blob: 85ea892d0133c37bb633660ae02b7646cdf4818c
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 | /** |
22 | * @file |
23 | * video crop filter |
24 | */ |
25 | |
26 | #include <stdio.h> |
27 | |
28 | #include "avfilter.h" |
29 | #include "formats.h" |
30 | #include "internal.h" |
31 | #include "video.h" |
32 | #include "libavutil/eval.h" |
33 | #include "libavutil/avstring.h" |
34 | #include "libavutil/internal.h" |
35 | #include "libavutil/libm.h" |
36 | #include "libavutil/imgutils.h" |
37 | #include "libavutil/mathematics.h" |
38 | #include "libavutil/opt.h" |
39 | |
40 | static const char *const var_names[] = { |
41 | "in_w", "iw", ///< width of the input video |
42 | "in_h", "ih", ///< height of the input video |
43 | "out_w", "ow", ///< width of the cropped video |
44 | "out_h", "oh", ///< height of the cropped video |
45 | "a", |
46 | "sar", |
47 | "dar", |
48 | "hsub", |
49 | "vsub", |
50 | "x", |
51 | "y", |
52 | "n", ///< number of frame |
53 | "pos", ///< position in the file |
54 | "t", ///< timestamp expressed in seconds |
55 | NULL |
56 | }; |
57 | |
58 | enum var_name { |
59 | VAR_IN_W, VAR_IW, |
60 | VAR_IN_H, VAR_IH, |
61 | VAR_OUT_W, VAR_OW, |
62 | VAR_OUT_H, VAR_OH, |
63 | VAR_A, |
64 | VAR_SAR, |
65 | VAR_DAR, |
66 | VAR_HSUB, |
67 | VAR_VSUB, |
68 | VAR_X, |
69 | VAR_Y, |
70 | VAR_N, |
71 | VAR_POS, |
72 | VAR_T, |
73 | VAR_VARS_NB |
74 | }; |
75 | |
76 | typedef struct CropContext { |
77 | const AVClass *class; |
78 | int x; ///< x offset of the non-cropped area with respect to the input area |
79 | int y; ///< y offset of the non-cropped area with respect to the input area |
80 | int w; ///< width of the cropped area |
81 | int h; ///< height of the cropped area |
82 | |
83 | AVRational out_sar; ///< output sample aspect ratio |
84 | int keep_aspect; ///< keep display aspect ratio when cropping |
85 | int exact; ///< exact cropping, for subsampled formats |
86 | |
87 | int max_step[4]; ///< max pixel step for each plane, expressed as a number of bytes |
88 | int hsub, vsub; ///< chroma subsampling |
89 | char *x_expr, *y_expr, *w_expr, *h_expr; |
90 | AVExpr *x_pexpr, *y_pexpr; /* parsed expressions for x and y */ |
91 | double var_values[VAR_VARS_NB]; |
92 | } CropContext; |
93 | |
94 | static int query_formats(AVFilterContext *ctx) |
95 | { |
96 | AVFilterFormats *formats = NULL; |
97 | int fmt, ret; |
98 | |
99 | for (fmt = 0; av_pix_fmt_desc_get(fmt); fmt++) { |
100 | const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(fmt); |
101 | if (!(desc->flags & (AV_PIX_FMT_FLAG_HWACCEL | AV_PIX_FMT_FLAG_BITSTREAM)) && |
102 | !((desc->log2_chroma_w || desc->log2_chroma_h) && !(desc->flags & AV_PIX_FMT_FLAG_PLANAR)) && |
103 | (ret = ff_add_format(&formats, fmt)) < 0) |
104 | return ret; |
105 | } |
106 | |
107 | return ff_set_common_formats(ctx, formats); |
108 | } |
109 | |
110 | static av_cold void uninit(AVFilterContext *ctx) |
111 | { |
112 | CropContext *s = ctx->priv; |
113 | |
114 | av_expr_free(s->x_pexpr); |
115 | s->x_pexpr = NULL; |
116 | av_expr_free(s->y_pexpr); |
117 | s->y_pexpr = NULL; |
118 | } |
119 | |
120 | static inline int normalize_double(int *n, double d) |
121 | { |
122 | int ret = 0; |
123 | |
124 | if (isnan(d)) { |
125 | ret = AVERROR(EINVAL); |
126 | } else if (d > INT_MAX || d < INT_MIN) { |
