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[] = {
74 typedef struct ScaleContext {
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
84 * -N = try to keep aspect but make sure it is divisible by N
88 unsigned int flags; ///sws flags
90 int hsub, vsub; ///< chroma subsampling
91 int slice_y; ///< top of current output slice
92 int input_is_pal; ///< set to 1 if the input format is paletted
93 int output_is_pal; ///< set to 1 if the output format is paletted
96 char *w_expr; ///< width expression string
97 char *h_expr; ///< height expression string
100 char *in_color_matrix;
101 char *out_color_matrix;
111 int force_original_aspect_ratio;
114 static av_cold int init_dict(AVFilterContext *ctx, AVDictionary **opts)
116 ScaleContext *scale = ctx->priv;
119 if (scale->size_str && (scale->w_expr || scale->h_expr)) {
120 av_log(ctx, AV_LOG_ERROR,
121 "Size and width/height expressions cannot be set at the same time.\n");
122 return AVERROR(EINVAL);
125 if (scale->w_expr && !scale->h_expr)
126 FFSWAP(char *, scale->w_expr, scale->size_str);
128 if (scale->size_str) {
130 if ((ret = av_parse_video_size(&scale->w, &scale->h, scale->size_str)) < 0) {
131 av_log(ctx, AV_LOG_ERROR,
132 "Invalid size '%s'\n", scale->size_str);
135 snprintf(buf, sizeof(buf)-1, "%d", scale->w);
136 av_opt_set(scale, "w", buf, 0);
137 snprintf(buf, sizeof(buf)-1, "%d", scale->h);
138 av_opt_set(scale, "h", buf, 0);
141 av_opt_set(scale, "w", "iw", 0);
143 av_opt_set(scale, "h", "ih", 0);
145 av_log(ctx, AV_LOG_VERBOSE, "w:%s h:%s flags:'%s' interl:%d\n",
146 scale->w_expr, scale->h_expr, (char *)av_x_if_null(scale->flags_str, ""), scale->interlaced);
150 if (scale->flags_str) {
151 const AVClass *class = sws_get_class();
152 const AVOption *o = av_opt_find(&class, "sws_flags", NULL, 0,
153 AV_OPT_SEARCH_FAKE_OBJ);
154 int ret = av_opt_eval_flags(&class, o, scale->flags_str, &scale->flags);
164 static av_cold void uninit(AVFilterContext *ctx)
166 ScaleContext *scale = ctx->priv;
167 sws_freeContext(scale->sws);
168 sws_freeContext(scale->isws[0]);
169 sws_freeContext(scale->isws[1]);
171 av_dict_free(&scale->opts);
174 static int query_formats(AVFilterContext *ctx)
176 AVFilterFormats *formats;
177 enum AVPixelFormat pix_fmt;
180 if (ctx->inputs[0]) {
181 const AVPixFmtDescriptor *desc = NULL;
183 while ((desc = av_pix_fmt_desc_next(desc))) {
184 pix_fmt = av_pix_fmt_desc_get_id(desc);
185 if ((sws_isSupportedInput(pix_fmt) ||
186 sws_isSupportedEndiannessConversion(pix_fmt))
187 && (ret = ff_add_format(&formats, pix_fmt)) < 0) {
188 ff_formats_unref(&formats);
192 ff_formats_ref(formats, &ctx->inputs[0]->out_formats);
194 if (ctx->outputs[0]) {
195 const AVPixFmtDescriptor *desc = NULL;
197 while ((desc = av_pix_fmt_desc_next(desc))) {
198 pix_fmt = av_pix_fmt_desc_get_id(desc);
199 if ((sws_isSupportedOutput(pix_fmt) || pix_fmt == AV_PIX_FMT_PAL8 ||
200 sws_isSupportedEndiannessConversion(pix_fmt))
201 && (ret = ff_add_format(&formats, pix_fmt)) < 0) {
202 ff_formats_unref(&formats);
206 ff_formats_ref(formats, &ctx->outputs[0]->in_formats);
212 static const int *parse_yuv_type(const char *s, enum AVColorSpace colorspace)
217 if (s && strstr(s, "bt709")) {
218 colorspace = AVCOL_SPC_BT709;
219 } else if (s && strstr(s, "fcc")) {
