| 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 | |
| 86 | int max_step[4]; ///< max pixel step for each plane, expressed as a number of bytes |
| 87 | int hsub, vsub; ///< chroma subsampling |
| 88 | char *x_expr, *y_expr, *w_expr, *h_expr; |
| 89 | AVExpr *x_pexpr, *y_pexpr; /* parsed expressions for x and y */ |
| 90 | double var_values[VAR_VARS_NB]; |
| 91 | } CropContext; |
| 92 | |
| 93 | static int query_formats(AVFilterContext *ctx) |
| 94 | { |
| 95 | AVFilterFormats *formats = NULL; |
| 96 | int fmt; |
| 97 | |
| 98 | for (fmt = 0; av_pix_fmt_desc_get(fmt); fmt++) { |
| 99 | const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(fmt); |
| 100 | if (!(desc->flags & (AV_PIX_FMT_FLAG_HWACCEL | AV_PIX_FMT_FLAG_BITSTREAM)) && |
| 101 | !((desc->log2_chroma_w || desc->log2_chroma_h) && !(desc->flags & AV_PIX_FMT_FLAG_PLANAR))) |
| 102 | ff_add_format(&formats, fmt); |
| 103 | } |
| 104 | |
| 105 | ff_set_common_formats(ctx, formats); |
| 106 | return 0; |
| 107 | } |
| 108 | |
| 109 | static av_cold void uninit(AVFilterContext *ctx) |
| 110 | { |
| 111 | CropContext *s = ctx->priv; |
| 112 | |
| 113 | av_expr_free(s->x_pexpr); |
| 114 | s->x_pexpr = NULL; |
| 115 | av_expr_free(s->y_pexpr); |
| 116 | s->y_pexpr = NULL; |
| 117 | } |
| 118 | |
| 119 | static inline int normalize_double(int *n, double d) |
| 120 | { |
| 121 | int ret = 0; |
| 122 | |
| 123 | if (isnan(d)) { |
| 124 | ret = AVERROR(EINVAL); |
| 125 | } else if (d > INT_MAX || d < INT_MIN) { |
| 126 | *n = d > INT_MAX ? INT_MAX : INT_MIN; |
| 127 | ret = AVERROR(EINVAL); |
| 128 | } else |
| 129 | *n = round(d); |
| 130 | |
| 131 | return ret; |
| 132 | } |
| 133 | |
| 134 | static int config_input(AVFilterLink *link) |
| 135 | { |
| 136 | AVFilterContext *ctx = link->dst; |
| 137 | CropContext *s = ctx->priv; |
| 138 | const AVPixFmtDescriptor *pix_desc = av_pix_fmt_desc_get(link->format); |
| 139 | int ret; |
| 140 | const char *expr; |
| 141 | double res; |
| 142 | |
| 143 | s->var_values[VAR_IN_W] = s->var_values[VAR_IW] = ctx->inputs[0]->w; |
| 144 | s->var_values[VAR_IN_H] = s->var_values[VAR_IH] = ctx->inputs[0]->h; |
| 145 | s->var_values[VAR_A] = (float) link->w / link->h; |
| 146 | s->var_values[VAR_SAR] = link->sample_aspect_ratio.num ? av_q2d(link->sample_aspect_ratio) : 1; |
| 147 | s->var_values[VAR_DAR] = s->var_values[VAR_A] * s->var_values[VAR_SAR]; |
| 148 | s->var_values[VAR_HSUB] = 1<<pix_desc->log2_chroma_w; |
| 149 | s->var_values[VAR_VSUB] = 1<<pix_desc->log2_chroma_h; |
| 150 | s->var_values[VAR_X] = NAN; |
| 151 | s->var_values[VAR_Y] = NAN; |
| 152 | s->var_values[VAR_OUT_W] = s->var_values[VAR_OW] = NAN; |
| 153 | s->var_values[VAR_OUT_H] = s->var_values[VAR_OH] = NAN; |
| 154 | s->var_values[VAR_N] = 0; |
| 155 | s->var_values[VAR_T] = NAN; |
| 156 | s->var_values[VAR_POS] = NAN; |
| 157 | |
| 158 | av_image_fill_max_pixsteps(s->max_step, NULL, pix_desc); |
| 159 | s->hsub = pix_desc->log2_chroma_w; |
| 160 | s->vsub = pix_desc->log2_chroma_h; |
| 161 | |
| 162 | if ((ret = av_expr_parse_and_eval(&res, (expr = s->w_expr), |
| 163 | var_names, s->var_values, |
| 164 | NULL, NULL, NULL, NULL, NULL, 0, ctx)) < 0) |
| 165 | goto fail_expr; |
| 166 | s->var_values[VAR_OUT_W] = s->var_values[VAR_OW] = res; |
| 167 | if ((ret = av_expr_parse_and_eval(&res, (expr = s->h_expr), |
| 168 | var_names, s->var_values, |
| 169 | NULL, NULL, NULL, NULL, NULL, 0, ctx)) < 0) |
| 170 | goto fail_expr; |
| 171 | s->var_values[VAR_OUT_H] = s->var_values[VAR_OH] = res; |
