vf_scale.c 14 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396
  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 "avfilter.h"
  25. #include "libavutil/avstring.h"
  26. #include "libavutil/eval.h"
  27. #include "libavutil/mathematics.h"
  28. #include "libavutil/opt.h"
  29. #include "libavutil/pixdesc.h"
  30. #include "libavutil/imgutils.h"
  31. #include "libavutil/avassert.h"
  32. #include "libswscale/swscale.h"
  33. static const char *const var_names[] = {
  34. "in_w", "iw",
  35. "in_h", "ih",
  36. "out_w", "ow",
  37. "out_h", "oh",
  38. "a",
  39. "sar",
  40. "dar",
  41. "hsub",
  42. "vsub",
  43. NULL
  44. };
  45. enum var_name {
  46. VAR_IN_W, VAR_IW,
  47. VAR_IN_H, VAR_IH,
  48. VAR_OUT_W, VAR_OW,
  49. VAR_OUT_H, VAR_OH,
  50. VAR_A,
  51. VAR_SAR,
  52. VAR_DAR,
  53. VAR_HSUB,
  54. VAR_VSUB,
  55. VARS_NB
  56. };
  57. typedef struct {
  58. struct SwsContext *sws; ///< software scaler context
  59. struct SwsContext *isws[2]; ///< software scaler context for interlaced material
  60. /**
  61. * New dimensions. Special values are:
  62. * 0 = original width/height
  63. * -1 = keep original aspect
  64. */
  65. int w, h;
  66. unsigned int flags; ///sws flags
  67. int hsub, vsub; ///< chroma subsampling
  68. int slice_y; ///< top of current output slice
  69. int input_is_pal; ///< set to 1 if the input format is paletted
  70. int output_is_pal; ///< set to 1 if the output format is paletted
  71. int interlaced;
  72. char w_expr[256]; ///< width expression string
  73. char h_expr[256]; ///< height expression string
  74. } ScaleContext;
  75. static av_cold int init(AVFilterContext *ctx, const char *args, void *opaque)
  76. {
  77. ScaleContext *scale = ctx->priv;
  78. const char *p;
  79. av_strlcpy(scale->w_expr, "iw", sizeof(scale->w_expr));
  80. av_strlcpy(scale->h_expr, "ih", sizeof(scale->h_expr));
  81. scale->flags = SWS_BILINEAR;
  82. if (args) {
  83. sscanf(args, "%255[^:]:%255[^:]", scale->w_expr, scale->h_expr);
  84. p = strstr(args,"flags=");
  85. if (p) {
  86. const AVClass *class = sws_get_class();
  87. const AVOption *o = av_opt_find(&class, "sws_flags", NULL, 0,
  88. AV_OPT_SEARCH_FAKE_OBJ);
  89. int ret = av_opt_eval_flags(&class, o, p + 6, &scale->flags);
  90. if (ret < 0)
  91. return ret;
  92. }
  93. if(strstr(args,"interl=1")){
  94. scale->interlaced=1;
  95. }else if(strstr(args,"interl=-1"))
  96. scale->interlaced=-1;
  97. }
  98. return 0;
  99. }
  100. static av_cold void uninit(AVFilterContext *ctx)
  101. {
  102. ScaleContext *scale = ctx->priv;
  103. sws_freeContext(scale->sws);
  104. sws_freeContext(scale->isws[0]);
  105. sws_freeContext(scale->isws[1]);
  106. scale->sws = NULL;
  107. }
  108. static int query_formats(AVFilterContext *ctx)
  109. {
  110. AVFilterFormats *formats;
  111. enum PixelFormat pix_fmt;
  112. int ret;
  113. if (ctx->inputs[0]) {
  114. formats = NULL;
  115. for (pix_fmt = 0; pix_fmt < PIX_FMT_NB; pix_fmt++)
  116. if ( sws_isSupportedInput(pix_fmt)
  117. && (ret = avfilter_add_format(&formats, pix_fmt)) < 0) {
