You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

448 lines
15KB

  1. /*
  2. * Copyright (c) 2007 Bobby Bingham
  3. *
  4. * This file is part of FFmpeg.
  5. *
  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.
  10. *
  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.
  15. *
  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
  19. */
  20. /**
  21. * @file
  22. * scale video filter
  23. */
  24. #include <stdio.h>
  25. #include <string.h>
  26. #include "avfilter.h"
  27. #include "formats.h"
  28. #include "internal.h"
  29. #include "video.h"
  30. #include "libavutil/avstring.h"
  31. #include "libavutil/eval.h"
  32. #include "libavutil/internal.h"
  33. #include "libavutil/mathematics.h"
  34. #include "libavutil/opt.h"
  35. #include "libavutil/parseutils.h"
  36. #include "libavutil/pixdesc.h"
  37. #include "libavutil/imgutils.h"
  38. #include "libavutil/avassert.h"
  39. #include "libswscale/swscale.h"
  40. static const char *const var_names[] = {
  41. "in_w", "iw",
  42. "in_h", "ih",
  43. "out_w", "ow",
  44. "out_h", "oh",
  45. "a",
  46. "sar",
  47. "dar",
  48. "hsub",
  49. "vsub",
  50. NULL
  51. };
  52. enum var_name {
  53. VAR_IN_W, VAR_IW,
  54. VAR_IN_H, VAR_IH,
  55. VAR_OUT_W, VAR_OW,
  56. VAR_OUT_H, VAR_OH,
  57. VAR_A,
  58. VAR_SAR,
  59. VAR_DAR,
  60. VAR_HSUB,
  61. VAR_VSUB,
  62. VARS_NB
  63. };
  64. typedef struct {
  65. const AVClass *class;
  66. struct SwsContext *sws; ///< software scaler context
  67. struct SwsContext *isws[2]; ///< software scaler context for interlaced material
  68. /**
  69. * New dimensions. Special values are:
  70. * 0 = original width/height
  71. * -1 = keep original aspect
  72. */
  73. int w, h;
  74. char *size_str;
  75. unsigned int flags; ///sws flags
  76. int hsub, vsub; ///< chroma subsampling
  77. int slice_y; ///< top of current output slice
  78. int input_is_pal; ///< set to 1 if the input format is paletted
  79. int output_is_pal; ///< set to 1 if the output format is paletted
  80. int interlaced;
  81. char *w_expr; ///< width expression string
  82. char *h_expr; ///< height expression string
  83. char *flags_str;
  84. } ScaleContext;
  85. static av_cold int init(AVFilterContext *ctx, const char *args)
  86. {
  87. ScaleContext *scale = ctx->priv;
  88. #if 1
  89. static const char *shorthand[] = { "w", "h", NULL };
  90. int ret;
  91. const char *args0 = args;
  92. if (args && (scale->size_str = av_get_token(&args, ":"))) {
  93. if (av_parse_video_size(&scale->w, &scale->h, scale->size_str) < 0) {
  94. av_freep(&scale->size_str);
  95. args = args0;
  96. } else if (*args)
  97. args++;
  98. }
  99. if ((ret = av_opt_set_from_string(scale, args, shorthand, "=", ":")) < 0)
  100. return ret;
  101. if (scale->size_str && (scale->w_expr || scale->h_expr)) {
  102. av_log(ctx, AV_LOG_ERROR,
  103. "Size and width/height expressions cannot be set at the same time.\n");
  104. return AVERROR(EINVAL);
  105. }
  106. if (scale->size_str) {
  107. char buf[32];
  108. if ((ret = av_parse_video_size(&scale->w, &scale->h, scale->size_str)) < 0) {
  109. av_log(ctx, AV_LOG_ERROR,
  110. "Invalid size '%s'\n", scale->size_str);
  111. return ret;
  112. }
  113. snprintf(buf, sizeof(buf)-1, "%d", scale->w);
  114. av_opt_set(scale, "w", buf, 0);
  115. snprintf(buf, sizeof(buf)-1, "%d", scale->h);
  116. av_opt_set(scale, "h", buf, 0);
  117. }
  118. if (!scale->w_expr)
  119. av_opt_set(scale, "w", "iw", 0);
  120. if (!scale->h_expr)
  121. av_opt_set(scale, "h", "ih", 0);
  122. av_log(ctx, AV_LOG_VERBOSE, "w:%s h:%s flags:'%s' interl:%d\n",
  123. scale->w_expr, scale->h_expr, (char *)av_x_if_null(scale->flags_str, ""), scale->interlaced);