220 colorspace = AVCOL_SPC_FCC;
221 } else if (s && strstr(s, "smpte240m")) {
222 colorspace = AVCOL_SPC_SMPTE240M;
223 } else if (s && (strstr(s, "bt601") || strstr(s, "bt470") || strstr(s, "smpte170m"))) {
224 colorspace = AVCOL_SPC_BT470BG;
227 if (colorspace < 1 || colorspace > 7) {
228 colorspace = AVCOL_SPC_BT470BG;
231 return sws_getCoefficients(colorspace);
234 static int config_props(AVFilterLink *outlink)
236 AVFilterContext *ctx = outlink->src;
237 AVFilterLink *inlink = outlink->src->inputs[0];
238 enum AVPixelFormat outfmt = outlink->format;
239 ScaleContext *scale = ctx->priv;
240 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(inlink->format);
241 const AVPixFmtDescriptor *out_desc = av_pix_fmt_desc_get(outlink->format);
243 double var_values[VARS_NB], res;
246 int factor_w, factor_h;
248 var_values[VAR_IN_W] = var_values[VAR_IW] = inlink->w;
249 var_values[VAR_IN_H] = var_values[VAR_IH] = inlink->h;
250 var_values[VAR_OUT_W] = var_values[VAR_OW] = NAN;
251 var_values[VAR_OUT_H] = var_values[VAR_OH] = NAN;
252 var_values[VAR_A] = (double) inlink->w / inlink->h;
253 var_values[VAR_SAR] = inlink->sample_aspect_ratio.num ?
254 (double) inlink->sample_aspect_ratio.num / inlink->sample_aspect_ratio.den : 1;
255 var_values[VAR_DAR] = var_values[VAR_A] * var_values[VAR_SAR];
256 var_values[VAR_HSUB] = 1 << desc->log2_chroma_w;
257 var_values[VAR_VSUB] = 1 << desc->log2_chroma_h;
258 var_values[VAR_OHSUB] = 1 << out_desc->log2_chroma_w;
259 var_values[VAR_OVSUB] = 1 << out_desc->log2_chroma_h;
261 /* evaluate width and height */
262 av_expr_parse_and_eval(&res, (expr = scale->w_expr),
263 var_names, var_values,
264 NULL, NULL, NULL, NULL, NULL, 0, ctx);
265 scale->w = var_values[VAR_OUT_W] = var_values[VAR_OW] = res;
266 if ((ret = av_expr_parse_and_eval(&res, (expr = scale->h_expr),
267 var_names, var_values,
268 NULL, NULL, NULL, NULL, NULL, 0, ctx)) < 0)
270 scale->h = var_values[VAR_OUT_H] = var_values[VAR_OH] = res;
271 /* evaluate again the width, as it may depend on the output height */
272 if ((ret = av_expr_parse_and_eval(&res, (expr = scale->w_expr),
273 var_names, var_values,
274 NULL, NULL, NULL, NULL, NULL, 0, ctx)) < 0)
281 /* Check if it is requested that the result has to be divisible by a some
282 * factor (w or h = -n with n being the factor). */
293 scale->w = scale->h = 0;
300 /* Make sure that the result is divisible by the factor we determined
301 * earlier. If no factor was set, it is nothing will happen as the default
304 w = av_rescale(h, inlink->w, inlink->h * factor_w) * factor_w;
306 h = av_rescale(w, inlink->h, inlink->w * factor_h) * factor_h;
308 /* Note that force_original_aspect_ratio may overwrite the previous set
309 * dimensions so that it is not divisible by the set factors anymore. */
310 if (scale->force_original_aspect_ratio) {
311 int tmp_w = av_rescale(h, inlink->w, inlink->h);
312 int tmp_h = av_rescale(w, inlink->h, inlink->w);
314 if (scale->force_original_aspect_ratio == 1) {
323 if (w > INT_MAX || h > INT_MAX ||
324 (h * inlink->w) > INT_MAX ||
325 (w * inlink->h) > INT_MAX)
326 av_log(ctx, AV_LOG_ERROR, "Rescaled value for width or height is too big.\n");
331 /* TODO: make algorithm configurable */
333 scale->input_is_pal = desc->flags & AV_PIX_FMT_FLAG_PAL ||