  118. avfilter_formats_unref(&formats);
  119. return ret;
  120. }
  121. avfilter_formats_ref(formats, &ctx->inputs[0]->out_formats);
  122. }
  123. if (ctx->outputs[0]) {
  124. formats = NULL;
  125. for (pix_fmt = 0; pix_fmt < PIX_FMT_NB; pix_fmt++)
  126. if ( (sws_isSupportedOutput(pix_fmt) || pix_fmt == PIX_FMT_PAL8)
  127. && (ret = avfilter_add_format(&formats, pix_fmt)) < 0) {
  128. avfilter_formats_unref(&formats);
  129. return ret;
  130. }
  131. avfilter_formats_ref(formats, &ctx->outputs[0]->in_formats);
  132. }
  133. return 0;
  134. }
  135. static int config_props(AVFilterLink *outlink)
  136. {
  137. AVFilterContext *ctx = outlink->src;
  138. AVFilterLink *inlink = outlink->src->inputs[0];
  139. enum PixelFormat outfmt = outlink->format;
  140. ScaleContext *scale = ctx->priv;
  141. int64_t w, h;
  142. double var_values[VARS_NB], res;
  143. char *expr;
  144. int ret;
  145. var_values[VAR_IN_W] = var_values[VAR_IW] = inlink->w;
  146. var_values[VAR_IN_H] = var_values[VAR_IH] = inlink->h;
  147. var_values[VAR_OUT_W] = var_values[VAR_OW] = NAN;
  148. var_values[VAR_OUT_H] = var_values[VAR_OH] = NAN;
  149. var_values[VAR_A] = (float) inlink->w / inlink->h;
  150. var_values[VAR_SAR] = inlink->sample_aspect_ratio.num ?
  151. (float) inlink->sample_aspect_ratio.num / inlink->sample_aspect_ratio.den : 1;
  152. var_values[VAR_DAR] = var_values[VAR_A] * var_values[VAR_SAR];
  153. var_values[VAR_HSUB] = 1<<av_pix_fmt_descriptors[inlink->format].log2_chroma_w;
  154. var_values[VAR_VSUB] = 1<<av_pix_fmt_descriptors[inlink->format].log2_chroma_h;
  155. /* evaluate width and height */
  156. av_expr_parse_and_eval(&res, (expr = scale->w_expr),
  157. var_names, var_values,
  158. NULL, NULL, NULL, NULL, NULL, 0, ctx);
  159. scale->w = var_values[VAR_OUT_W] = var_values[VAR_OW] = res;
  160. if ((ret = av_expr_parse_and_eval(&res, (expr = scale->h_expr),
  161. var_names, var_values,
  162. NULL, NULL, NULL, NULL, NULL, 0, ctx)) < 0)
  163. goto fail;
  164. scale->h = var_values[VAR_OUT_H] = var_values[VAR_OH] = res;
  165. /* evaluate again the width, as it may depend on the output height */
  166. if ((ret = av_expr_parse_and_eval(&res, (expr = scale->w_expr),
  167. var_names, var_values,
  168. NULL, NULL, NULL, NULL, NULL, 0, ctx)) < 0)
  169. goto fail;
  170. scale->w = res;
  171. w = scale->w;
  172. h = scale->h;
  173. /* sanity check params */
  174. if (w < -1 || h < -1) {
  175. av_log(ctx, AV_LOG_ERROR, "Size values less than -1 are not acceptable.\n");
  176. return AVERROR(EINVAL);
  177. }
  178. if (w == -1 && h == -1)
  179. scale->w = scale->h = 0;
  180. if (!(w = scale->w))
  181. w = inlink->w;
  182. if (!(h = scale->h))
  183. h = inlink->h;
  184. if (w == -1)
  185. w = av_rescale(h, inlink->w, inlink->h);
  186. if (h == -1)
  187. h = av_rescale(w, inlink->h, inlink->w);
  188. if (w > INT_MAX || h > INT_MAX ||
  189. (h * inlink->w) > INT_MAX ||
  190. (w * inlink->h) > INT_MAX)
  191. av_log(ctx, AV_LOG_ERROR, "Rescaled value for width or height is too big.\n");
  192. outlink->w = w;
  193. outlink->h = h;