| 172 | /* evaluate again ow as it may depend on oh */ |
| 173 | if ((ret = av_expr_parse_and_eval(&res, (expr = s->w_expr), |
| 174 | var_names, s->var_values, |
| 175 | NULL, NULL, NULL, NULL, NULL, 0, ctx)) < 0) |
| 176 | goto fail_expr; |
| 177 | |
| 178 | s->var_values[VAR_OUT_W] = s->var_values[VAR_OW] = res; |
| 179 | if (normalize_double(&s->w, s->var_values[VAR_OUT_W]) < 0 || |
| 180 | normalize_double(&s->h, s->var_values[VAR_OUT_H]) < 0) { |
| 181 | av_log(ctx, AV_LOG_ERROR, |
| 182 | "Too big value or invalid expression for out_w/ow or out_h/oh. " |
| 183 | "Maybe the expression for out_w:'%s' or for out_h:'%s' is self-referencing.\n", |
| 184 | s->w_expr, s->h_expr); |
| 185 | return AVERROR(EINVAL); |
| 186 | } |
| 187 | s->w &= ~((1 << s->hsub) - 1); |
| 188 | s->h &= ~((1 << s->vsub) - 1); |
| 189 | |
| 190 | av_expr_free(s->x_pexpr); |
| 191 | av_expr_free(s->y_pexpr); |
| 192 | s->x_pexpr = s->y_pexpr = NULL; |
| 193 | if ((ret = av_expr_parse(&s->x_pexpr, s->x_expr, var_names, |
| 194 | NULL, NULL, NULL, NULL, 0, ctx)) < 0 || |
| 195 | (ret = av_expr_parse(&s->y_pexpr, s->y_expr, var_names, |
| 196 | NULL, NULL, NULL, NULL, 0, ctx)) < 0) |
| 197 | return AVERROR(EINVAL); |
| 198 | |
| 199 | if (s->keep_aspect) { |
| 200 | AVRational dar = av_mul_q(link->sample_aspect_ratio, |
| 201 | (AVRational){ link->w, link->h }); |
| 202 | av_reduce(&s->out_sar.num, &s->out_sar.den, |
| 203 | dar.num * s->h, dar.den * s->w, INT_MAX); |
| 204 | } else |
| 205 | s->out_sar = link->sample_aspect_ratio; |
| 206 | |
| 207 | av_log(ctx, AV_LOG_VERBOSE, "w:%d h:%d sar:%d/%d -> w:%d h:%d sar:%d/%d\n", |
| 208 | link->w, link->h, link->sample_aspect_ratio.num, link->sample_aspect_ratio.den, |
| 209 | s->w, s->h, s->out_sar.num, s->out_sar.den); |
| 210 | |
| 211 | if (s->w <= 0 || s->h <= 0 || |
| 212 | s->w > link->w || s->h > link->h) { |
| 213 | av_log(ctx, AV_LOG_ERROR, |
| 214 | "Invalid too big or non positive size for width '%d' or height '%d'\n", |
| 215 | s->w, s->h); |
| 216 | return AVERROR(EINVAL); |
| 217 | } |
| 218 | |
| 219 | /* set default, required in the case the first computed value for x/y is NAN */ |
| 220 | s->x = (link->w - s->w) / 2; |
| 221 | s->y = (link->h - s->h) / 2; |
| 222 | s->x &= ~((1 << s->hsub) - 1); |
| 223 | s->y &= ~((1 << s->vsub) - 1); |
| 224 | return 0; |
| 225 | |
| 226 | fail_expr: |
| 227 | av_log(NULL, AV_LOG_ERROR, "Error when evaluating the expression '%s'\n", expr); |
| 228 | return ret; |
| 229 | } |
| 230 | |
| 231 | static int config_output(AVFilterLink *link) |
| 232 | { |
| 233 | CropContext *s = link->src->priv; |
| 234 | |
| 235 | link->w = s->w; |
| 236 | link->h = s->h; |
| 237 | link->sample_aspect_ratio = s->out_sar; |
| 238 | |
| 239 | return 0; |
| 240 | } |
| 241 | |
| 242 | static int filter_frame(AVFilterLink *link, AVFrame *frame) |
| 243 | { |
| 244 | AVFilterContext *ctx = link->dst; |
| 245 | CropContext *s = ctx->priv; |
| 246 | const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(link->format); |
| 247 | int i; |
| 248 | |
| 249 | frame->width = s->w; |
| 250 | frame->height = s->h; |
| 251 | |
| 252 | s->var_values[VAR_N] = link->frame_count; |
| 253 | s->var_values[VAR_T] = frame->pts == AV_NOPTS_VALUE ? |
| 254 | NAN : frame->pts * av_q2d(link->time_base); |
| 255 | s->var_values[VAR_POS] = av_frame_get_pkt_pos(frame) == -1 ? |
| 256 | NAN : av_frame_get_pkt_pos(frame); |
| 257 | s->var_values[VAR_X] = av_expr_eval(s->x_pexpr, s->var_values, NULL); |
| 258 | s->var_values[VAR_Y] = av_expr_eval(s->y_pexpr, s->var_values, NULL); |
