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 "libavutil/avstring.h"
34 #include "libavutil/eval.h"
35 #include "libavutil/internal.h"
36 #include "libavutil/mathematics.h"
37 #include "libavutil/opt.h"
38 #include "libavutil/parseutils.h"
39 #include "libavutil/pixdesc.h"
40 #include "libavutil/imgutils.h"
41 #include "libavutil/avassert.h"
42 #include "libswscale/swscale.h"
44 static const char *const var_names[] = {
76 struct SwsContext *sws; ///< software scaler context
77 struct SwsContext *isws[2]; ///< software scaler context for interlaced material
81 * New dimensions. Special values are:
82 * 0 = original width/height
83 * -1 = keep original aspect
87 unsigned int flags; ///sws flags
89 int hsub, vsub; ///< chroma subsampling
90 int slice_y; ///< top of current output slice
91 int input_is_pal; ///< set to 1 if the input format is paletted
92 int output_is_pal; ///< set to 1 if the output format is paletted
95 char *w_expr; ///< width expression string
96 char *h_expr; ///< height expression string
99 char *in_color_matrix;
100 char *out_color_matrix;
110 int force_original_aspect_ratio;
113 static av_cold int init_dict(AVFilterContext *ctx, AVDictionary **opts)
115 ScaleContext *scale = ctx->priv;
118 if (scale->size_str && (scale->w_expr || scale->h_expr)) {
119 av_log(ctx, AV_LOG_ERROR,
120 "Size and width/height expressions cannot be set at the same time.\n");
121 return AVERROR(EINVAL);
124 if (scale->w_expr && !scale->h_expr)
125 FFSWAP(char *, scale->w_expr, scale->size_str);
127 if (scale->size_str) {
129 if ((ret = av_parse_video_size(&scale->w, &scale->h, scale->size_str)) < 0) {
130 av_log(ctx, AV_LOG_ERROR,
131 "Invalid size '%s'\n", scale->size_str);
134 snprintf(buf, sizeof(buf)-1, "%d", scale->w);
135 av_opt_set(scale, "w", buf, 0);
136 snprintf(buf, sizeof(buf)-1, "%d", scale->h);
137 av_opt_set(scale, "h", buf, 0);
140 av_opt_set(scale, "w", "iw", 0);
142 av_opt_set(scale, "h", "ih", 0);
144 av_log(ctx, AV_LOG_VERBOSE, "w:%s h:%s flags:'%s' interl:%d\n",
145 scale->w_expr, scale->h_expr, (char *)av_x_if_null(scale->flags_str, ""), scale->interlaced);
149 if (scale->flags_str) {
150 const AVClass *class = sws_get_class();
151 const AVOption *o = av_opt_find(&class, "sws_flags", NULL, 0,
152 AV_OPT_SEARCH_FAKE_OBJ);
153 int ret = av_opt_eval_flags(&class, o, scale->flags_str, &scale->flags);
163 static av_cold void uninit(AVFilterContext *ctx)
165 ScaleContext *scale = ctx->priv;
166 sws_freeContext(scale->sws);
167 sws_freeContext(scale->isws[0]);
168 sws_freeContext(scale->isws[1]);
170 av_dict_free(&scale->opts);
173 static int query_formats(AVFilterContext *ctx)
175 AVFilterFormats *formats;
176 enum AVPixelFormat pix_fmt;
179 if (ctx->inputs[0]) {
181 for (pix_fmt = 0; pix_fmt < AV_PIX_FMT_NB; pix_fmt++)
182 if ((sws_isSupportedInput(pix_fmt) ||
183 sws_isSupportedEndiannessConversion(pix_fmt))
184 && (ret = ff_add_format(&formats, pix_fmt)) < 0) {
185 ff_formats_unref(&formats);
188 ff_formats_ref(formats, &ctx->inputs[0]->out_formats);
190 if (ctx->outputs[0]) {
192 for (pix_fmt = 0; pix_fmt < AV_PIX_FMT_NB; pix_fmt++)
193 if ((sws_isSupportedOutput(pix_fmt) || pix_fmt == AV_PIX_FMT_PAL8 ||
194 sws_isSupportedEndiannessConversion(pix_fmt))
195 && (ret = ff_add_format(&formats, pix_fmt)) < 0) {
196 ff_formats_unref(&formats);