  194. /* TODO: make algorithm configurable */
  195. scale->input_is_pal = av_pix_fmt_descriptors[inlink->format].flags & PIX_FMT_PAL ||
  196. av_pix_fmt_descriptors[inlink->format].flags & PIX_FMT_PSEUDOPAL;
  197. if (outfmt == PIX_FMT_PAL8) outfmt = PIX_FMT_BGR8;
  198. scale->output_is_pal = av_pix_fmt_descriptors[outfmt].flags & PIX_FMT_PAL ||
  199. av_pix_fmt_descriptors[outfmt].flags & PIX_FMT_PSEUDOPAL;
  200. if (scale->sws)
  201. sws_freeContext(scale->sws);
  202. if (inlink->w == outlink->w && inlink->h == outlink->h &&
  203. inlink->format == outlink->format)
  204. scale->sws = NULL;
  205. else {
  206. scale->sws = sws_getContext(inlink ->w, inlink ->h, inlink ->format,
  207. outlink->w, outlink->h, outfmt,
  208. scale->flags, NULL, NULL, NULL);
  209. if (scale->isws[0])
  210. sws_freeContext(scale->isws[0]);
  211. scale->isws[0] = sws_getContext(inlink ->w, inlink ->h/2, inlink ->format,
  212. outlink->w, outlink->h/2, outfmt,
  213. scale->flags, NULL, NULL, NULL);
  214. if (scale->isws[1])
  215. sws_freeContext(scale->isws[1]);
  216. scale->isws[1] = sws_getContext(inlink ->w, inlink ->h/2, inlink ->format,
  217. outlink->w, outlink->h/2, outfmt,
  218. scale->flags, NULL, NULL, NULL);
  219. if (!scale->sws || !scale->isws[0] || !scale->isws[1])
  220. return AVERROR(EINVAL);
  221. }
  222. if (inlink->sample_aspect_ratio.num){
  223. outlink->sample_aspect_ratio = av_mul_q((AVRational){outlink->h * inlink->w, outlink->w * inlink->h}, inlink->sample_aspect_ratio);
  224. } else
  225. outlink->sample_aspect_ratio = inlink->sample_aspect_ratio;
  226. av_log(ctx, AV_LOG_INFO, "w:%d h:%d fmt:%s sar:%d/%d -> w:%d h:%d fmt:%s sar:%d/%d flags:0x%0x\n",
  227. inlink ->w, inlink ->h, av_pix_fmt_descriptors[ inlink->format].name,
  228. inlink->sample_aspect_ratio.num, inlink->sample_aspect_ratio.den,
  229. outlink->w, outlink->h, av_pix_fmt_descriptors[outlink->format].name,
  230. outlink->sample_aspect_ratio.num, outlink->sample_aspect_ratio.den,
  231. scale->flags);
  232. return 0;
  233. fail:
  234. av_log(NULL, AV_LOG_ERROR,
  235. "Error when evaluating the expression '%s'.\n"
  236. "Maybe the expression for out_w:'%s' or for out_h:'%s' is self-referencing.\n",
  237. expr, scale->w_expr, scale->h_expr);
  238. return ret;
  239. }
  240. static void start_frame(AVFilterLink *link, AVFilterBufferRef *picref)
  241. {
  242. ScaleContext *scale = link->dst->priv;
  243. AVFilterLink *outlink = link->dst->outputs[0];
  244. AVFilterBufferRef *outpicref;
  245. if( picref->video->w != link->w
  246. || picref->video->h != link->h
  247. || picref->format != link->format) {
  248. AVFilterLink *out_link;
  249. int ret;
  250. snprintf(scale->w_expr, sizeof(scale->w_expr)-1, "%d", outlink->w);
  251. snprintf(scale->h_expr, sizeof(scale->h_expr)-1, "%d", outlink->h);
  252. link->dst->inputs[0]->format = picref->format;
  253. link->dst->inputs[0]->w = picref->video->w;
  254. link->dst->inputs[0]->h = picref->video->h;
  255. if ((ret = config_props(outlink)) < 0)
  256. av_assert0(0); //what to do here ?
