libavfilter/vf_libopencv.c

   1 /*
   2  * Copyright (c) 2010 Stefano Sabatini
   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  * libopencv wrapper functions
  24  */
  25
  26 /* #define DEBUG */
  27
  28 #include <opencv/cv.h>
  29 #include <opencv/cxcore.h>
  30 #include "libavutil/avstring.h"
  31 #include "libavutil/common.h"
  32 #include "libavutil/file.h"
  33 #include "avfilter.h"
  34 #include "formats.h"
  35 #include "internal.h"
  36 #include "video.h"
  37
  38 static void fill_iplimage_from_picref(IplImage *img, const AVFilterBufferRef *picref, enum AVPixelFormat pixfmt)
  39 {
  40     IplImage *tmpimg;
  41     int depth, channels_nb;
  42
  43     if      (pixfmt == AV_PIX_FMT_GRAY8) { depth = IPL_DEPTH_8U;  channels_nb = 1; }
  44     else if (pixfmt == AV_PIX_FMT_BGRA)  { depth = IPL_DEPTH_8U;  channels_nb = 4; }
  45     else if (pixfmt == AV_PIX_FMT_BGR24) { depth = IPL_DEPTH_8U;  channels_nb = 3; }
  46     else return;
  47
  48     tmpimg = cvCreateImageHeader((CvSize){picref->video->w, picref->video->h}, depth, channels_nb);
  49     *img = *tmpimg;
  50     img->imageData = img->imageDataOrigin = picref->data[0];
  51     img->dataOrder = IPL_DATA_ORDER_PIXEL;
  52     img->origin    = IPL_ORIGIN_TL;
  53     img->widthStep = picref->linesize[0];
  54 }
  55
  56 static void fill_picref_from_iplimage(AVFilterBufferRef *picref, const IplImage *img, enum AVPixelFormat pixfmt)
  57 {
  58     picref->linesize[0] = img->widthStep;
  59     picref->data[0]     = img->imageData;
  60 }
  61
  62 static int query_formats(AVFilterContext *ctx)
  63 {
  64     static const enum AVPixelFormat pix_fmts[] = {
  65         AV_PIX_FMT_BGR24, AV_PIX_FMT_BGRA, AV_PIX_FMT_GRAY8, AV_PIX_FMT_NONE
  66     };
  67
  68     ff_set_common_formats(ctx, ff_make_format_list(pix_fmts));
  69     return 0;
  70 }
  71
  72 typedef struct {
  73     const char *name;
  74     int (*init)(AVFilterContext *ctx, const char *args);
  75     void (*uninit)(AVFilterContext *ctx);
  76     void (*end_frame_filter)(AVFilterContext *ctx, IplImage *inimg, IplImage *outimg);
  77     void *priv;
  78 } OCVContext;
  79
  80 typedef struct {
  81     int type;
  82     int    param1, param2;
  83     double param3, param4;
  84 } SmoothContext;
  85
  86 static av_cold int smooth_init(AVFilterContext *ctx, const char *args)
  87 {
  88     OCVContext *ocv = ctx->priv;
  89     SmoothContext *smooth = ocv->priv;
  90     char type_str[128] = "gaussian";
  91
  92     smooth->param1 = 3;
  93     smooth->param2 = 0;
  94     smooth->param3 = 0.0;
  95     smooth->param4 = 0.0;
  96
  97     if (args)
  98         sscanf(args, "%127[^:]:%d:%d:%lf:%lf", type_str, &smooth->param1, &smooth->param2, &smooth->param3, &smooth->param4);
  99
 100     if      (!strcmp(type_str, "blur"         )) smooth->type = CV_BLUR;
 101     else if (!strcmp(type_str, "blur_no_scale")) smooth->type = CV_BLUR_NO_SCALE;
 102     else if (!strcmp(type_str, "median"       )) smooth->type = CV_MEDIAN;
 103     else if (!strcmp(type_str, "gaussian"     )) smooth->type = CV_GAUSSIAN;
 104     else if (!strcmp(type_str, "bilateral"    )) smooth->type = CV_BILATERAL;
 105     else {
 106         av_log(ctx, AV_LOG_ERROR, "Smoothing type '%s' unknown.\n", type_str);
 107         return AVERROR(EINVAL);
 108     }
 109
 110     if (smooth->param1 < 0 || !(smooth->param1%2)) {
 111         av_log(ctx, AV_LOG_ERROR,
 112                "Invalid value '%d' for param1, it has to be a positive odd number\n",
 113                smooth->param1);