199 ff_formats_ref(formats, &ctx->outputs[0]->in_formats);
205 static const int *parse_yuv_type(const char *s, enum AVColorSpace colorspace)
210 if (s && strstr(s, "bt709")) {
211 colorspace = AVCOL_SPC_BT709;
212 } else if (s && strstr(s, "fcc")) {
213 colorspace = AVCOL_SPC_FCC;
214 } else if (s && strstr(s, "smpte240m")) {
215 colorspace = AVCOL_SPC_SMPTE240M;
216 } else if (s && (strstr(s, "bt601") || strstr(s, "bt470") || strstr(s, "smpte170m"))) {
217 colorspace = AVCOL_SPC_BT470BG;
220 if (colorspace < 1 || colorspace > 7) {
221 colorspace = AVCOL_SPC_BT470BG;
224 return sws_getCoefficients(colorspace);
227 static int config_props(AVFilterLink *outlink)
229 AVFilterContext *ctx = outlink->src;
230 AVFilterLink *inlink = outlink->src->inputs[0];
231 enum AVPixelFormat outfmt = outlink->format;
232 ScaleContext *scale = ctx->priv;
233 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(inlink->format);
234 const AVPixFmtDescriptor *out_desc = av_pix_fmt_desc_get(outlink->format);
236 double var_values[VARS_NB], res;
240 var_values[VAR_IN_W] = var_values[VAR_IW] = inlink->w;
241 var_values[VAR_IN_H] = var_values[VAR_IH] = inlink->h;
242 var_values[VAR_OUT_W] = var_values[VAR_OW] = NAN;
243 var_values[VAR_OUT_H] = var_values[VAR_OH] = NAN;
244 var_values[VAR_A] = (double) inlink->w / inlink->h;
245 var_values[VAR_SAR] = inlink->sample_aspect_ratio.num ?
246 (double) inlink->sample_aspect_ratio.num / inlink->sample_aspect_ratio.den : 1;
247 var_values[VAR_DAR] = var_values[VAR_A] * var_values[VAR_SAR];
248 var_values[VAR_HSUB] = 1 << desc->log2_chroma_w;
249 var_values[VAR_VSUB] = 1 << desc->log2_chroma_h;
250 var_values[VAR_OHSUB] = 1 << out_desc->log2_chroma_w;
251 var_values[VAR_OVSUB] = 1 << out_desc->log2_chroma_h;
253 /* evaluate width and height */
254 av_expr_parse_and_eval(&res, (expr = scale->w_expr),
255 var_names, var_values,
256 NULL, NULL, NULL, NULL, NULL, 0, ctx);
257 scale->w = var_values[VAR_OUT_W] = var_values[VAR_OW] = res;
258 if ((ret = av_expr_parse_and_eval(&res, (expr = scale->h_expr),
259 var_names, var_values,
260 NULL, NULL, NULL, NULL, NULL, 0, ctx)) < 0)
262 scale->h = var_values[VAR_OUT_H] = var_values[VAR_OH] = res;
263 /* evaluate again the width, as it may depend on the output height */
264 if ((ret = av_expr_parse_and_eval(&res, (expr = scale->w_expr),
265 var_names, var_values,
266 NULL, NULL, NULL, NULL, NULL, 0, ctx)) < 0)
273 /* sanity check params */
274 if (w < -1 || h < -1) {
275 av_log(ctx, AV_LOG_ERROR, "Size values less than -1 are not acceptable.\n");
276 return AVERROR(EINVAL);
278 if (w == -1 && h == -1)
279 scale->w = scale->h = 0;
286 w = av_rescale(h, inlink->w, inlink->h);
288 h = av_rescale(w, inlink->h, inlink->w);
290 if (scale->force_original_aspect_ratio) {
291 int tmp_w = av_rescale(h, inlink->w, inlink->h);
292 int tmp_h = av_rescale(w, inlink->h, inlink->w);
294 if (scale->force_original_aspect_ratio == 1) {
303 if (w > INT_MAX || h > INT_MAX ||
304 (h * inlink->w) > INT_MAX ||
305 (w * inlink->h) > INT_MAX)
306 av_log(ctx, AV_LOG_ERROR, "Rescaled value for width or height is too big.\n");
311 /* TODO: make algorithm configurable */
313 scale->input_is_pal = desc->flags & AV_PIX_FMT_FLAG_PAL ||
314 desc->flags & AV_PIX_FMT_FLAG_PSEUDOPAL;
315 if (outfmt == AV_PIX_FMT_PAL8) outfmt = AV_PIX_FMT_BGR8;
