73ea9d23e067323cddc17d70fe9a41de00822328
[libav.git] / libavfilter / vf_scale.c
1 /*
2 * Copyright (c) 2007 Bobby Bingham
3 *
4 * This file is part of Libav.
5 *
6 * Libav 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 * Libav 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 Libav; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 */
20
21 /**
22 * @file
23 * scale video filter
24 */
25
26 #include <stdio.h>
27 #include <string.h>
28
29 #include "avfilter.h"
30 #include "formats.h"
31 #include "internal.h"
32 #include "video.h"
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/pixdesc.h"
39 #include "libswscale/swscale.h"
40
41 static const char *const var_names[] = {
42 "PI",
43 "PHI",
44 "E",
45 "in_w", "iw",
46 "in_h", "ih",
47 "out_w", "ow",
48 "out_h", "oh",
49 "a", "dar",
50 "sar",
51 "hsub",
52 "vsub",
53 NULL
54 };
55
56 enum var_name {
57 VAR_PI,
58 VAR_PHI,
59 VAR_E,
60 VAR_IN_W, VAR_IW,
61 VAR_IN_H, VAR_IH,
62 VAR_OUT_W, VAR_OW,
63 VAR_OUT_H, VAR_OH,
64 VAR_A, VAR_DAR,
65 VAR_SAR,
66 VAR_HSUB,
67 VAR_VSUB,
68 VARS_NB
69 };
70
71 typedef struct ScaleContext {
72 const AVClass *class;
73 struct SwsContext *sws; ///< software scaler context
74
75 /**
76 * New dimensions. Special values are:
77 * 0 = original width/height
78 * -1 = keep original aspect
79 */
80 int w, h;
81 unsigned int flags; ///sws flags
82
83 int hsub, vsub; ///< chroma subsampling
84 int slice_y; ///< top of current output slice
85 int input_is_pal; ///< set to 1 if the input format is paletted
86
87 char *w_expr; ///< width expression string
88 char *h_expr; ///< height expression string
89 char *flags_str;
90 } ScaleContext;
91
92 static av_cold int init(AVFilterContext *ctx)
93 {
94 ScaleContext *scale = ctx->priv;
95
96 if (scale->flags_str) {
97 const AVClass *class = sws_get_class();
98 const AVOption *o = av_opt_find(&class, "sws_flags", NULL, 0,
99 AV_OPT_SEARCH_FAKE_OBJ);
100 int ret = av_opt_eval_flags(&class, o, scale->flags_str, &scale->flags);
101
102 if (ret < 0)
103 return ret;
104 }
105
106 return 0;
107 }
108
109 static av_cold void uninit(AVFilterContext *ctx)
110 {
111 ScaleContext *scale = ctx->priv;
112 sws_freeContext(scale->sws);
113 scale->sws = NULL;
114 }
115
116 static int query_formats(AVFilterContext *ctx)
117 {
118 AVFilterFormats *formats;
119 enum AVPixelFormat pix_fmt;
120 int ret;
121
122 if (ctx->inputs[0]) {
123 const AVPixFmtDescriptor *desc = NULL;
124 formats = NULL;
125 while ((desc = av_pix_fmt_desc_next(desc))) {
126 pix_fmt = av_pix_fmt_desc_get_id(desc);
127 if ((sws_isSupportedInput(pix_fmt) ||
128 sws_isSupportedEndiannessConversion(pix_fmt))
129 && (ret = ff_add_format(&formats, pix_fmt)) < 0) {
130 ff_formats_unref(&formats);
131 return ret;
132 }
133 }
134 ff_formats_ref(formats, &ctx->inputs[0]->out_formats);
135 }
136 if (ctx->outputs[0]) {
137 const AVPixFmtDescriptor *desc = NULL;
138 formats = NULL;
139 while ((desc = av_pix_fmt_desc_next(desc))) {
140 pix_fmt = av_pix_fmt_desc_get_id(desc);
141 if ((sws_isSupportedOutput(pix_fmt) ||
142 sws_isSupportedEndiannessConversion(pix_fmt))
143 && (ret = ff_add_format(&formats, pix_fmt)) < 0) {
144 ff_formats_unref(&formats);
145 return ret;
146 }
147 }
148 ff_formats_ref(formats, &ctx->outputs[0]->in_formats);
149 }
150
151 return 0;
152 }
153
154 static int config_props(AVFilterLink *outlink)
155 {
156 AVFilterContext *ctx = outlink->src;
157 AVFilterLink *inlink = outlink->src->inputs[0];
158 ScaleContext *scale = ctx->priv;
159 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(inlink->format);
160 int64_t w, h;
161 double var_values[VARS_NB], res;
162 char *expr;
163 int ret;
164
165 var_values[VAR_PI] = M_PI;
166 var_values[VAR_PHI] = M_PHI;
167 var_values[VAR_E] = M_E;
168 var_values[VAR_IN_W] = var_values[VAR_IW] = inlink->w;
169 var_values[VAR_IN_H] = var_values[VAR_IH] = inlink->h;
170 var_values[VAR_OUT_W] = var_values[VAR_OW] = NAN;
171 var_values[VAR_OUT_H] = var_values[VAR_OH] = NAN;
172 var_values[VAR_A] = (double) inlink->w / inlink->h;
173 var_values[VAR_SAR] = inlink->sample_aspect_ratio.num ?