127 | *n = d > INT_MAX ? INT_MAX : INT_MIN; |
128 | ret = AVERROR(EINVAL); |
129 | } else |
130 | *n = lrint(d); |
131 | |
132 | return ret; |
133 | } |
134 | |
135 | static int config_input(AVFilterLink *link) |
136 | { |
137 | AVFilterContext *ctx = link->dst; |
138 | CropContext *s = ctx->priv; |
139 | const AVPixFmtDescriptor *pix_desc = av_pix_fmt_desc_get(link->format); |
140 | int ret; |
141 | const char *expr; |
142 | double res; |
143 | |
144 | s->var_values[VAR_IN_W] = s->var_values[VAR_IW] = ctx->inputs[0]->w; |
145 | s->var_values[VAR_IN_H] = s->var_values[VAR_IH] = ctx->inputs[0]->h; |
146 | s->var_values[VAR_A] = (float) link->w / link->h; |
147 | s->var_values[VAR_SAR] = link->sample_aspect_ratio.num ? av_q2d(link->sample_aspect_ratio) : 1; |
148 | s->var_values[VAR_DAR] = s->var_values[VAR_A] * s->var_values[VAR_SAR]; |
149 | s->var_values[VAR_HSUB] = 1<<pix_desc->log2_chroma_w; |
150 | s->var_values[VAR_VSUB] = 1<<pix_desc->log2_chroma_h; |
151 | s->var_values[VAR_X] = NAN; |
152 | s->var_values[VAR_Y] = NAN; |
153 | s->var_values[VAR_OUT_W] = s->var_values[VAR_OW] = NAN; |
154 | s->var_values[VAR_OUT_H] = s->var_values[VAR_OH] = NAN; |
155 | s->var_values[VAR_N] = 0; |
156 | s->var_values[VAR_T] = NAN; |
157 | s->var_values[VAR_POS] = NAN; |
158 | |
159 | av_image_fill_max_pixsteps(s->max_step, NULL, pix_desc); |
160 | s->hsub = pix_desc->log2_chroma_w; |
161 | s->vsub = pix_desc->log2_chroma_h; |
162 | |
163 | if ((ret = av_expr_parse_and_eval(&res, (expr = s->w_expr), |
164 | var_names, s->var_values, |
165 | NULL, NULL, NULL, NULL, NULL, 0, ctx)) < 0) |
166 | goto fail_expr; |
167 | s->var_values[VAR_OUT_W] = s->var_values[VAR_OW] = res; |
168 | if ((ret = av_expr_parse_and_eval(&res, (expr = s->h_expr), |
169 | var_names, s->var_values, |
170 | NULL, NULL, NULL, NULL, NULL, 0, ctx)) < 0) |
171 | goto fail_expr; |
172 | s->var_values[VAR_OUT_H] = s->var_values[VAR_OH] = res; |
173 | /* evaluate again ow as it may depend on oh */ |
174 | if ((ret = av_expr_parse_and_eval(&res, (expr = s->w_expr), |
175 | var_names, s->var_values, |
176 | NULL, NULL, NULL, NULL, NULL, 0, ctx)) < 0) |
177 | goto fail_expr; |
178 | |
179 | s->var_values[VAR_OUT_W] = s->var_values[VAR_OW] = res; |
180 | if (normalize_double(&s->w, s->var_values[VAR_OUT_W]) < 0 || |
181 | normalize_double(&s->h, s->var_values[VAR_OUT_H]) < 0) { |
182 | av_log(ctx, AV_LOG_ERROR, |
183 | "Too big value or invalid expression for out_w/ow or out_h/oh. " |
184 | "Maybe the expression for out_w:'%s' or for out_h:'%s' is self-referencing.\n", |
185 | s->w_expr, s->h_expr); |
186 | return AVERROR(EINVAL); |
187 | } |
188 | |
189 | if (!s->exact) { |
190 | s->w &= ~((1 << s->hsub) - 1); |
191 | s->h &= ~((1 << s->vsub) - 1); |
192 | } |
193 | |
194 | av_expr_free(s->x_pexpr); |
195 | av_expr_free(s->y_pexpr); |
196 | s->x_pexpr = s->y_pexpr = NULL; |
197 | if ((ret = av_expr_parse(&s->x_pexpr, s->x_expr, var_names, |
198 | NULL, NULL, NULL, NULL, 0, ctx)) < 0 || |
199 | (ret = av_expr_parse(&s->y_pexpr, s->y_expr, var_names, |
200 | NULL, NULL, NULL, NULL, 0, ctx)) < 0) |
201 | return AVERROR(EINVAL); |
202 | |
203 | if (s->keep_aspect) { |
204 | AVRational dar = av_mul_q(link->sample_aspect_ratio, |
205 | (AVRational){ link->w, link->h }); |
206 | av_reduce(&s->out_sar.num, &s->out_sar.den, |
207 | dar.num * s->h, dar.den * s->w, INT_MAX); |