316 scale->output_is_pal = av_pix_fmt_desc_get(outfmt)->flags & AV_PIX_FMT_FLAG_PAL ||
317 av_pix_fmt_desc_get(outfmt)->flags & AV_PIX_FMT_FLAG_PSEUDOPAL;
320 sws_freeContext(scale->sws);
322 sws_freeContext(scale->isws[0]);
324 sws_freeContext(scale->isws[1]);
325 scale->isws[0] = scale->isws[1] = scale->sws = NULL;
326 if (inlink->w == outlink->w && inlink->h == outlink->h &&
327 inlink->format == outlink->format)
330 struct SwsContext **swscs[3] = {&scale->sws, &scale->isws[0], &scale->isws[1]};
333 for (i = 0; i < 3; i++) {
334 struct SwsContext **s = swscs[i];
335 *s = sws_alloc_context();
337 return AVERROR(ENOMEM);
340 AVDictionaryEntry *e = NULL;
342 while ((e = av_dict_get(scale->opts, "", e, AV_DICT_IGNORE_SUFFIX))) {
343 if ((ret = av_opt_set(*s, e->key, e->value, 0)) < 0)
348 av_opt_set_int(*s, "srcw", inlink ->w, 0);
349 av_opt_set_int(*s, "srch", inlink ->h >> !!i, 0);
350 av_opt_set_int(*s, "src_format", inlink->format, 0);
351 av_opt_set_int(*s, "dstw", outlink->w, 0);
352 av_opt_set_int(*s, "dsth", outlink->h >> !!i, 0);
353 av_opt_set_int(*s, "dst_format", outfmt, 0);
354 av_opt_set_int(*s, "sws_flags", scale->flags, 0);
356 av_opt_set_int(*s, "src_h_chr_pos", scale->in_h_chr_pos, 0);
357 av_opt_set_int(*s, "src_v_chr_pos", scale->in_v_chr_pos, 0);
358 av_opt_set_int(*s, "dst_h_chr_pos", scale->out_h_chr_pos, 0);
359 av_opt_set_int(*s, "dst_v_chr_pos", scale->out_v_chr_pos, 0);
361 if ((ret = sws_init_context(*s, NULL, NULL)) < 0)
363 if (!scale->interlaced)
368 if (inlink->sample_aspect_ratio.num){
369 outlink->sample_aspect_ratio = av_mul_q((AVRational){outlink->h * inlink->w, outlink->w * inlink->h}, inlink->sample_aspect_ratio);
371 outlink->sample_aspect_ratio = inlink->sample_aspect_ratio;
373 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:0x%0x\n",
374 inlink ->w, inlink ->h, av_get_pix_fmt_name( inlink->format),
375 inlink->sample_aspect_ratio.num, inlink->sample_aspect_ratio.den,
376 outlink->w, outlink->h, av_get_pix_fmt_name(outlink->format),
377 outlink->sample_aspect_ratio.num, outlink->sample_aspect_ratio.den,
382 av_log(NULL, AV_LOG_ERROR,
383 "Error when evaluating the expression '%s'.\n"
384 "Maybe the expression for out_w:'%s' or for out_h:'%s' is self-referencing.\n",
385 expr, scale->w_expr, scale->h_expr);
389 static int scale_slice(AVFilterLink *link, AVFrame *out_buf, AVFrame *cur_pic, struct SwsContext *sws, int y, int h, int mul, int field)
391 ScaleContext *scale = link->dst->priv;
392 const uint8_t *in[4];
394 int in_stride[4],out_stride[4];
398 int vsub= ((i+1)&2) ? scale->vsub : 0;
399 in_stride[i] = cur_pic->linesize[i] * mul;
400 out_stride[i] = out_buf->linesize[i] * mul;
401 in[i] = cur_pic->data[i] + ((y>>vsub)+field) * cur_pic->linesize[i];
402 out[i] = out_buf->data[i] + field * out_buf->linesize[i];
404 if(scale->input_is_pal)
405 in[1] = cur_pic->data[1];
406 if(scale->output_is_pal)
407 out[1] = out_buf->data[1];
409 return sws_scale(sws, in, in_stride, y/mul, h,
413 static int filter_frame(AVFilterLink *link, AVFrame *in)
415 ScaleContext *scale = link->dst->priv;
416 AVFilterLink *outlink = link->dst->outputs[0];
418 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(link->format);
422 if( in->width != link->w
423 || in->height != link->h
424 || in->format != link->format) {
426 snprintf(buf, sizeof(buf)-1, "%d", outlink->w);