  257. }
  258. if (!scale->sws) {
  259. avfilter_start_frame(outlink, avfilter_ref_buffer(picref, ~0));
  260. return;
  261. }
  262. scale->hsub = av_pix_fmt_descriptors[link->format].log2_chroma_w;
  263. scale->vsub = av_pix_fmt_descriptors[link->format].log2_chroma_h;
  264. outpicref = avfilter_get_video_buffer(outlink, AV_PERM_WRITE|AV_PERM_ALIGN, outlink->w, outlink->h);
  265. avfilter_copy_buffer_ref_props(outpicref, picref);
  266. outpicref->video->w = outlink->w;
  267. outpicref->video->h = outlink->h;
  268. outlink->out_buf = outpicref;
  269. if(scale->output_is_pal)
  270. ff_set_systematic_pal2(outpicref->data[1], outlink->format == PIX_FMT_PAL8 ? PIX_FMT_BGR8 : outlink->format);
  271. av_reduce(&outpicref->video->sample_aspect_ratio.num, &outpicref->video->sample_aspect_ratio.den,
  272. (int64_t)picref->video->sample_aspect_ratio.num * outlink->h * link->w,
  273. (int64_t)picref->video->sample_aspect_ratio.den * outlink->w * link->h,
  274. INT_MAX);
  275. scale->slice_y = 0;
  276. avfilter_start_frame(outlink, avfilter_ref_buffer(outpicref, ~0));
  277. }
  278. static int scale_slice(AVFilterLink *link, struct SwsContext *sws, int y, int h, int mul, int field)
  279. {
  280. ScaleContext *scale = link->dst->priv;
  281. AVFilterBufferRef *cur_pic = link->cur_buf;
  282. AVFilterBufferRef *out_buf = link->dst->outputs[0]->out_buf;
  283. const uint8_t *in[4];
  284. uint8_t *out[4];
  285. int in_stride[4],out_stride[4];
  286. int i;
  287. for(i=0; i<4; i++){
  288. int vsub= ((i+1)&2) ? scale->vsub : 0;
  289. in_stride[i] = cur_pic->linesize[i] * mul;
  290. out_stride[i] = out_buf->linesize[i] * mul;
  291. in[i] = cur_pic->data[i] + ((y>>vsub)+field) * cur_pic->linesize[i];
  292. out[i] = out_buf->data[i] + field * out_buf->linesize[i];
  293. }
  294. if(scale->input_is_pal)
  295. in[1] = cur_pic->data[1];
  296. if(scale->output_is_pal)
  297. out[1] = out_buf->data[1];
  298. return sws_scale(sws, in, in_stride, y/mul, h,
  299. out,out_stride);
  300. }
  301. static void draw_slice(AVFilterLink *link, int y, int h, int slice_dir)
  302. {
  303. ScaleContext *scale = link->dst->priv;
  304. int out_h;
  305. if (!scale->sws) {
  306. avfilter_draw_slice(link->dst->outputs[0], y, h, slice_dir);
  307. return;
  308. }
  309. if (scale->slice_y == 0 && slice_dir == -1)
  310. scale->slice_y = link->dst->outputs[0]->h;
  311. if(scale->interlaced>0 || (scale->interlaced<0 && link->cur_buf->video->interlaced)){
  312. av_assert0(y%(2<<scale->vsub) == 0);
  313. out_h = scale_slice(link, scale->isws[0], y, (h+1)/2, 2, 0);
  314. out_h+= scale_slice(link, scale->isws[1], y, h /2, 2, 1);
  315. }else{
  316. out_h = scale_slice(link, scale->sws, y, h, 1, 0);
  317. }
  318. if (slice_dir == -1)
  319. scale->slice_y -= out_h;
  320. avfilter_draw_slice(link->dst->outputs[0], scale->slice_y, out_h, slice_dir);
  321. if (slice_dir == 1)
  322. scale->slice_y += out_h;
  323. }
  324. AVFilter avfilter_vf_scale = {
  325. .name = "scale",
  326. .description = NULL_IF_CONFIG_SMALL("Scale the input video to width:height size and/or convert the image format."),
  327. .init = init,
  328. .uninit = uninit,
  329. .query_formats = query_formats,
  330. .priv_size = sizeof(ScaleContext),
  331. .inputs = (const AVFilterPad[]) {{ .name = "default",
  332. .type = AVMEDIA_TYPE_VIDEO,
  333. .start_frame = start_frame,
  334. .draw_slice = draw_slice,
  335. .min_perms = AV_PERM_READ, },
  336. { .name = NULL}},
  337. .outputs = (const AVFilterPad[]) {{ .name = "default",
  338. .type = AVMEDIA_TYPE_VIDEO,
  339. .config_props = config_props, },
  340. { .name = NULL}},
  341. };