libavfilter/vf_framepack.c

   1 /*
   2  * Copyright (c) 2013 Vittorio Giovara
   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  * Generate a frame packed video, by combining two views in a single surface.
  24  */
  25
  26 #include <string.h>
  27
  28 #include "libavutil/imgutils.h"
  29 #include "libavutil/opt.h"
  30 #include "libavutil/pixdesc.h"
  31 #include "libavutil/rational.h"
  32 #include "libavutil/stereo3d.h"
  33
  34 #include "avfilter.h"
  35 #include "formats.h"
  36 #include "internal.h"
  37 #include "video.h"
  38
  39 #define LEFT  0
  40 #define RIGHT 1
  41
  42 typedef struct FramepackContext {
  43     const AVClass *class;
  44
  45     const AVPixFmtDescriptor *pix_desc; ///< agreed pixel format
  46
  47     enum AVStereo3DType format;         ///< frame pack type output
  48
  49     AVFrame *input_views[2];            ///< input frames
  50
  51     int64_t double_pts;                 ///< new pts for frameseq mode
  52 } FramepackContext;
  53
  54 static const enum AVPixelFormat formats_supported[] = {
  55     AV_PIX_FMT_YUV420P,  AV_PIX_FMT_YUV422P,  AV_PIX_FMT_YUV444P,
  56     AV_PIX_FMT_YUV410P,  AV_PIX_FMT_YUVA420P, AV_PIX_FMT_YUVJ420P,
  57     AV_PIX_FMT_YUVJ422P, AV_PIX_FMT_YUVJ444P, AV_PIX_FMT_YUVJ440P,
  58     AV_PIX_FMT_NONE
  59 };
  60
  61 static int query_formats(AVFilterContext *ctx)
  62 {
  63     // this will ensure that formats are the same on all pads
  64     ff_set_common_formats(ctx, ff_make_format_list(formats_supported));
  65     return 0;
  66 }
  67
  68 static av_cold void framepack_uninit(AVFilterContext *ctx)
  69 {
  70     FramepackContext *s = ctx->priv;
  71
  72     // clean any leftover frame
  73     av_frame_free(&s->input_views[LEFT]);
  74     av_frame_free(&s->input_views[RIGHT]);
  75 }
  76
  77 static int config_output(AVFilterLink *outlink)
  78 {
  79     AVFilterContext *ctx  = outlink->src;
  80     FramepackContext *s   = outlink->src->priv;
  81
  82     int width             = ctx->inputs[LEFT]->w;
  83     int height            = ctx->inputs[LEFT]->h;
  84     AVRational time_base  = ctx->inputs[LEFT]->time_base;
  85     AVRational frame_rate = ctx->inputs[LEFT]->frame_rate;
  86
  87     // check size and fps match on the other input
  88     if (width  != ctx->inputs[RIGHT]->w ||
  89         height != ctx->inputs[RIGHT]->h) {
  90         av_log(ctx, AV_LOG_ERROR,
  91                "Left and right sizes differ (%dx%d vs %dx%d).\n",
  92                width, height,
  93                ctx->inputs[RIGHT]->w, ctx->inputs[RIGHT]->h);
  94         return AVERROR_INVALIDDATA;
  95     } else if (av_cmp_q(time_base, ctx->inputs[RIGHT]->time_base) != 0) {
  96         av_log(ctx, AV_LOG_ERROR,
  97                "Left and right time bases differ (%d/%d vs %d/%d).\n",
  98                time_base.num, time_base.den,
  99                ctx->inputs[RIGHT]->time_base.num,
 100                ctx->inputs[RIGHT]->time_base.den);
 101         return AVERROR_INVALIDDATA;
 102     } else if (av_cmp_q(frame_rate, ctx->inputs[RIGHT]->frame_rate) != 0) {
 103         av_log(ctx, AV_LOG_ERROR,
 104                "Left and right framerates differ (%d/%d vs %d/%d).\n",
 105                frame_rate.num, frame_rate.den,
 106                ctx->inputs[RIGHT]->frame_rate.num,
 107                ctx->inputs[RIGHT]->frame_rate.den);
 108         return AVERROR_INVALIDDATA;
 109     }
 110
 111     s->pix_desc = av_pix_fmt_desc_get(outlink->format);
 112     if (!s->pix_desc)
 113         return AVERROR_BUG;
 114
 115     // modify output properties as needed
 116     switch (s->format) {
 117     case AV_STEREO3D_FRAMESEQUENCE:
 118         time_base.den  *= 2;
 119         frame_rate.num *= 2;