174 (double) inlink->sample_aspect_ratio.num / inlink->sample_aspect_ratio.den : 1;
175 var_values[VAR_DAR] = var_values[VAR_A] * var_values[VAR_SAR];
176 var_values[VAR_HSUB] = 1 << desc->log2_chroma_w;
177 var_values[VAR_VSUB] = 1 << desc->log2_chroma_h;
178
179 /* evaluate width and height */
180 av_expr_parse_and_eval(&res, (expr = scale->w_expr),
181 var_names, var_values,
182 NULL, NULL, NULL, NULL, NULL, 0, ctx);
183 scale->w = var_values[VAR_OUT_W] = var_values[VAR_OW] = res;
184 if ((ret = av_expr_parse_and_eval(&res, (expr = scale->h_expr),
185 var_names, var_values,
186 NULL, NULL, NULL, NULL, NULL, 0, ctx)) < 0)
187 goto fail;
188 scale->h = var_values[VAR_OUT_H] = var_values[VAR_OH] = res;
189 /* evaluate again the width, as it may depend on the output height */
190 if ((ret = av_expr_parse_and_eval(&res, (expr = scale->w_expr),
191 var_names, var_values,
192 NULL, NULL, NULL, NULL, NULL, 0, ctx)) < 0)
193 goto fail;
194 scale->w = res;
195
196 w = scale->w;
197 h = scale->h;
198
199 /* sanity check params */
200 if (w < -1 || h < -1) {
201 av_log(ctx, AV_LOG_ERROR, "Size values less than -1 are not acceptable.\n");
202 return AVERROR(EINVAL);
203 }
204 if (w == -1 && h == -1)
205 scale->w = scale->h = 0;
206
207 if (!(w = scale->w))
208 w = inlink->w;
209 if (!(h = scale->h))
210 h = inlink->h;
211 if (w == -1)
212 w = av_rescale(h, inlink->w, inlink->h);
213 if (h == -1)
214 h = av_rescale(w, inlink->h, inlink->w);
215
216 if (w > INT_MAX || h > INT_MAX ||
217 (h * inlink->w) > INT_MAX ||
218 (w * inlink->h) > INT_MAX)
219 av_log(ctx, AV_LOG_ERROR, "Rescaled value for width or height is too big.\n");
220
221 outlink->w = w;
222 outlink->h = h;
223
224 /* TODO: make algorithm configurable */
225 av_log(ctx, AV_LOG_VERBOSE, "w:%d h:%d fmt:%s -> w:%d h:%d fmt:%s flags:0x%0x\n",
226 inlink ->w, inlink ->h, av_get_pix_fmt_name(inlink->format),
227 outlink->w, outlink->h, av_get_pix_fmt_name(outlink->format),
228 scale->flags);
229
230 scale->input_is_pal = desc->flags & AV_PIX_FMT_FLAG_PAL ||
231 desc->flags & AV_PIX_FMT_FLAG_PSEUDOPAL;
232
233 if (scale->sws)
234 sws_freeContext(scale->sws);
235 if (inlink->w == outlink->w && inlink->h == outlink->h &&
236 inlink->format == outlink->format)
237 scale->sws = NULL;
238 else {
239 scale->sws = sws_getContext(inlink ->w, inlink ->h, inlink ->format,
240 outlink->w, outlink->h, outlink->format,
241 scale->flags, NULL, NULL, NULL);
242 if (!scale->sws)
243 return AVERROR(EINVAL);
244 }
245
246
247 if (inlink->sample_aspect_ratio.num)
248 outlink->sample_aspect_ratio = av_mul_q((AVRational){outlink->h*inlink->w,
249 outlink->w*inlink->h},
250 inlink->sample_aspect_ratio);
251 else
