, 0);
579 av_opt_set_int(s, "dstw", outlink->w, 0);
580 av_opt_set_int(s, "dsth", outlink->h >> !!i, 0);
581 av_opt_set_int(s, "dst_format", outfmt, 0);
582 if (scale->in_range != AVCOL_RANGE_UNSPECIFIED)
583 av_opt_set_int(s, "src_range",
584 scale->in_range == AVCOL_RANGE_JPEG, 0);
585 else if (scale->in_frame_range != AVCOL_RANGE_UNSPECIFIED)
586 av_opt_set_int(s, "src_range",
587 scale->in_frame_range == AVCOL_RANGE_JPEG, 0);
588 if (scale->out_range != AVCOL_RANGE_UNSPECIFIED)
589 av_opt_set_int(s, "dst_range",
590 scale->out_range == AVCOL_RANGE_JPEG, 0);
592 /* Override chroma location default settings to have the correct
593 * chroma positions. MPEG chroma positions are used by convention.
594 * Note that this works for both MPEG-1/JPEG and MPEG-2/4 chroma
595 * locations, since they share a vertical alignment */
596 if (desc->log2_chroma_h == 1 && scale->in_v_chr_pos == -513) {
597 in_v_chr_pos = (i == 0) ? 128 : (i == 1) ? 64 : 192;
600 if (outdesc->log2_chroma_h == 1 && scale->out_v_chr_pos == -513) {
601 out_v_chr_pos = (i == 0) ? 128 : (i == 1) ? 64 : 192;
604 av_opt_set_int(s, "src_h_chr_pos", scale->in_h_chr_pos, 0);
605 av_opt_set_int(s, "src_v_chr_pos", in_v_chr_pos, 0);
606 av_opt_set_int(s, "dst_h_chr_pos", scale->out_h_chr_pos, 0);
607 av_opt_set_int(s, "dst_v_chr_pos", out_v_chr_pos, 0);
609 if ((ret = sws_init_context(s, NULL, NULL)) < 0)
611 if (!scale->interlaced)
616 if (inlink0->sample_aspect_ratio.num){
617 outlink->sample_aspect_ratio = av_mul_q((AVRational){outlink->h * inlink0->w, outlink->w * inlink0->h}, inlink0->sample_aspect_ratio);
619 outlink->sample_aspect_ratio = inlink0->sample_aspect_ratio;
622 av_opt_get(scale->sws, "sws_flags", 0, &flags_val);
624 av_log(ctx, AV_LOG_VERBOSE, "w:%d h:%d fmt:%s sar:%d/%d -> w:%d h:%d fmt:%s sar:%d/%d flags:%s\n",
625 inlink ->w, inlink ->h, av_get_pix_fmt_name( inlink->format),
626 inlink->sample_aspect_ratio.num, inlink->sample_aspect_ratio.den,
627 outlink->w, outlink->h, av_get_pix_fmt_name(outlink->format),
628 outlink->sample_aspect_ratio.num, outlink->sample_aspect_ratio.den,
630 av_freep(&flags_val);
638 static int config_props_ref(AVFilterLink *outlink)
640 AVFilterLink *inlink = outlink->src->inputs[1];
642 outlink->w = inlink->w;
643 outlink->h = inlink->h;
644 outlink->sample_aspect_ratio = inlink->sample_aspect_ratio;
645 outlink->time_base = inlink->time_base;
646 outlink->frame_rate = inlink->frame_rate;
651 static int request_frame(AVFilterLink *outlink)
653 return ff_request_frame(outlink->src->inputs[0]);
656 static int request_frame_ref(AVFilterLink *outlink)
658 return ff_request_frame(outlink->src->inputs[1]);
661 static void frame_offset(AVFrame *frame, int dir, int is_pal)
663 for (int i = 0; i < 4 && frame->data[i]; i++) {
664 if (i == 1 && is_pal)
666 frame->data[i] += frame->linesize[i] * dir;
670 static int scale_field(ScaleContext *scale, AVFrame *dst, AVFrame *src,
673 int orig_h_src = src->height;
674 int orig_h_dst = dst->height;
677 // offset the data pointers for the bottom field
679 frame_offset(src, 1, scale->input_is_pal);
680 frame_offset(dst, 1, scale->output_is_pal);
683 // take every second line
684 for (int i = 0; i < 4; i++) {
685 src->linesize[i] *= 2;
686 dst->linesize[i] *= 2;