 120
 121         s->double_pts = AV_NOPTS_VALUE;
 122         break;
 123     case AV_STEREO3D_COLUMNS:
 124     case AV_STEREO3D_SIDEBYSIDE:
 125         width *= 2;
 126         break;
 127     case AV_STEREO3D_LINES:
 128     case AV_STEREO3D_TOPBOTTOM:
 129         height *= 2;
 130         break;
 131     default:
 132         av_log(ctx, AV_LOG_ERROR, "Unknown packing mode.");
 133         return AVERROR_INVALIDDATA;
 134     }
 135
 136     outlink->w          = width;
 137     outlink->h          = height;
 138     outlink->time_base  = time_base;
 139     outlink->frame_rate = frame_rate;
 140
 141     return 0;
 142 }
 143
 144 static void horizontal_frame_pack(AVFilterLink *outlink,
 145                                   AVFrame *out,
 146                                   int interleaved)
 147 {
 148     AVFilterContext *ctx = outlink->src;
 149     FramepackContext *s = ctx->priv;
 150     int i, plane;
 151
 152     if (interleaved) {
 153         const uint8_t *leftp  = s->input_views[LEFT]->data[0];
 154         const uint8_t *rightp = s->input_views[RIGHT]->data[0];
 155         uint8_t *dstp         = out->data[0];
 156         int length = out->width / 2;
 157         int lines  = out->height;
 158
 159         for (plane = 0; plane < s->pix_desc->nb_components; plane++) {
 160             if (plane == 1 || plane == 2) {
 161                 length = -(-(out->width / 2) >> s->pix_desc->log2_chroma_w);
 162                 lines  = -(-(out->height)    >> s->pix_desc->log2_chroma_h);
 163             }
 164             for (i = 0; i < lines; i++) {
 165                 int j;
 166                 leftp  = s->input_views[LEFT]->data[plane] +
 167                          s->input_views[LEFT]->linesize[plane] * i;
 168                 rightp = s->input_views[RIGHT]->data[plane] +
 169                          s->input_views[RIGHT]->linesize[plane] * i;
 170                 dstp   = out->data[plane] + out->linesize[plane] * i;
 171                 for (j = 0; j < length; j++) {
 172                     // interpolate chroma as necessary
 173                     if ((s->pix_desc->log2_chroma_w ||
 174                          s->pix_desc->log2_chroma_h) &&
 175                         (plane == 1 || plane == 2)) {
 176                         *dstp++ = (*leftp + *rightp) / 2;
 177                         *dstp++ = (*leftp + *rightp) / 2;
 178                     } else {
 179                         *dstp++ = *leftp;
 180                         *dstp++ = *rightp;
 181                     }
 182                     leftp += 1;
 183                     rightp += 1;
 184                 }
 185             }
 186         }
 187     } else {
 188         for (i = 0; i < 2; i++) {
 189             const uint8_t *src[4];
 190             uint8_t *dst[4];
 191             int sub_w = s->input_views[i]->width >> s->pix_desc->log2_chroma_w;
 192
 193             src[0] = s->input_views[i]->data[0];
 194             src[1] = s->input_views[i]->data[1];
 195             src[2] = s->input_views[i]->data[2];
 196
 197             dst[0] = out->data[0] + i * s->input_views[i]->width;
 198             dst[1] = out->data[1] + i * sub_w;
 199             dst[2] = out->data[2] + i * sub_w;
 200
 201             av_image_copy(dst, out->linesize, src, s->input_views[i]->linesize,
 202                           s->input_views[i]->format,
 203                           s->input_views[i]->width,
 204                           s->input_views[i]->height);
 205         }
 206     }
 207 }
 208
 209 static void vertical_frame_pack(AVFilterLink *outlink,
 210                                 AVFrame *out,
 211                                 int interleaved)
 212 {