252 outlink->sample_aspect_ratio = inlink->sample_aspect_ratio;
253
254 return 0;
255
256 fail:
257 av_log(NULL, AV_LOG_ERROR,
258 "Error when evaluating the expression '%s'\n", expr);
259 return ret;
260 }
261
262 static int filter_frame(AVFilterLink *link, AVFrame *in)
263 {
264 ScaleContext *scale = link->dst->priv;
265 AVFilterLink *outlink = link->dst->outputs[0];
266 AVFrame *out;
267 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(link->format);
268
269 if (!scale->sws)
270 return ff_filter_frame(outlink, in);
271
272 scale->hsub = desc->log2_chroma_w;
273 scale->vsub = desc->log2_chroma_h;
274
275 out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
276 if (!out) {
277 av_frame_free(&in);
278 return AVERROR(ENOMEM);
279 }
280
281 av_frame_copy_props(out, in);
282 out->width = outlink->w;
283 out->height = outlink->h;
284
285 av_reduce(&out->sample_aspect_ratio.num, &out->sample_aspect_ratio.den,
286 (int64_t)in->sample_aspect_ratio.num * outlink->h * link->w,
287 (int64_t)in->sample_aspect_ratio.den * outlink->w * link->h,
288 INT_MAX);
289
290 sws_scale(scale->sws, in->data, in->linesize, 0, in->height,
291 out->data, out->linesize);
292
293 av_frame_free(&in);
294 return ff_filter_frame(outlink, out);
295 }
296
297 #define OFFSET(x) offsetof(ScaleContext, x)
298 #define FLAGS AV_OPT_FLAG_VIDEO_PARAM
299 static const AVOption options[] = {
300 { "w", "Output video width", OFFSET(w_expr), AV_OPT_TYPE_STRING, { .str = "iw" }, .flags = FLAGS },
301 { "h", "Output video height", OFFSET(h_expr), AV_OPT_TYPE_STRING, { .str = "ih" }, .flags = FLAGS },
302 { "flags", "Flags to pass to libswscale", OFFSET(flags_str), AV_OPT_TYPE_STRING, { .str = "bilinear" }, .flags = FLAGS },
303 { NULL },
304 };
305
306 static const AVClass scale_class = {
307 .class_name = "scale",
308 .item_name = av_default_item_name,
309 .option = options,
310 .version = LIBAVUTIL_VERSION_INT,
311 };
312
313 static const AVFilterPad avfilter_vf_scale_inputs[] = {
314 {
315 .name = "default",
316 .type = AVMEDIA_TYPE_VIDEO,
317 .filter_frame = filter_frame,
318 },
319 { NULL }
320 };
321
322 static const AVFilterPad avfilter_vf_scale_outputs[] = {
323 {
324 .name = "default",
325 .type = AVMEDIA_TYPE_VIDEO,
326 .config_props = config_props,
327 },
328 { NULL }
329 };
330
331 AVFilter ff_vf_scale = {
332 .name = "scale",
333 .description = NULL_IF_CONFIG_SMALL("Scale the input video to width:height size and/or convert the image format."),
334
335 .init = init,
336 .uninit = uninit,
337
338 .query_formats = query_formats,
339
340 .priv_size = sizeof(ScaleContext),
341 .priv_class = &scale_class,
342
343 .inputs = avfilter_vf_scale_inputs,
344 .outputs = avfilter_vf_scale_outputs,
345 };