| 259 | s->var_values[VAR_X] = av_expr_eval(s->x_pexpr, s->var_values, NULL); |
| 260 | |
| 261 | normalize_double(&s->x, s->var_values[VAR_X]); |
| 262 | normalize_double(&s->y, s->var_values[VAR_Y]); |
| 263 | |
| 264 | if (s->x < 0) |
| 265 | s->x = 0; |
| 266 | if (s->y < 0) |
| 267 | s->y = 0; |
| 268 | if ((unsigned)s->x + (unsigned)s->w > link->w) |
| 269 | s->x = link->w - s->w; |
| 270 | if ((unsigned)s->y + (unsigned)s->h > link->h) |
| 271 | s->y = link->h - s->h; |
| 272 | s->x &= ~((1 << s->hsub) - 1); |
| 273 | s->y &= ~((1 << s->vsub) - 1); |
| 274 | |
| 275 | av_dlog(ctx, "n:%d t:%f pos:%f x:%d y:%d x+w:%d y+h:%d\n", |
| 276 | (int)s->var_values[VAR_N], s->var_values[VAR_T], s->var_values[VAR_POS], |
| 277 | s->x, s->y, s->x+s->w, s->y+s->h); |
| 278 | |
| 279 | frame->data[0] += s->y * frame->linesize[0]; |
| 280 | frame->data[0] += s->x * s->max_step[0]; |
| 281 | |
| 282 | if (!(desc->flags & AV_PIX_FMT_FLAG_PAL || desc->flags & AV_PIX_FMT_FLAG_PSEUDOPAL)) { |
| 283 | for (i = 1; i < 3; i ++) { |
| 284 | if (frame->data[i]) { |
| 285 | frame->data[i] += (s->y >> s->vsub) * frame->linesize[i]; |
| 286 | frame->data[i] += (s->x * s->max_step[i]) >> s->hsub; |
| 287 | } |
| 288 | } |
| 289 | } |
| 290 | |
| 291 | /* alpha plane */ |
| 292 | if (frame->data[3]) { |
| 293 | frame->data[3] += s->y * frame->linesize[3]; |
| 294 | frame->data[3] += s->x * s->max_step[3]; |
| 295 | } |
| 296 | |
| 297 | return ff_filter_frame(link->dst->outputs[0], frame); |
| 298 | } |
| 299 | |
| 300 | #define OFFSET(x) offsetof(CropContext, x) |
| 301 | #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM |
| 302 | |
| 303 | static const AVOption crop_options[] = { |
| 304 | { "out_w", "set the width crop area expression", OFFSET(w_expr), AV_OPT_TYPE_STRING, {.str = "iw"}, CHAR_MIN, CHAR_MAX, FLAGS }, |
| 305 | { "w", "set the width crop area expression", OFFSET(w_expr), AV_OPT_TYPE_STRING, {.str = "iw"}, CHAR_MIN, CHAR_MAX, FLAGS }, |
| 306 | { "out_h", "set the height crop area expression", OFFSET(h_expr), AV_OPT_TYPE_STRING, {.str = "ih"}, CHAR_MIN, CHAR_MAX, FLAGS }, |
| 307 | { "h", "set the height crop area expression", OFFSET(h_expr), AV_OPT_TYPE_STRING, {.str = "ih"}, CHAR_MIN, CHAR_MAX, FLAGS }, |
| 308 | { "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 }, |
| 309 | { "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 }, |
| 310 | { "keep_aspect", "keep aspect ratio", OFFSET(keep_aspect), AV_OPT_TYPE_INT, {.i64=0}, 0, 1, FLAGS }, |
| 311 | { NULL } |
| 312 | }; |
| 313 | |
| 314 | AVFILTER_DEFINE_CLASS(crop); |
| 315 | |
| 316 | static const AVFilterPad avfilter_vf_crop_inputs[] = { |
| 317 | { |
| 318 | .name = "default", |
| 319 | .type = AVMEDIA_TYPE_VIDEO, |
| 320 | .filter_frame = filter_frame, |
| 321 | .config_props = config_input, |
| 322 | }, |
| 323 | { NULL } |
| 324 | }; |
| 325 | |
| 326 | static const AVFilterPad avfilter_vf_crop_outputs[] = { |
| 327 | { |
| 328 | .name = "default", |
| 329 | .type = AVMEDIA_TYPE_VIDEO, |
| 330 | .config_props = config_output, |
| 331 | }, |
| 332 | { NULL } |
| 333 | }; |
| 334 | |
| 335 | AVFilter ff_vf_crop = { |
| 336 | .name = "crop", |
| 337 | .description = NULL_IF_CONFIG_SMALL("Crop the input video."), |
| 338 | .priv_size = sizeof(CropContext), |
| 339 | .priv_class = &crop_class, |
| 340 | .query_formats = query_formats, |
| 341 | .uninit = uninit, |
| 342 | .inputs = avfilter_vf_crop_inputs, |
| 343 | .outputs = avfilter_vf_crop_outputs, |
| 344 | }; |