427 av_opt_set(scale, "w", buf, 0);
428 snprintf(buf, sizeof(buf)-1, "%d", outlink->h);
429 av_opt_set(scale, "h", buf, 0);
431 link->dst->inputs[0]->format = in->format;
432 link->dst->inputs[0]->w = in->width;
433 link->dst->inputs[0]->h = in->height;
435 if ((ret = config_props(outlink)) < 0)
440 return ff_filter_frame(outlink, in);
442 scale->hsub = desc->log2_chroma_w;
443 scale->vsub = desc->log2_chroma_h;
445 out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
448 return AVERROR(ENOMEM);
451 av_frame_copy_props(out, in);
452 out->width = outlink->w;
453 out->height = outlink->h;
455 if(scale->output_is_pal)
456 avpriv_set_systematic_pal2((uint32_t*)out->data[1], outlink->format == AV_PIX_FMT_PAL8 ? AV_PIX_FMT_BGR8 : outlink->format);
458 in_range = av_frame_get_color_range(in);
460 if ( scale->in_color_matrix
461 || scale->out_color_matrix
462 || scale-> in_range != AVCOL_RANGE_UNSPECIFIED
463 || in_range != AVCOL_RANGE_UNSPECIFIED
464 || scale->out_range != AVCOL_RANGE_UNSPECIFIED) {
465 int in_full, out_full, brightness, contrast, saturation;
466 const int *inv_table, *table;
468 sws_getColorspaceDetails(scale->sws, (int **)&inv_table, &in_full,
469 (int **)&table, &out_full,
470 &brightness, &contrast, &saturation);
472 if (scale->in_color_matrix)
473 inv_table = parse_yuv_type(scale->in_color_matrix, av_frame_get_colorspace(in));
474 if (scale->out_color_matrix)
475 table = parse_yuv_type(scale->out_color_matrix, AVCOL_SPC_UNSPECIFIED);
477 if (scale-> in_range != AVCOL_RANGE_UNSPECIFIED)
478 in_full = (scale-> in_range == AVCOL_RANGE_JPEG);
479 else if (in_range != AVCOL_RANGE_UNSPECIFIED)
480 in_full = (in_range == AVCOL_RANGE_JPEG);
481 if (scale->out_range != AVCOL_RANGE_UNSPECIFIED)
482 out_full = (scale->out_range == AVCOL_RANGE_JPEG);
484 sws_setColorspaceDetails(scale->sws, inv_table, in_full,
486 brightness, contrast, saturation);
488 sws_setColorspaceDetails(scale->isws[0], inv_table, in_full,
490 brightness, contrast, saturation);
492 sws_setColorspaceDetails(scale->isws[1], inv_table, in_full,
494 brightness, contrast, saturation);
497 av_reduce(&out->sample_aspect_ratio.num, &out->sample_aspect_ratio.den,
498 (int64_t)in->sample_aspect_ratio.num * outlink->h * link->w,
499 (int64_t)in->sample_aspect_ratio.den * outlink->w * link->h,
502 if(scale->interlaced>0 || (scale->interlaced<0 && in->interlaced_frame)){
503 scale_slice(link, out, in, scale->isws[0], 0, (link->h+1)/2, 2, 0);
504 scale_slice(link, out, in, scale->isws[1], 0, link->h /2, 2, 1);
506 scale_slice(link, out, in, scale->sws, 0, link->h, 1, 0);
510 return ff_filter_frame(outlink, out);
513 static const AVClass *child_class_next(const AVClass *prev)
515 return prev ? NULL : sws_get_class();
518 #define OFFSET(x) offsetof(ScaleContext, x)
519 #define FLAGS AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
521 static const AVOption scale_options[] = {
522 { "w", "Output video width", OFFSET(w_expr), AV_OPT_TYPE_STRING, .flags = FLAGS },
523 { "width", "Output video width", OFFSET(w_expr), AV_OPT_TYPE_STRING, .flags = FLAGS },
524 { "h", "Output video height", OFFSET(h_expr), AV_OPT_TYPE_STRING, .flags = FLAGS },
525 { "height","Output video height", OFFSET(h_expr), AV_OPT_TYPE_STRING, .flags = FLAGS },
526 { "flags", "Flags to pass to libswscale", OFFSET(flags_str), AV_OPT_TYPE_STRING, { .str = "bilinear" }, .flags = FLAGS },