  124. scale->flags = SWS_BILINEAR;
  125. #endif
  126. if (scale->flags_str) {
  127. const AVClass *class = sws_get_class();
  128. const AVOption *o = av_opt_find(&class, "sws_flags", NULL, 0,
  129. AV_OPT_SEARCH_FAKE_OBJ);
  130. int ret = av_opt_eval_flags(&class, o, scale->flags_str, &scale->flags);
  131. if (ret < 0)
  132. return ret;
  133. }
  134. return 0;
  135. }
  136. static av_cold void uninit(AVFilterContext *ctx)
  137. {
  138. ScaleContext *scale = ctx->priv;
  139. sws_freeContext(scale->sws);
  140. sws_freeContext(scale->isws[0]);
  141. sws_freeContext(scale->isws[1]);
  142. scale->sws = NULL;
  143. av_opt_free(scale);
  144. }
  145. static int query_formats(AVFilterContext *ctx)
  146. {
  147. AVFilterFormats *formats;
  148. enum AVPixelFormat pix_fmt;
  149. int ret;
  150. if (ctx->inputs[0]) {
  151. formats = NULL;
  152. for (pix_fmt = 0; pix_fmt < AV_PIX_FMT_NB; pix_fmt++)
  153. if ( sws_isSupportedInput(pix_fmt)
  154. && (ret = ff_add_format(&formats, pix_fmt)) < 0) {
  155. ff_formats_unref(&formats);
  156. return ret;
  157. }
  158. ff_formats_ref(formats, &ctx->inputs[0]->out_formats);
  159. }
  160. if (ctx->outputs[0]) {
  161. formats = NULL;
  162. for (pix_fmt = 0; pix_fmt < AV_PIX_FMT_NB; pix_fmt++)
  163. if ( (sws_isSupportedOutput(pix_fmt) || pix_fmt == AV_PIX_FMT_PAL8)
  164. && (ret = ff_add_format(&formats, pix_fmt)) < 0) {
  165. ff_formats_unref(&formats);
  166. return ret;
  167. }
  168. ff_formats_ref(formats, &ctx->outputs[0]->in_formats);
  169. }
  170. return 0;
  171. }
  172. static int config_props(AVFilterLink *outlink)
  173. {
  174. AVFilterContext *ctx = outlink->src;
  175. AVFilterLink *inlink = outlink->src->inputs[0];
  176. enum AVPixelFormat outfmt = outlink->format;
  177. ScaleContext *scale = ctx->priv;
  178. const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(inlink->format);
  179. int64_t w, h;
  180. double var_values[VARS_NB], res;
  181. char *expr;
  182. int ret;
  183. var_values[VAR_IN_W] = var_values[VAR_IW] = inlink->w;
  184. var_values[VAR_IN_H] = var_values[VAR_IH] = inlink->h;
  185. var_values[VAR_OUT_W] = var_values[VAR_OW] = NAN;
  186. var_values[VAR_OUT_H] = var_values[VAR_OH] = NAN;
  187. var_values[VAR_A] = (double) inlink->w / inlink->h;
  188. var_values[VAR_SAR] = inlink->sample_aspect_ratio.num ?
  189. (double) inlink->sample_aspect_ratio.num / inlink->sample_aspect_ratio.den : 1;
  190. var_values[VAR_DAR] = var_values[VAR_A] * var_values[VAR_SAR];
  191. var_values[VAR_HSUB] = 1 << desc->log2_chroma_w;
  192. var_values[VAR_VSUB] = 1 << desc->log2_chroma_h;
  193. /* evaluate width and height */
  194. av_expr_parse_and_eval(&res, (expr = scale->w_expr),
  195. var_names, var_values,
  196. NULL, NULL, NULL, NULL, NULL, 0, ctx);
  197. scale->w = var_values[VAR_OUT_W] = var_values[VAR_OW] = res;
  198. if ((ret = av_expr_parse_and_eval(&res, (expr = scale->h_expr),
  199. var_names, var_values,
  200. NULL, NULL, NULL, NULL, NULL, 0, ctx)) < 0)
  201. goto fail;
  202. scale->h = var_values[VAR_OUT_H] = var_values[VAR_OH] = res;
  203. /* evaluate again the width, as it may depend on the output height */
  204. if ((ret = av_expr_parse_and_eval(&res, (expr = scale->w_expr),
  205. var_names, var_values,
  206. NULL, NULL, NULL, NULL, NULL, 0, ctx)) < 0)
  207. goto fail;
  208. scale->w = res;
  209. w = scale->w;
  210. h = scale->h;
  211. /* sanity check params */
  212. if (w < -1 || h < -1) {
  213. av_log(ctx, AV_LOG_ERROR, "Size values less than -1 are not acceptable.\n");
  214. return AVERROR(EINVAL);
  215. }
  216. if (w == -1 && h == -1)
  217. scale->w = scale->h = 0;
  218. if (!(w = scale->w))
  219. w = inlink->w;