691 ret = sws_scale_frame(scale->isws[field], dst, src);
695 // undo the changes we made above
696 for (int i = 0; i < 4; i++) {
697 src->linesize[i] /= 2;
698 dst->linesize[i] /= 2;
700 src->height = orig_h_src;
701 dst->height = orig_h_dst;
704 frame_offset(src, -1, scale->input_is_pal);
705 frame_offset(dst, -1, scale->output_is_pal);
711 static int scale_frame(AVFilterLink *link, AVFrame *in, AVFrame **frame_out)
713 AVFilterContext *ctx = link->dst;
714 ScaleContext *scale = ctx->priv;
715 AVFilterLink *outlink = ctx->outputs[0];
717 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(link->format);
724 if (in->colorspace == AVCOL_SPC_YCGCO)
725 av_log(link->dst, AV_LOG_WARNING, "Detected unsupported YCgCo colorspace.\n");
727 frame_changed = in->width != link->w ||
728 in->height != link->h ||
729 in->format != link->format ||
730 in->sample_aspect_ratio.den != link->sample_aspect_ratio.den ||
731 in->sample_aspect_ratio.num != link->sample_aspect_ratio.num;
733 if (in->color_range != AVCOL_RANGE_UNSPECIFIED &&
734 scale->in_range == AVCOL_RANGE_UNSPECIFIED &&
735 in->color_range != scale->in_frame_range) {
736 scale->in_frame_range = in->color_range;
740 if (scale->eval_mode == EVAL_MODE_FRAME || frame_changed) {
741 unsigned vars_w[VARS_NB] = { 0 }, vars_h[VARS_NB] = { 0 };
743 av_expr_count_vars(scale->w_pexpr, vars_w, VARS_NB);
744 av_expr_count_vars(scale->h_pexpr, vars_h, VARS_NB);
746 if (scale->eval_mode == EVAL_MODE_FRAME &&
748 ctx->filter != &ff_vf_scale2ref &&
749 !(vars_w[VAR_N] || vars_w[VAR_T]
754 !(vars_h[VAR_N] || vars_h[VAR_T]
759 scale->w && scale->h)
762 if (scale->eval_mode == EVAL_MODE_INIT) {
763 snprintf(buf, sizeof(buf) - 1, "%d", scale->w);
764 av_opt_set(scale, "w", buf, 0);
765 snprintf(buf, sizeof(buf) - 1, "%d", scale->h);
766 av_opt_set(scale, "h", buf, 0);
768 ret = scale_parse_expr(ctx, NULL, &scale->w_pexpr, "width", scale->w_expr);
772 ret = scale_parse_expr(ctx, NULL, &scale->h_pexpr, "height", scale->h_expr);
777 if (ctx->filter == &ff_vf_scale2ref) {
778 scale->var_values[VAR_S2R_MAIN_N] = link->frame_count_out;
779 scale->var_values[VAR_S2R_MAIN_T] = TS2T(in->pts, link->time_base);
781 FF_DISABLE_DEPRECATION_WARNINGS
782 scale->var_values[VAR_S2R_MAIN_POS] = in->pkt_pos == -1 ? NAN : in->pkt_pos;
783 FF_ENABLE_DEPRECATION_WARNINGS
786 scale->var_values[VAR_N] = link->frame_count_out;
787 scale->var_values[VAR_T] = TS2T(in->pts, link->time_base);
789 FF_DISABLE_DEPRECATION_WARNINGS
790 scale->var_values[VAR_POS] = in->pkt_pos == -1 ? NAN : in->pkt_pos;
791 FF_ENABLE_DEPRECATION_WARNINGS
795 link->dst->inputs[0]->format = in->format;
796 link->dst->inputs[0]->w = in->width;
797 link->dst->inputs[0]->h = in->height;
799 link->dst->inputs[0]->sample_aspect_ratio.den = in->sample_aspect_ratio.den;
800 link->dst->inputs[0]->sample_aspect_ratio.num = in->sample_aspect_ratio.num;
802 if ((ret = config_props(outlink)) < 0)
812 scale->hsub = desc->log2_chroma_w;
813 scale->vsub = desc->log2_chroma_h;
815 out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
818 return AVERROR(ENOMEM);
822 av_frame_copy_props(out, in);
823 out->width = outlink->w;
824 out->height = outlink->h;
827 // 1. If the output is RGB, set the matrix coefficients to RGB.