 114         return AVERROR(EINVAL);
 115     }
 116     if ((smooth->type == CV_BLUR || smooth->type == CV_BLUR_NO_SCALE || smooth->type == CV_GAUSSIAN) &&
 117         (smooth->param2 < 0 || (smooth->param2 && !(smooth->param2%2)))) {
 118         av_log(ctx, AV_LOG_ERROR,
 119                "Invalid value '%d' for param2, it has to be zero or a positive odd number\n",
 120                smooth->param2);
 121         return AVERROR(EINVAL);
 122     }
 123
 124     av_log(ctx, AV_LOG_VERBOSE, "type:%s param1:%d param2:%d param3:%f param4:%f\n",
 125            type_str, smooth->param1, smooth->param2, smooth->param3, smooth->param4);
 126     return 0;
 127 }
 128
 129 static void smooth_end_frame_filter(AVFilterContext *ctx, IplImage *inimg, IplImage *outimg)
 130 {
 131     OCVContext *ocv = ctx->priv;
 132     SmoothContext *smooth = ocv->priv;
 133     cvSmooth(inimg, outimg, smooth->type, smooth->param1, smooth->param2, smooth->param3, smooth->param4);
 134 }
 135
 136 static int read_shape_from_file(int *cols, int *rows, int **values, const char *filename,
 137                                 void *log_ctx)
 138 {
 139     uint8_t *buf, *p, *pend;
 140     size_t size;
 141     int ret, i, j, w;
 142
 143     if ((ret = av_file_map(filename, &buf, &size, 0, log_ctx)) < 0)
 144         return ret;
 145
 146     /* prescan file to get the number of lines and the maximum width */
 147     w = 0;
 148     for (i = 0; i < size; i++) {
 149         if (buf[i] == '\n') {
 150             if (*rows == INT_MAX) {
 151                 av_log(log_ctx, AV_LOG_ERROR, "Overflow on the number of rows in the file\n");
 152                 return AVERROR_INVALIDDATA;
 153             }
 154             ++(*rows);
 155             *cols = FFMAX(*cols, w);
 156             w = 0;
 157         } else if (w == INT_MAX) {
 158             av_log(log_ctx, AV_LOG_ERROR, "Overflow on the number of columns in the file\n");
 159             return AVERROR_INVALIDDATA;
 160         }
 161         w++;
 162     }
 163     if (*rows > (SIZE_MAX / sizeof(int) / *cols)) {
 164         av_log(log_ctx, AV_LOG_ERROR, "File with size %dx%d is too big\n",
 165                *rows, *cols);
 166         return AVERROR_INVALIDDATA;
 167     }
 168     if (!(*values = av_mallocz(sizeof(int) * *rows * *cols)))
 169         return AVERROR(ENOMEM);
 170
 171     /* fill *values */
 172     p    = buf;
 173     pend = buf + size-1;
 174     for (i = 0; i < *rows; i++) {
 175         for (j = 0;; j++) {
 176             if (p > pend || *p == '\n') {
 177                 p++;
 178                 break;
 179             } else
 180                 (*values)[*cols*i + j] = !!isgraph(*(p++));
 181         }
 182     }
 183     av_file_unmap(buf, size);
 184
 185 #ifdef DEBUG
 186     {
 187         char *line;
 188         if (!(line = av_malloc(*cols + 1)))
 189             return AVERROR(ENOMEM);
 190         for (i = 0; i < *rows; i++) {
 191             for (j = 0; j < *cols; j++)
 192                 line[j] = (*values)[i * *cols + j] ? '@' : ' ';
 193             line[j] = 0;
 194             av_log(log_ctx, AV_LOG_DEBUG, "%3d: %s\n", i, line);
 195         }
 196         av_free(line);
 197     }
 198 #endif
 199
 200     return 0;
 201 }
 202
 203 static int parse_iplconvkernel(IplConvKernel **kernel, char *buf, void *log_ctx)
 204 {
 205     char shape_filename[128] = "", shape_str[32] = "rect";
 206     int cols = 0, rows = 0, anchor_x = 0, anchor_y = 0, shape = CV_SHAPE_RECT;
 207     int *values = NULL, ret;
 208