  220. if (!(h = scale->h))
  221. h = inlink->h;
  222. if (w == -1)
  223. w = av_rescale(h, inlink->w, inlink->h);
  224. if (h == -1)
  225. h = av_rescale(w, inlink->h, inlink->w);
  226. if (w > INT_MAX || h > INT_MAX ||
  227. (h * inlink->w) > INT_MAX ||
  228. (w * inlink->h) > INT_MAX)
  229. av_log(ctx, AV_LOG_ERROR, "Rescaled value for width or height is too big.\n");
  230. outlink->w = w;
  231. outlink->h = h;
  232. /* TODO: make algorithm configurable */
  233. scale->input_is_pal = desc->flags & PIX_FMT_PAL ||
  234. desc->flags & PIX_FMT_PSEUDOPAL;
  235. if (outfmt == AV_PIX_FMT_PAL8) outfmt = AV_PIX_FMT_BGR8;
  236. scale->output_is_pal = av_pix_fmt_desc_get(outfmt)->flags & PIX_FMT_PAL ||
  237. av_pix_fmt_desc_get(outfmt)->flags & PIX_FMT_PSEUDOPAL;
  238. if (scale->sws)
  239. sws_freeContext(scale->sws);
  240. if (inlink->w == outlink->w && inlink->h == outlink->h &&
  241. inlink->format == outlink->format)
  242. scale->sws = NULL;
  243. else {
  244. scale->sws = sws_getContext(inlink ->w, inlink ->h, inlink ->format,
  245. outlink->w, outlink->h, outfmt,
  246. scale->flags, NULL, NULL, NULL);
  247. if (scale->isws[0])
  248. sws_freeContext(scale->isws[0]);
  249. scale->isws[0] = sws_getContext(inlink ->w, inlink ->h/2, inlink ->format,
  250. outlink->w, outlink->h/2, outfmt,
  251. scale->flags, NULL, NULL, NULL);
  252. if (scale->isws[1])
  253. sws_freeContext(scale->isws[1]);
  254. scale->isws[1] = sws_getContext(inlink ->w, inlink ->h/2, inlink ->format,
  255. outlink->w, outlink->h/2, outfmt,
  256. scale->flags, NULL, NULL, NULL);
  257. if (!scale->sws || !scale->isws[0] || !scale->isws[1])
  258. return AVERROR(EINVAL);
  259. }
  260. if (inlink->sample_aspect_ratio.num){
  261. outlink->sample_aspect_ratio = av_mul_q((AVRational){outlink->h * inlink->w, outlink->w * inlink->h}, inlink->sample_aspect_ratio);
  262. } else
  263. outlink->sample_aspect_ratio = inlink->sample_aspect_ratio;
  264. 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",
  265. inlink ->w, inlink ->h, av_get_pix_fmt_name( inlink->format),
  266. inlink->sample_aspect_ratio.num, inlink->sample_aspect_ratio.den,
  267. outlink->w, outlink->h, av_get_pix_fmt_name(outlink->format),
  268. outlink->sample_aspect_ratio.num, outlink->sample_aspect_ratio.den,
  269. scale->flags);
  270. return 0;
  271. fail:
  272. av_log(NULL, AV_LOG_ERROR,
  273. "Error when evaluating the expression '%s'.\n"
  274. "Maybe the expression for out_w:'%s' or for out_h:'%s' is self-referencing.\n",
  275. expr, scale->w_expr, scale->h_expr);
  276. return ret;
  277. }
  278. static int scale_slice(AVFilterLink *link, AVFrame *out_buf, AVFrame *cur_pic, struct SwsContext *sws, int y, int h, int mul, int field)
  279. {
  280. ScaleContext *scale = link->dst->priv;
  281. const uint8_t *in[4];
  282. uint8_t *out[4];
  283. int in_stride[4],out_stride[4];
  284. int i;
  285. for(i=0; i<4; i++){
  286. int vsub= ((i+1)&2) ? scale->vsub : 0;
  287. in_stride[i] = cur_pic->linesize[i] * mul;
  288. out_stride[i] = out_buf->linesize[i] * mul;
  289. in[i] = cur_pic->data[i] + ((y>>vsub)+field) * cur_pic->linesize[i];
  290. out[i] = out_buf->data[i] + field * out_buf->linesize[i];
  291. }
  292. if(scale->input_is_pal)
  293. in[1] = cur_pic->data[1];
  294. if(scale->output_is_pal)
  295. out[1] = out_buf->data[1];
  296. return sws_scale(sws, in, in_stride, y/mul, h,
  297. out,out_stride);
  298. }
  299. static int filter_frame(AVFilterLink *link, AVFrame *in)
  300. {
  301. ScaleContext *scale = link->dst->priv;
  302. AVFilterLink *outlink = link->dst->outputs[0];
  303. AVFrame *out;
  304. const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(link->format);