828 // 2. If the output is not RGB and we've got the RGB/XYZ (identity)
829 // matrix configured, unset the matrix.
830 // In theory these should be in swscale itself as the AVFrame
831 // based API gets in, so that not every swscale API user has
832 // to go through duplicating such sanity checks.
833 if (av_pix_fmt_desc_get(out->format)->flags & AV_PIX_FMT_FLAG_RGB)
834 out->colorspace = AVCOL_SPC_RGB;
835 else if (out->colorspace == AVCOL_SPC_RGB)
836 out->colorspace = AVCOL_SPC_UNSPECIFIED;
838 if (scale->output_is_pal)
839 avpriv_set_systematic_pal2((uint32_t*)out->data[1], outlink->format == AV_PIX_FMT_PAL8 ? AV_PIX_FMT_BGR8 : outlink->format);
841 in_range = in->color_range;
843 if ( scale->in_color_matrix
844 || scale->out_color_matrix
845 || scale-> in_range != AVCOL_RANGE_UNSPECIFIED
846 || in_range != AVCOL_RANGE_UNSPECIFIED
847 || scale->out_range != AVCOL_RANGE_UNSPECIFIED) {
848 int in_full, out_full, brightness, contrast, saturation;
849 const int *inv_table, *table;
851 sws_getColorspaceDetails(scale->sws, (int **)&inv_table, &in_full,
852 (int **)&table, &out_full,
853 &brightness, &contrast, &saturation);
855 if (scale->in_color_matrix)
856 inv_table = parse_yuv_type(scale->in_color_matrix, in->colorspace);
857 if (scale->out_color_matrix)
858 table = parse_yuv_type(scale->out_color_matrix, AVCOL_SPC_UNSPECIFIED);
859 else if (scale->in_color_matrix)
862 if (scale-> in_range != AVCOL_RANGE_UNSPECIFIED)
863 in_full = (scale-> in_range == AVCOL_RANGE_JPEG);
864 else if (in_range != AVCOL_RANGE_UNSPECIFIED)
865 in_full = (in_range == AVCOL_RANGE_JPEG);
866 if (scale->out_range != AVCOL_RANGE_UNSPECIFIED)
867 out_full = (scale->out_range == AVCOL_RANGE_JPEG);
869 sws_setColorspaceDetails(scale->sws, inv_table, in_full,
871 brightness, contrast, saturation);
873 sws_setColorspaceDetails(scale->isws[0], inv_table, in_full,
875 brightness, contrast, saturation);
877 sws_setColorspaceDetails(scale->isws[1], inv_table, in_full,
879 brightness, contrast, saturation);
881 out->color_range = out_full ? AVCOL_RANGE_JPEG : AVCOL_RANGE_MPEG;
884 av_reduce(&out->sample_aspect_ratio.num, &out->sample_aspect_ratio.den,
885 (int64_t)in->sample_aspect_ratio.num * outlink->h * link->w,
886 (int64_t)in->sample_aspect_ratio.den * outlink->w * link->h,
889 if (scale->interlaced>0 || (scale->interlaced<0 &&
890 (in->flags & AV_FRAME_FLAG_INTERLACED))) {
891 ret = scale_field(scale, out, in, 0);
893 ret = scale_field(scale, out, in, 1);
895 ret = sws_scale_frame(scale->sws, out, in);
900 av_frame_free(frame_out);
904 static int filter_frame(AVFilterLink *link, AVFrame *in)
906 AVFilterContext *ctx = link->dst;
907 AVFilterLink *outlink = ctx->outputs[0];
911 ret = scale_frame(link, in, &out);
913 return ff_filter_frame(outlink, out);
918 static int filter_frame_ref(AVFilterLink *link, AVFrame *in)
920 ScaleContext *scale = link->dst->priv;
921 AVFilterLink *outlink = link->dst->outputs[1];