527 { "interl", "set interlacing", OFFSET(interlaced), AV_OPT_TYPE_INT, {.i64 = 0 }, -1, 1, FLAGS },
528 { "size", "set video size", OFFSET(size_str), AV_OPT_TYPE_STRING, {.str = NULL}, 0, FLAGS },
529 { "s", "set video size", OFFSET(size_str), AV_OPT_TYPE_STRING, {.str = NULL}, 0, FLAGS },
530 { "in_color_matrix", "set input YCbCr type", OFFSET(in_color_matrix), AV_OPT_TYPE_STRING, { .str = "auto" }, .flags = FLAGS },
531 { "out_color_matrix", "set output YCbCr type", OFFSET(out_color_matrix), AV_OPT_TYPE_STRING, { .str = NULL }, .flags = FLAGS },
532 { "in_range", "set input color range", OFFSET( in_range), AV_OPT_TYPE_INT, {.i64 = AVCOL_RANGE_UNSPECIFIED }, 0, 2, FLAGS, "range" },
533 { "out_range", "set output color range", OFFSET(out_range), AV_OPT_TYPE_INT, {.i64 = AVCOL_RANGE_UNSPECIFIED }, 0, 2, FLAGS, "range" },
534 { "auto", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = AVCOL_RANGE_UNSPECIFIED }, 0, 0, FLAGS, "range" },
535 { "full", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = AVCOL_RANGE_JPEG}, 0, 0, FLAGS, "range" },
536 { "jpeg", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = AVCOL_RANGE_JPEG}, 0, 0, FLAGS, "range" },
537 { "mpeg", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = AVCOL_RANGE_MPEG}, 0, 0, FLAGS, "range" },
538 { "tv", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = AVCOL_RANGE_MPEG}, 0, 0, FLAGS, "range" },
539 { "pc", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = AVCOL_RANGE_JPEG}, 0, 0, FLAGS, "range" },
540 { "in_v_chr_pos", "input vertical chroma position in luma grid/256" , OFFSET(in_v_chr_pos), AV_OPT_TYPE_INT, { .i64 = -1}, -1, 512, FLAGS },
541 { "in_h_chr_pos", "input horizontal chroma position in luma grid/256", OFFSET(in_h_chr_pos), AV_OPT_TYPE_INT, { .i64 = -1}, -1, 512, FLAGS },
542 { "out_v_chr_pos", "output vertical chroma position in luma grid/256" , OFFSET(out_v_chr_pos), AV_OPT_TYPE_INT, { .i64 = -1}, -1, 512, FLAGS },
543 { "out_h_chr_pos", "output horizontal chroma position in luma grid/256", OFFSET(out_h_chr_pos), AV_OPT_TYPE_INT, { .i64 = -1}, -1, 512, FLAGS },
544 { "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" },
545 { "disable", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 0 }, 0, 0, FLAGS, "force_oar" },
546 { "decrease", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 1 }, 0, 0, FLAGS, "force_oar" },
547 { "increase", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 2 }, 0, 0, FLAGS, "force_oar" },
551 static const AVClass scale_class = {
552 .class_name = "scale",
553 .item_name = av_default_item_name,
554 .option = scale_options,
555 .version = LIBAVUTIL_VERSION_INT,
556 .child_class_next = child_class_next,
559 static const AVFilterPad avfilter_vf_scale_inputs[] = {
562 .type = AVMEDIA_TYPE_VIDEO,
563 .filter_frame = filter_frame,
568 static const AVFilterPad avfilter_vf_scale_outputs[] = {
571 .type = AVMEDIA_TYPE_VIDEO,
572 .config_props = config_props,
577 AVFilter ff_vf_scale = {
579 .description = NULL_IF_CONFIG_SMALL("Scale the input video size and/or convert the image format."),
580 .init_dict = init_dict,
582 .query_formats = query_formats,
583 .priv_size = sizeof(ScaleContext),
584 .priv_class = &scale_class,
585 .inputs = avfilter_vf_scale_inputs,
586 .outputs = avfilter_vf_scale_outputs,