334 desc->flags & AV_PIX_FMT_FLAG_PSEUDOPAL;
335 if (outfmt == AV_PIX_FMT_PAL8) outfmt = AV_PIX_FMT_BGR8;
336 scale->output_is_pal = av_pix_fmt_desc_get(outfmt)->flags & AV_PIX_FMT_FLAG_PAL ||
337 av_pix_fmt_desc_get(outfmt)->flags & AV_PIX_FMT_FLAG_PSEUDOPAL;
340 sws_freeContext(scale->sws);
342 sws_freeContext(scale->isws[0]);
344 sws_freeContext(scale->isws[1]);
345 scale->isws[0] = scale->isws[1] = scale->sws = NULL;
346 if (inlink->w == outlink->w && inlink->h == outlink->h &&
347 inlink->format == outlink->format)
350 struct SwsContext **swscs[3] = {&scale->sws, &scale->isws[0], &scale->isws[1]};
353 for (i = 0; i < 3; i++) {
354 struct SwsContext **s = swscs[i];
355 *s = sws_alloc_context();
357 return AVERROR(ENOMEM);
360 AVDictionaryEntry *e = NULL;
362 while ((e = av_dict_get(scale->opts, "", e, AV_DICT_IGNORE_SUFFIX))) {
363 if ((ret = av_opt_set(*s, e->key, e->value, 0)) < 0)
368 av_opt_set_int(*s, "srcw", inlink ->w, 0);
369 av_opt_set_int(*s, "srch", inlink ->h >> !!i, 0);
370 av_opt_set_int(*s, "src_format", inlink->format, 0);
371 av_opt_set_int(*s, "dstw", outlink->w, 0);
372 av_opt_set_int(*s, "dsth", outlink->h >> !!i, 0);
373 av_opt_set_int(*s, "dst_format", outfmt, 0);
374 av_opt_set_int(*s, "sws_flags", scale->flags, 0);
376 av_opt_set_int(*s, "src_h_chr_pos", scale->in_h_chr_pos, 0);
377 av_opt_set_int(*s, "src_v_chr_pos", scale->in_v_chr_pos, 0);
378 av_opt_set_int(*s, "dst_h_chr_pos", scale->out_h_chr_pos, 0);
379 av_opt_set_int(*s, "dst_v_chr_pos", scale->out_v_chr_pos, 0);
381 if ((ret = sws_init_context(*s, NULL, NULL)) < 0)
383 if (!scale->interlaced)
388 if (inlink->sample_aspect_ratio.num){
389 outlink->sample_aspect_ratio = av_mul_q((AVRational){outlink->h * inlink->w, outlink->w * inlink->h}, inlink->sample_aspect_ratio);
391 outlink->sample_aspect_ratio = inlink->sample_aspect_ratio;
393 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",
394 inlink ->w, inlink ->h, av_get_pix_fmt_name( inlink->format),
395 inlink->sample_aspect_ratio.num, inlink->sample_aspect_ratio.den,
396 outlink->w, outlink->h, av_get_pix_fmt_name(outlink->format),
397 outlink->sample_aspect_ratio.num, outlink->sample_aspect_ratio.den,
402 av_log(NULL, AV_LOG_ERROR,
403 "Error when evaluating the expression '%s'.\n"
404 "Maybe the expression for out_w:'%s' or for out_h:'%s' is self-referencing.\n",
405 expr, scale->w_expr, scale->h_expr);
409 static int scale_slice(AVFilterLink *link, AVFrame *out_buf, AVFrame *cur_pic, struct SwsContext *sws, int y, int h, int mul, int field)
411 ScaleContext *scale = link->dst->priv;
412 const uint8_t *in[4];
414 int in_stride[4],out_stride[4];
418 int vsub= ((i+1)&2) ? scale->vsub : 0;
419 in_stride[i] = cur_pic->linesize[i] * mul;
420 out_stride[i] = out_buf->linesize[i] * mul;
421 in[i] = cur_pic->data[i] + ((y>>vsub)+field) * cur_pic->linesize[i];
422 out[i] = out_buf->data[i] + field * out_buf->linesize[i];
424 if(scale->input_is_pal)
425 in[1] = cur_pic->data[1];
426 if(scale->output_is_pal)
427 out[1] = out_buf->data[1];
429 return sws_scale(sws, in, in_stride, y/mul, h,
433 static int filter_frame(AVFilterLink *link, AVFrame *in)
435 ScaleContext *scale = link->dst->priv;
436 AVFilterLink *outlink = link->dst->outputs[0];