924 frame_changed = in->width != link->w ||
925 in->height != link->h ||
926 in->format != link->format ||
927 in->sample_aspect_ratio.den != link->sample_aspect_ratio.den ||
928 in->sample_aspect_ratio.num != link->sample_aspect_ratio.num;
931 link->format = in->format;
933 link->h = in->height;
934 link->sample_aspect_ratio.num = in->sample_aspect_ratio.num;
935 link->sample_aspect_ratio.den = in->sample_aspect_ratio.den;
937 config_props_ref(outlink);
940 if (scale->eval_mode == EVAL_MODE_FRAME) {
941 scale->var_values[VAR_N] = link->frame_count_out;
942 scale->var_values[VAR_T] = TS2T(in->pts, link->time_base);
944 FF_DISABLE_DEPRECATION_WARNINGS
945 scale->var_values[VAR_POS] = in->pkt_pos == -1 ? NAN : in->pkt_pos;
946 FF_ENABLE_DEPRECATION_WARNINGS
950 return ff_filter_frame(outlink, in);
953 static int process_command(AVFilterContext *ctx, const char *cmd, const char *args,
954 char *res, int res_len, int flags)
956 ScaleContext *scale = ctx->priv;
961 w = !strcmp(cmd, "width") || !strcmp(cmd, "w");
962 h = !strcmp(cmd, "height") || !strcmp(cmd, "h");
965 str_expr = w ? scale->w_expr : scale->h_expr;
966 pexpr_ptr = w ? &scale->w_pexpr : &scale->h_pexpr;
968 ret = scale_parse_expr(ctx, str_expr, pexpr_ptr, cmd, args);
970 ret = AVERROR(ENOSYS);
973 av_log(ctx, AV_LOG_ERROR, "Failed to process command. Continuing with existing parameters.\n");
978 static const AVClass *child_class_iterate(void **iter)
980 const AVClass *c = *iter ? NULL : sws_get_class();
981 *iter = (void*)(uintptr_t)c;
985 static void *child_next(void *obj, void *prev)
987 ScaleContext *s = obj;
993 #define OFFSET(x) offsetof(ScaleContext, x)
994 #define FLAGS AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
995 #define TFLAGS AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_RUNTIME_PARAM
997 static const AVOption scale_options[] = {
998 { "w", "Output video width", OFFSET(w_expr), AV_OPT_TYPE_STRING, .flags = TFLAGS },
999 { "width", "Output video width", OFFSET(w_expr), AV_OPT_TYPE_STRING, .flags = TFLAGS },
1000 { "h", "Output video height", OFFSET(h_expr), AV_OPT_TYPE_STRING, .flags = TFLAGS },
1001 { "height","Output video height", OFFSET(h_expr), AV_OPT_TYPE_STRING, .flags = TFLAGS },
1002 { "flags", "Flags to pass to libswscale", OFFSET(flags_str), AV_OPT_TYPE_STRING, { .str = "" }, .flags = FLAGS },
1003 { "interl", "set interlacing", OFFSET(interlaced), AV_OPT_TYPE_BOOL, {.i64 = 0 }, -1, 1, FLAGS },
1004 { "size", "set video size", OFFSET(size_str), AV_OPT_TYPE_STRING, {.str = NULL}, 0, FLAGS },
1005 { "s", "set video size", OFFSET(size_str), AV_OPT_TYPE_STRING, {.str = NULL}, 0, FLAGS },
1006 { "in_color_matrix", "set input YCbCr type", OFFSET(in_color_matrix), AV_OPT_TYPE_STRING, { .str = "auto" }, .flags = FLAGS, "color" },
1007 { "out_color_matrix", "set output YCbCr type", OFFSET(out_color_matrix), AV_OPT_TYPE_STRING, { .str = NULL }, .flags = FLAGS, "color"},