208 | } else |
209 | s->out_sar = link->sample_aspect_ratio; |
210 | |
211 | av_log(ctx, AV_LOG_VERBOSE, "w:%d h:%d sar:%d/%d -> w:%d h:%d sar:%d/%d\n", |
212 | link->w, link->h, link->sample_aspect_ratio.num, link->sample_aspect_ratio.den, |
213 | s->w, s->h, s->out_sar.num, s->out_sar.den); |
214 | |
215 | if (s->w <= 0 || s->h <= 0 || |
216 | s->w > link->w || s->h > link->h) { |
217 | av_log(ctx, AV_LOG_ERROR, |
218 | "Invalid too big or non positive size for width '%d' or height '%d'\n", |
219 | s->w, s->h); |
220 | return AVERROR(EINVAL); |
221 | } |
222 | |
223 | /* set default, required in the case the first computed value for x/y is NAN */ |
224 | s->x = (link->w - s->w) / 2; |
225 | s->y = (link->h - s->h) / 2; |
226 | if (!s->exact) { |
227 | s->x &= ~((1 << s->hsub) - 1); |
228 | s->y &= ~((1 << s->vsub) - 1); |
229 | } |
230 | return 0; |
231 | |
232 | fail_expr: |
233 | av_log(NULL, AV_LOG_ERROR, "Error when evaluating the expression '%s'\n", expr); |
234 | return ret; |
235 | } |
236 | |
237 | static int config_output(AVFilterLink *link) |
238 | { |
239 | CropContext *s = link->src->priv; |
240 | |
241 | link->w = s->w; |
242 | link->h = s->h; |
243 | link->sample_aspect_ratio = s->out_sar; |
244 | |
245 | return 0; |
246 | } |
247 | |
248 | static int filter_frame(AVFilterLink *link, AVFrame *frame) |
249 | { |
250 | AVFilterContext *ctx = link->dst; |
251 | CropContext *s = ctx->priv; |
252 | const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(link->format); |
253 | int i; |
254 | |
255 | frame->width = s->w; |
256 | frame->height = s->h; |
257 | |
258 | s->var_values[VAR_N] = link->frame_count_out; |
259 | s->var_values[VAR_T] = frame->pts == AV_NOPTS_VALUE ? |
260 | NAN : frame->pts * av_q2d(link->time_base); |
261 | s->var_values[VAR_POS] = av_frame_get_pkt_pos(frame) == -1 ? |
262 | NAN : av_frame_get_pkt_pos(frame); |
263 | s->var_values[VAR_X] = av_expr_eval(s->x_pexpr, s->var_values, NULL); |
264 | s->var_values[VAR_Y] = av_expr_eval(s->y_pexpr, s->var_values, NULL); |
265 | s->var_values[VAR_X] = av_expr_eval(s->x_pexpr, s->var_values, NULL); |
266 | |
267 | normalize_double(&s->x, s->var_values[VAR_X]); |
268 | normalize_double(&s->y, s->var_values[VAR_Y]); |
269 | |
270 | if (s->x < 0) |
271 | s->x = 0; |
272 | if (s->y < 0) |
273 | s->y = 0; |
274 | if ((unsigned)s->x + (unsigned)s->w > link->w) |
275 | s->x = link->w - s->w; |
276 | if ((unsigned)s->y + (unsigned)s->h > link->h) |
277 | s->y = link->h - s->h; |
278 | if (!s->exact) { |
279 | s->x &= ~((1 << s->hsub) - 1); |
280 | s->y &= ~((1 << s->vsub) - 1); |
281 | } |
282 | |
283 | av_log(ctx, AV_LOG_TRACE, "n:%d t:%f pos:%f x:%d y:%d x+w:%d y+h:%d\n", |
284 | (int)s->var_values[VAR_N], s->var_values[VAR_T], s->var_values[VAR_POS], |
285 | s->x, s->y, s->x+s->w, s->y+s->h); |
286 | |
287 | frame->data[0] += s->y * frame->linesize[0]; |
288 | frame->data[0] += s->x * s->max_step[0]; |
289 | |
290 | if (!(desc->flags & AV_PIX_FMT_FLAG_PAL || desc->flags & AV_PIX_FMT_FLAG_PSEUDOPAL)) { |
291 | for (i = 1; i < 3; i ++) { |
292 | if (frame->data[i]) { |
293 | frame->data[i] += (s->y >> s->vsub) * frame->linesize[i]; |
294 | frame->data[i] += (s->x * s->max_step[i]) >> s->hsub; |
295 | } |
296 | } |
297 | } |
298 | |
299 | /* alpha plane */ |
300 | if (frame->data[3]) { |