 213     AVFilterContext *ctx = outlink->src;
 214     FramepackContext *s = ctx->priv;
 215     int i;
 216
 217     for (i = 0; i < 2; i++) {
 218         const uint8_t *src[4];
 219         uint8_t *dst[4];
 220         int linesizes[4];
 221         int sub_h = s->input_views[i]->height >> s->pix_desc->log2_chroma_h;
 222
 223         src[0] = s->input_views[i]->data[0];
 224         src[1] = s->input_views[i]->data[1];
 225         src[2] = s->input_views[i]->data[2];
 226
 227         dst[0] = out->data[0] + i * out->linesize[0] *
 228                  (interleaved + s->input_views[i]->height * (1 - interleaved));
 229         dst[1] = out->data[1] + i * out->linesize[1] *
 230                  (interleaved + sub_h * (1 - interleaved));
 231         dst[2] = out->data[2] + i * out->linesize[2] *
 232                  (interleaved + sub_h * (1 - interleaved));
 233
 234         linesizes[0] = out->linesize[0] +
 235                        interleaved * out->linesize[0];
 236         linesizes[1] = out->linesize[1] +
 237                        interleaved * out->linesize[1];
 238         linesizes[2] = out->linesize[2] +
 239                        interleaved * out->linesize[2];
 240
 241         av_image_copy(dst, linesizes, src, s->input_views[i]->linesize,
 242                       s->input_views[i]->format,
 243                       s->input_views[i]->width,
 244                       s->input_views[i]->height);
 245     }
 246 }
 247
 248 static av_always_inline void spatial_frame_pack(AVFilterLink *outlink,
 249                                                 AVFrame *dst)
 250 {
 251     AVFilterContext *ctx = outlink->src;
 252     FramepackContext *s = ctx->priv;
 253     switch (s->format) {
 254     case AV_STEREO3D_SIDEBYSIDE:
 255         horizontal_frame_pack(outlink, dst, 0);
 256         break;
 257     case AV_STEREO3D_COLUMNS:
 258         horizontal_frame_pack(outlink, dst, 1);
 259         break;
 260     case AV_STEREO3D_TOPBOTTOM:
 261         vertical_frame_pack(outlink, dst, 0);
 262         break;
 263     case AV_STEREO3D_LINES:
 264         vertical_frame_pack(outlink, dst, 1);
 265         break;
 266     }
 267 }
 268
 269 static int filter_frame_left(AVFilterLink *inlink, AVFrame *frame)
 270 {
 271     FramepackContext *s = inlink->dst->priv;
 272     s->input_views[LEFT] = frame;
 273     return 0;
 274 }
 275
 276 static int filter_frame_right(AVFilterLink *inlink, AVFrame *frame)
 277 {
 278     FramepackContext *s = inlink->dst->priv;
 279     s->input_views[RIGHT] = frame;
 280     return 0;
 281 }
 282
 283 static int request_frame(AVFilterLink *outlink)
 284 {
 285     AVFilterContext *ctx = outlink->src;
 286     FramepackContext *s = ctx->priv;
 287     AVStereo3D *stereo;
 288     int ret, i;
 289
 290     /* get a frame on the either input, stop as soon as a video ends */
 291     for (i = 0; i < 2; i++) {
 292         if (!s->input_views[i]) {
 293             ret = ff_request_frame(ctx->inputs[i]);
 294             if (ret < 0)
 295                 return ret;
 296         }
 297     }
 298
 299     if (s->format == AV_STEREO3D_FRAMESEQUENCE) {
 300         if (s->double_pts == AV_NOPTS_VALUE)
 301             s->double_pts = s->input_views[LEFT]->pts;
 302
 303         for (i = 0; i < 2; i++) {
 304             // set correct timestamps
 305             s->input_views[i]->pts = s->double_pts++;
 306
 307             // set stereo3d side data
 308             stereo = av_stereo3d_create_side_data(s->input_views[i]);
 309             if (!stereo)
 310                 return AVERROR(ENOMEM);
 311             stereo->type = s->format;
 312
 313             // filter the frame and immediately relinquish its pointer