1008 { "auto", NULL, 0, AV_OPT_TYPE_CONST, { .str = "auto" }, 0, 0, FLAGS, "color" },
1009 { "bt601", NULL, 0, AV_OPT_TYPE_CONST, { .str = "bt601" }, 0, 0, FLAGS, "color" },
1010 { "bt470", NULL, 0, AV_OPT_TYPE_CONST, { .str = "bt470" }, 0, 0, FLAGS, "color" },
1011 { "smpte170m", NULL, 0, AV_OPT_TYPE_CONST, { .str = "smpte170m" }, 0, 0, FLAGS, "color" },
1012 { "bt709", NULL, 0, AV_OPT_TYPE_CONST, { .str = "bt709" }, 0, 0, FLAGS, "color" },
1013 { "fcc", NULL, 0, AV_OPT_TYPE_CONST, { .str = "fcc" }, 0, 0, FLAGS, "color" },
1014 { "smpte240m", NULL, 0, AV_OPT_TYPE_CONST, { .str = "smpte240m" }, 0, 0, FLAGS, "color" },
1015 { "bt2020", NULL, 0, AV_OPT_TYPE_CONST, { .str = "bt2020" }, 0, 0, FLAGS, "color" },
1016 { "in_range", "set input color range", OFFSET( in_range), AV_OPT_TYPE_INT, {.i64 = AVCOL_RANGE_UNSPECIFIED }, 0, 2, FLAGS, "range" },
1017 { "out_range", "set output color range", OFFSET(out_range), AV_OPT_TYPE_INT, {.i64 = AVCOL_RANGE_UNSPECIFIED }, 0, 2, FLAGS, "range" },
1018 { "auto", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = AVCOL_RANGE_UNSPECIFIED }, 0, 0, FLAGS, "range" },
1019 { "unknown", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = AVCOL_RANGE_UNSPECIFIED }, 0, 0, FLAGS, "range" },
1020 { "full", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = AVCOL_RANGE_JPEG}, 0, 0, FLAGS, "range" },
1021 { "limited",NULL, 0, AV_OPT_TYPE_CONST, {.i64 = AVCOL_RANGE_MPEG}, 0, 0, FLAGS, "range" },
1022 { "jpeg", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = AVCOL_RANGE_JPEG}, 0, 0, FLAGS, "range" },
1023 { "mpeg", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = AVCOL_RANGE_MPEG}, 0, 0, FLAGS, "range" },
1024 { "tv", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = AVCOL_RANGE_MPEG}, 0, 0, FLAGS, "range" },
1025 { "pc", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = AVCOL_RANGE_JPEG}, 0, 0, FLAGS, "range" },
1026 { "in_v_chr_pos", "input vertical chroma position in luma grid/256" , OFFSET(in_v_chr_pos), AV_OPT_TYPE_INT, { .i64 = -513}, -513, 512, FLAGS },
1027 { "in_h_chr_pos", "input horizontal chroma position in luma grid/256", OFFSET(in_h_chr_pos), AV_OPT_TYPE_INT, { .i64 = -513}, -513, 512, FLAGS },
1028 { "out_v_chr_pos", "output vertical chroma position in luma grid/256" , OFFSET(out_v_chr_pos), AV_OPT_TYPE_INT, { .i64 = -513}, -513, 512, FLAGS },
1029 { "out_h_chr_pos", "output horizontal chroma position in luma grid/256", OFFSET(out_h_chr_pos), AV_OPT_TYPE_INT, { .i64 = -513}, -513, 512, FLAGS },
1030 { "force_original_aspect_ratio", "decrease or increase w/h if necessary to keep the original AR", OFFSET(force_original_aspect_ratio), AV_OPT_TYPE_INT, { .i64 = 0}, 0, 2, FLAGS, "force_oar" },
1031 { "disable", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 0 }, 0, 0, FLAGS, "force_oar" },
1032 { "decrease", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 1 }, 0, 0, FLAGS, "force_oar" },