301 | frame->data[3] += s->y * frame->linesize[3]; |
302 | frame->data[3] += s->x * s->max_step[3]; |
303 | } |
304 | |
305 | return ff_filter_frame(link->dst->outputs[0], frame); |
306 | } |
307 | |
308 | static int process_command(AVFilterContext *ctx, const char *cmd, const char *args, |
309 | char *res, int res_len, int flags) |
310 | { |
311 | CropContext *s = ctx->priv; |
312 | int ret; |
313 | |
314 | if ( !strcmp(cmd, "out_w") || !strcmp(cmd, "w") |
315 | || !strcmp(cmd, "out_h") || !strcmp(cmd, "h") |
316 | || !strcmp(cmd, "x") || !strcmp(cmd, "y")) { |
317 | |
318 | int old_x = s->x; |
319 | int old_y = s->y; |
320 | int old_w = s->w; |
321 | int old_h = s->h; |
322 | |
323 | AVFilterLink *outlink = ctx->outputs[0]; |
324 | AVFilterLink *inlink = ctx->inputs[0]; |
325 | |
326 | av_opt_set(s, cmd, args, 0); |
327 | |
328 | if ((ret = config_input(inlink)) < 0) { |
329 | s->x = old_x; |
330 | s->y = old_y; |
331 | s->w = old_w; |
332 | s->h = old_h; |
333 | return ret; |
334 | } |
335 | |
336 | ret = config_output(outlink); |
337 | |
338 | } else |
339 | ret = AVERROR(ENOSYS); |
340 | |
341 | return ret; |
342 | } |
343 | |
344 | #define OFFSET(x) offsetof(CropContext, x) |
345 | #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM |
346 | |
347 | static const AVOption crop_options[] = { |
348 | { "out_w", "set the width crop area expression", OFFSET(w_expr), AV_OPT_TYPE_STRING, {.str = "iw"}, CHAR_MIN, CHAR_MAX, FLAGS }, |
349 | { "w", "set the width crop area expression", OFFSET(w_expr), AV_OPT_TYPE_STRING, {.str = "iw"}, CHAR_MIN, CHAR_MAX, FLAGS }, |
350 | { "out_h", "set the height crop area expression", OFFSET(h_expr), AV_OPT_TYPE_STRING, {.str = "ih"}, CHAR_MIN, CHAR_MAX, FLAGS }, |
351 | { "h", "set the height crop area expression", OFFSET(h_expr), AV_OPT_TYPE_STRING, {.str = "ih"}, CHAR_MIN, CHAR_MAX, FLAGS }, |
352 | { "x", "set the x crop area expression", OFFSET(x_expr), AV_OPT_TYPE_STRING, {.str = "(in_w-out_w)/2"}, CHAR_MIN, CHAR_MAX, FLAGS }, |
353 | { "y", "set the y crop area expression", OFFSET(y_expr), AV_OPT_TYPE_STRING, {.str = "(in_h-out_h)/2"}, CHAR_MIN, CHAR_MAX, FLAGS }, |
354 | { "keep_aspect", "keep aspect ratio", OFFSET(keep_aspect), AV_OPT_TYPE_BOOL, {.i64=0}, 0, 1, FLAGS }, |
355 | { "exact", "do exact cropping", OFFSET(exact), AV_OPT_TYPE_BOOL, {.i64=0}, 0, 1, FLAGS }, |
356 | { NULL } |
357 | }; |
358 | |
359 | AVFILTER_DEFINE_CLASS(crop); |
360 | |
361 | static const AVFilterPad avfilter_vf_crop_inputs[] = { |
362 | { |
363 | .name = "default", |
364 | .type = AVMEDIA_TYPE_VIDEO, |
365 | .filter_frame = filter_frame, |
366 | .config_props = config_input, |
367 | }, |
368 | { NULL } |
369 | }; |
370 | |
371 | static const AVFilterPad avfilter_vf_crop_outputs[] = { |
372 | { |
373 | .name = "default", |
374 | .type = AVMEDIA_TYPE_VIDEO, |
375 | .config_props = config_output, |
376 | }, |
377 | { NULL } |
378 | }; |
379 | |
380 | AVFilter ff_vf_crop = { |
381 | .name = "crop", |
382 | .description = NULL_IF_CONFIG_SMALL("Crop the input video."), |
383 | .priv_size = sizeof(CropContext), |
384 | .priv_class = &crop_class, |
385 | .query_formats = query_formats, |
386 | .uninit = uninit, |
387 | .inputs = avfilter_vf_crop_inputs, |
388 | .outputs = avfilter_vf_crop_outputs, |
389 | .process_command = process_command, |
390 | }; |
391 |