  305. char buf[32];
  306. if( in->width != link->w
  307. || in->height != link->h
  308. || in->format != link->format) {
  309. int ret;
  310. snprintf(buf, sizeof(buf)-1, "%d", outlink->w);
  311. av_opt_set(scale, "w", buf, 0);
  312. snprintf(buf, sizeof(buf)-1, "%d", outlink->h);
  313. av_opt_set(scale, "h", buf, 0);
  314. link->dst->inputs[0]->format = in->format;
  315. link->dst->inputs[0]->w = in->width;
  316. link->dst->inputs[0]->h = in->height;
  317. if ((ret = config_props(outlink)) < 0)
  318. return ret;
  319. }
  320. if (!scale->sws)
  321. return ff_filter_frame(outlink, in);
  322. scale->hsub = desc->log2_chroma_w;
  323. scale->vsub = desc->log2_chroma_h;
  324. out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
  325. if (!out) {
  326. av_frame_free(&in);
  327. return AVERROR(ENOMEM);
  328. }
  329. av_frame_copy_props(out, in);
  330. out->width = outlink->w;
  331. out->height = outlink->h;
  332. if(scale->output_is_pal)
  333. avpriv_set_systematic_pal2((uint32_t*)out->data[1], outlink->format == AV_PIX_FMT_PAL8 ? AV_PIX_FMT_BGR8 : outlink->format);
  334. av_reduce(&out->sample_aspect_ratio.num, &out->sample_aspect_ratio.den,
  335. (int64_t)in->sample_aspect_ratio.num * outlink->h * link->w,
  336. (int64_t)in->sample_aspect_ratio.den * outlink->w * link->h,
  337. INT_MAX);
  338. if(scale->interlaced>0 || (scale->interlaced<0 && in->interlaced_frame)){
  339. scale_slice(link, out, in, scale->isws[0], 0, (link->h+1)/2, 2, 0);
  340. scale_slice(link, out, in, scale->isws[1], 0, link->h /2, 2, 1);
  341. }else{
  342. scale_slice(link, out, in, scale->sws, 0, link->h, 1, 0);
  343. }
  344. av_frame_free(&in);
  345. return ff_filter_frame(outlink, out);
  346. }
  347. #define OFFSET(x) offsetof(ScaleContext, x)
  348. #define FLAGS AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
  349. static const AVOption scale_options[] = {
  350. { "w", "Output video width", OFFSET(w_expr), AV_OPT_TYPE_STRING, { .str = "iw" }, .flags = FLAGS },
  351. { "width", "Output video width", OFFSET(w_expr), AV_OPT_TYPE_STRING, { .str = "iw" }, .flags = FLAGS },
  352. { "h", "Output video height", OFFSET(h_expr), AV_OPT_TYPE_STRING, { .str = "ih" }, .flags = FLAGS },
  353. { "height","Output video height", OFFSET(h_expr), AV_OPT_TYPE_STRING, { .str = "ih" }, .flags = FLAGS },
  354. { "flags", "Flags to pass to libswscale", OFFSET(flags_str), AV_OPT_TYPE_STRING, { .str = "bilinear" }, .flags = FLAGS },
  355. { "interl", "set interlacing", OFFSET(interlaced), AV_OPT_TYPE_INT, {.i64 = 0 }, -1, 1, FLAGS },
  356. { "size", "set video size", OFFSET(size_str), AV_OPT_TYPE_STRING, {.str = NULL}, 0, FLAGS },
  357. { "s", "set video size", OFFSET(size_str), AV_OPT_TYPE_STRING, {.str = NULL}, 0, FLAGS },
  358. { NULL },
  359. };
  360. AVFILTER_DEFINE_CLASS(scale);
  361. static const AVFilterPad avfilter_vf_scale_inputs[] = {
  362. {
  363. .name = "default",
  364. .type = AVMEDIA_TYPE_VIDEO,
  365. .filter_frame = filter_frame,
  366. },
  367. { NULL }
  368. };
  369. static const AVFilterPad avfilter_vf_scale_outputs[] = {
  370. {
  371. .name = "default",
  372. .type = AVMEDIA_TYPE_VIDEO,
  373. .config_props = config_props,
  374. },
  375. { NULL }
  376. };
  377. AVFilter avfilter_vf_scale = {
  378. .name = "scale",
  379. .description = NULL_IF_CONFIG_SMALL("Scale the input video to width:height size and/or convert the image format."),
  380. .init = init,
  381. .uninit = uninit,
  382. .query_formats = query_formats,
  383. .priv_size = sizeof(ScaleContext),
  384. .priv_class = &scale_class,
  385. .inputs = avfilter_vf_scale_inputs,
  386. .outputs = avfilter_vf_scale_outputs,
  387. };