 314             ret = ff_filter_frame(outlink, s->input_views[i]);
 315             s->input_views[i] = NULL;
 316             if (ret < 0)
 317                 return ret;
 318         }
 319         return ret;
 320     } else {
 321         AVFrame *dst = ff_get_video_buffer(outlink, outlink->w, outlink->h);
 322         if (!dst)
 323             return AVERROR(ENOMEM);
 324
 325         spatial_frame_pack(outlink, dst);
 326
 327         // get any property from the original frame
 328         ret = av_frame_copy_props(dst, s->input_views[LEFT]);
 329         if (ret < 0) {
 330             av_frame_free(&dst);
 331             return ret;
 332         }
 333
 334         for (i = 0; i < 2; i++)
 335             av_frame_free(&s->input_views[i]);
 336
 337         // set stereo3d side data
 338         stereo = av_stereo3d_create_side_data(dst);
 339         if (!stereo) {
 340             av_frame_free(&dst);
 341             return AVERROR(ENOMEM);
 342         }
 343         stereo->type = s->format;
 344
 345         return ff_filter_frame(outlink, dst);
 346     }
 347 }
 348
 349 #define OFFSET(x) offsetof(FramepackContext, x)
 350 #define V AV_OPT_FLAG_VIDEO_PARAM
 351 static const AVOption options[] = {
 352     { "format", "Frame pack output format", OFFSET(format), AV_OPT_TYPE_INT,
 353         { .i64 = AV_STEREO3D_SIDEBYSIDE }, 0, INT_MAX, .flags = V, .unit = "format" },
 354     { "sbs", "Views are packed next to each other", 0, AV_OPT_TYPE_CONST,
 355         { .i64 = AV_STEREO3D_SIDEBYSIDE }, INT_MIN, INT_MAX, .flags = V, .unit = "format" },
 356     { "tab", "Views are packed on top of each other", 0, AV_OPT_TYPE_CONST,
 357         { .i64 = AV_STEREO3D_TOPBOTTOM }, INT_MIN, INT_MAX, .flags = V, .unit = "format" },
 358     { "frameseq", "Views are one after the other", 0, AV_OPT_TYPE_CONST,
 359         { .i64 = AV_STEREO3D_FRAMESEQUENCE }, INT_MIN, INT_MAX, .flags = V, .unit = "format" },
 360     { "lines", "Views are interleaved by lines", 0, AV_OPT_TYPE_CONST,
 361         { .i64 = AV_STEREO3D_LINES }, INT_MIN, INT_MAX, .flags = V, .unit = "format" },
 362     { "columns", "Views are interleaved by columns", 0, AV_OPT_TYPE_CONST,
 363         { .i64 = AV_STEREO3D_COLUMNS }, INT_MIN, INT_MAX, .flags = V, .unit = "format" },
 364     { NULL },
 365 };
 366
 367 static const AVClass framepack_class = {
 368     .class_name = "framepack",
 369     .item_name  = av_default_item_name,
 370     .option     = options,
 371     .version    = LIBAVUTIL_VERSION_INT,
 372 };
 373
 374 static const AVFilterPad framepack_inputs[] = {
 375     {
 376         .name         = "left",
 377         .type         = AVMEDIA_TYPE_VIDEO,
 378         .filter_frame = filter_frame_left,
 379         .needs_fifo   = 1,
 380     },
 381     {
 382         .name         = "right",
 383         .type         = AVMEDIA_TYPE_VIDEO,
 384         .filter_frame = filter_frame_right,
 385         .needs_fifo   = 1,
 386     },
 387     { NULL }
 388 };
 389
 390 static const AVFilterPad framepack_outputs[] = {
 391     {
 392         .name          = "packed",
 393         .type          = AVMEDIA_TYPE_VIDEO,
 394         .config_props  = config_output,
 395         .request_frame = request_frame,
 396     },
 397     { NULL }
 398 };
 399
 400 AVFilter ff_vf_framepack = {
 401     .name          = "framepack",
 402     .description   = NULL_IF_CONFIG_SMALL("Generate a frame packed stereoscopic video."),
 403     .priv_size     = sizeof(FramepackContext),
 404     .priv_class    = &framepack_class,
 405     .query_formats = query_formats,
 406     .inputs        = framepack_inputs,
 407     .outputs       = framepack_outputs,
 408     .uninit        = framepack_uninit,
 409 };