438 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(link->format);
442 if( in->width != link->w
443 || in->height != link->h
444 || in->format != link->format) {
446 snprintf(buf, sizeof(buf)-1, "%d", outlink->w);
447 av_opt_set(scale, "w", buf, 0);
448 snprintf(buf, sizeof(buf)-1, "%d", outlink->h);
449 av_opt_set(scale, "h", buf, 0);
451 link->dst->inputs[0]->format = in->format;
452 link->dst->inputs[0]->w = in->width;
453 link->dst->inputs[0]->h = in->height;
455 if ((ret = config_props(outlink)) < 0)
460 return ff_filter_frame(outlink, in);
462 scale->hsub = desc->log2_chroma_w;
463 scale->vsub = desc->log2_chroma_h;
465 out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
468 return AVERROR(ENOMEM);
471 av_frame_copy_props(out, in);
472 out->width = outlink->w;
473 out->height = outlink->h;
475 if(scale->output_is_pal)
476 avpriv_set_systematic_pal2((uint32_t*)out->data[1], outlink->format == AV_PIX_FMT_PAL8 ? AV_PIX_FMT_BGR8 : outlink->format);
478 in_range = av_frame_get_color_range(in);
480 if ( scale->in_color_matrix
481 || scale->out_color_matrix
482 || scale-> in_range != AVCOL_RANGE_UNSPECIFIED
483 || in_range != AVCOL_RANGE_UNSPECIFIED
484 || scale->out_range != AVCOL_RANGE_UNSPECIFIED) {
485 int in_full, out_full, brightness, contrast, saturation;
486 const int *inv_table, *table;
488 sws_getColorspaceDetails(scale->sws, (int **)&inv_table, &in_full,
489 (int **)&table, &out_full,
490 &brightness, &contrast, &saturation);
492 if (scale->in_color_matrix)
493 inv_table = parse_yuv_type(scale->in_color_matrix, av_frame_get_colorspace(in));
494 if (scale->out_color_matrix)
495 table = parse_yuv_type(scale->out_color_matrix, AVCOL_SPC_UNSPECIFIED);
497 if (scale-> in_range != AVCOL_RANGE_UNSPECIFIED)
498 in_full = (scale-> in_range == AVCOL_RANGE_JPEG);
499 else if (in_range != AVCOL_RANGE_UNSPECIFIED)
500 in_full = (in_range == AVCOL_RANGE_JPEG);
501 if (scale->out_range != AVCOL_RANGE_UNSPECIFIED)
502 out_full = (scale->out_range == AVCOL_RANGE_JPEG);
504 sws_setColorspaceDetails(scale->sws, inv_table, in_full,
506 brightness, contrast, saturation);
508 sws_setColorspaceDetails(scale->isws[0], inv_table, in_full,
510 brightness, contrast, saturation);
512 sws_setColorspaceDetails(scale->isws[1], inv_table, in_full,
514 brightness, contrast, saturation);
517 av_reduce(&out->sample_aspect_ratio.num, &out->sample_aspect_ratio.den,
518 (int64_t)in->sample_aspect_ratio.num * outlink->h * link->w,
519 (int64_t)in->sample_aspect_ratio.den * outlink->w * link->h,
522 if(scale->interlaced>0 || (scale->interlaced<0 && in->interlaced_frame)){
523 scale_slice(link, out, in, scale->isws[0], 0, (link->h+1)/2, 2, 0);
524 scale_slice(link, out, in, scale->isws[1], 0, link->h /2, 2, 1);
526 scale_slice(link, out, in, scale->sws, 0, link->h, 1, 0);
530 return ff_filter_frame(outlink, out);
533 static const AVClass *child_class_next(const AVClass *prev)
535 return prev ? NULL : sws_get_class();
538 #define OFFSET(x) offsetof(ScaleContext, x)
539 #define FLAGS AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
541 static const AVOption scale_options[] = {
542 { "w", "Output video width", OFFSET(w_expr), AV_OPT_TYPE_STRING, .flags = FLAGS },
543 { "width", "Output video width", OFFSET(w_expr), AV_OPT_TYPE_STRING, .flags = FLAGS },