1033 { "increase", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 2 }, 0, 0, FLAGS, "force_oar" },
1034 { "force_divisible_by", "enforce that the output resolution is divisible by a defined integer when force_original_aspect_ratio is used", OFFSET(force_divisible_by), AV_OPT_TYPE_INT, { .i64 = 1}, 1, 256, FLAGS },
1035 { "param0", "Scaler param 0", OFFSET(param[0]), AV_OPT_TYPE_DOUBLE, { .dbl = DBL_MAX }, -DBL_MAX, DBL_MAX, FLAGS },
1036 { "param1", "Scaler param 1", OFFSET(param[1]), AV_OPT_TYPE_DOUBLE, { .dbl = DBL_MAX }, -DBL_MAX, DBL_MAX, FLAGS },
1037 { "eval", "specify when to evaluate expressions", OFFSET(eval_mode), AV_OPT_TYPE_INT, {.i64 = EVAL_MODE_INIT}, 0, EVAL_MODE_NB-1, FLAGS, "eval" },
1038 { "init", "eval expressions once during initialization", 0, AV_OPT_TYPE_CONST, {.i64=EVAL_MODE_INIT}, .flags = FLAGS, .unit = "eval" },
1039 { "frame", "eval expressions during initialization and per-frame", 0, AV_OPT_TYPE_CONST, {.i64=EVAL_MODE_FRAME}, .flags = FLAGS, .unit = "eval" },
1043 static const AVClass scale_class = {
1044 .class_name = "scale(2ref)",
1045 .item_name = av_default_item_name,
1046 .option = scale_options,
1047 .version = LIBAVUTIL_VERSION_INT,
1048 .category = AV_CLASS_CATEGORY_FILTER,
1049 .child_class_iterate = child_class_iterate,
1050 .child_next = child_next,
1053 static const AVFilterPad avfilter_vf_scale_inputs[] = {
1056 .type = AVMEDIA_TYPE_VIDEO,
1057 .filter_frame = filter_frame,
1061 static const AVFilterPad avfilter_vf_scale_outputs[] = {
1064 .type = AVMEDIA_TYPE_VIDEO,
1065 .config_props = config_props,
1069 const AVFilter ff_vf_scale = {
1071 .description = NULL_IF_CONFIG_SMALL("Scale the input video size and/or convert the image format."),
1075 .priv_size = sizeof(ScaleContext),
1076 .priv_class = &scale_class,
1077 FILTER_INPUTS(avfilter_vf_scale_inputs),
1078 FILTER_OUTPUTS(avfilter_vf_scale_outputs),
1079 FILTER_QUERY_FUNC(query_formats),
1080 .process_command = process_command,
1083 static const AVFilterPad avfilter_vf_scale2ref_inputs[] = {
1086 .type = AVMEDIA_TYPE_VIDEO,
1087 .filter_frame = filter_frame,
1091 .type = AVMEDIA_TYPE_VIDEO,
1092 .filter_frame = filter_frame_ref,
1096 static const AVFilterPad avfilter_vf_scale2ref_outputs[] = {
1099 .type = AVMEDIA_TYPE_VIDEO,
1100 .config_props = config_props,
1101 .request_frame= request_frame,
1105 .type = AVMEDIA_TYPE_VIDEO,
1106 .config_props = config_props_ref,
1107 .request_frame= request_frame_ref,
1111 const AVFilter ff_vf_scale2ref = {
1112 .name = "scale2ref",
1113 .description = NULL_IF_CONFIG_SMALL("Scale the input video size and/or convert the image format to the given reference."),
1117 .priv_size = sizeof(ScaleContext),
1118 .priv_class = &scale_class,
1119 FILTER_INPUTS(avfilter_vf_scale2ref_inputs),
1120 FILTER_OUTPUTS(avfilter_vf_scale2ref_outputs),
1121 FILTER_QUERY_FUNC(query_formats),
1122 .process_command = process_command,