544 { "h", "Output video height", OFFSET(h_expr), AV_OPT_TYPE_STRING, .flags = FLAGS },
545 { "height","Output video height", OFFSET(h_expr), AV_OPT_TYPE_STRING, .flags = FLAGS },
546 { "flags", "Flags to pass to libswscale", OFFSET(flags_str), AV_OPT_TYPE_STRING, { .str = "bilinear" }, .flags = FLAGS },
547 { "interl", "set interlacing", OFFSET(interlaced), AV_OPT_TYPE_INT, {.i64 = 0 }, -1, 1, FLAGS },
548 { "size", "set video size", OFFSET(size_str), AV_OPT_TYPE_STRING, {.str = NULL}, 0, FLAGS },
549 { "s", "set video size", OFFSET(size_str), AV_OPT_TYPE_STRING, {.str = NULL}, 0, FLAGS },
550 { "in_color_matrix", "set input YCbCr type", OFFSET(in_color_matrix), AV_OPT_TYPE_STRING, { .str = "auto" }, .flags = FLAGS },
551 { "out_color_matrix", "set output YCbCr type", OFFSET(out_color_matrix), AV_OPT_TYPE_STRING, { .str = NULL }, .flags = FLAGS },
552 { "in_range", "set input color range", OFFSET( in_range), AV_OPT_TYPE_INT, {.i64 = AVCOL_RANGE_UNSPECIFIED }, 0, 2, FLAGS, "range" },
553 { "out_range", "set output color range", OFFSET(out_range), AV_OPT_TYPE_INT, {.i64 = AVCOL_RANGE_UNSPECIFIED }, 0, 2, FLAGS, "range" },
554 { "auto", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = AVCOL_RANGE_UNSPECIFIED }, 0, 0, FLAGS, "range" },
555 { "full", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = AVCOL_RANGE_JPEG}, 0, 0, FLAGS, "range" },
556 { "jpeg", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = AVCOL_RANGE_JPEG}, 0, 0, FLAGS, "range" },
557 { "mpeg", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = AVCOL_RANGE_MPEG}, 0, 0, FLAGS, "range" },
558 { "tv", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = AVCOL_RANGE_MPEG}, 0, 0, FLAGS, "range" },
559 { "pc", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = AVCOL_RANGE_JPEG}, 0, 0, FLAGS, "range" },
560 { "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 },
561 { "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 },
562 { "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 },
563 { "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 },
564 { "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" },
565 { "disable", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 0 }, 0, 0, FLAGS, "force_oar" },
566 { "decrease", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 1 }, 0, 0, FLAGS, "force_oar" },
567 { "increase", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 2 }, 0, 0, FLAGS, "force_oar" },
571 static const AVClass scale_class = {
572 .class_name = "scale",
573 .item_name = av_default_item_name,
574 .option = scale_options,
575 .version = LIBAVUTIL_VERSION_INT,
576 .child_class_next = child_class_next,
579 static const AVFilterPad avfilter_vf_scale_inputs[] = {
582 .type = AVMEDIA_TYPE_VIDEO,
583 .filter_frame = filter_frame,
588 static const AVFilterPad avfilter_vf_scale_outputs[] = {
591 .type = AVMEDIA_TYPE_VIDEO,
592 .config_props = config_props,
597 AVFilter ff_vf_scale = {
599 .description = NULL_IF_CONFIG_SMALL("Scale the input video size and/or convert the image format."),
600 .init_dict = init_dict,
602 .query_formats = query_formats,
603 .priv_size = sizeof(ScaleContext),
604 .priv_class = &scale_class,
605 .inputs = avfilter_vf_scale_inputs,
606 .outputs = avfilter_vf_scale_outputs,