 209     sscanf(buf, "%dx%d+%dx%d/%32[^=]=%127s", &cols, &rows, &anchor_x, &anchor_y, shape_str, shape_filename);
 210
 211     if      (!strcmp(shape_str, "rect"   )) shape = CV_SHAPE_RECT;
 212     else if (!strcmp(shape_str, "cross"  )) shape = CV_SHAPE_CROSS;
 213     else if (!strcmp(shape_str, "ellipse")) shape = CV_SHAPE_ELLIPSE;
 214     else if (!strcmp(shape_str, "custom" )) {
 215         shape = CV_SHAPE_CUSTOM;
 216         if ((ret = read_shape_from_file(&cols, &rows, &values, shape_filename, log_ctx)) < 0)
 217             return ret;
 218     } else {
 219         av_log(log_ctx, AV_LOG_ERROR,
 220                "Shape unspecified or type '%s' unknown.\n", shape_str);
 221         return AVERROR(EINVAL);
 222     }
 223
 224     if (rows <= 0 || cols <= 0) {
 225         av_log(log_ctx, AV_LOG_ERROR,
 226                "Invalid non-positive values for shape size %dx%d\n", cols, rows);
 227         return AVERROR(EINVAL);
 228     }
 229
 230     if (anchor_x < 0 || anchor_y < 0 || anchor_x >= cols || anchor_y >= rows) {
 231         av_log(log_ctx, AV_LOG_ERROR,
 232                "Shape anchor %dx%d is not inside the rectangle with size %dx%d.\n",
 233                anchor_x, anchor_y, cols, rows);
 234         return AVERROR(EINVAL);
 235     }
 236
 237     *kernel = cvCreateStructuringElementEx(cols, rows, anchor_x, anchor_y, shape, values);
 238     av_freep(&values);
 239     if (!*kernel)
 240         return AVERROR(ENOMEM);
 241
 242     av_log(log_ctx, AV_LOG_VERBOSE, "Structuring element: w:%d h:%d x:%d y:%d shape:%s\n",
 243            rows, cols, anchor_x, anchor_y, shape_str);
 244     return 0;
 245 }
 246
 247 typedef struct {
 248     int nb_iterations;
 249     IplConvKernel *kernel;
 250 } DilateContext;
 251
 252 static av_cold int dilate_init(AVFilterContext *ctx, const char *args)
 253 {
 254     OCVContext *ocv = ctx->priv;
 255     DilateContext *dilate = ocv->priv;
 256     char default_kernel_str[] = "3x3+0x0/rect";
 257     char *kernel_str;
 258     const char *buf = args;
 259     int ret;
 260
 261     dilate->nb_iterations = 1;
 262
 263     if (args)
 264         kernel_str = av_get_token(&buf, ":");
 265     if ((ret = parse_iplconvkernel(&dilate->kernel,
 266                                    *kernel_str ? kernel_str : default_kernel_str,
 267                                    ctx)) < 0)
 268         return ret;
 269     av_free(kernel_str);
 270
 271     sscanf(buf, ":%d", &dilate->nb_iterations);
 272     av_log(ctx, AV_LOG_VERBOSE, "iterations_nb:%d\n", dilate->nb_iterations);
 273     if (dilate->nb_iterations <= 0) {
 274         av_log(ctx, AV_LOG_ERROR, "Invalid non-positive value '%d' for nb_iterations\n",
 275                dilate->nb_iterations);
 276         return AVERROR(EINVAL);
 277     }
 278     return 0;
 279 }
 280
 281 static av_cold void dilate_uninit(AVFilterContext *ctx)
 282 {
 283     OCVContext *ocv = ctx->priv;
 284     DilateContext *dilate = ocv->priv;
 285
 286     cvReleaseStructuringElement(&dilate->kernel);
 287 }
 288
 289 static void dilate_end_frame_filter(AVFilterContext *ctx, IplImage *inimg, IplImage *outimg)
 290 {
 291     OCVContext *ocv = ctx->priv;
 292     DilateContext *dilate = ocv->priv;
 293     cvDilate(inimg, outimg, dilate->kernel, dilate->nb_iterations);
 294 }
 295
 296 static void erode_end_frame_filter(AVFilterContext *ctx, IplImage *inimg, IplImage *outimg)
 297 {
 298     OCVContext *ocv = ctx->priv;
 299     DilateContext *dilate = ocv->priv;
 300     cvErode(inimg, outimg, dilate->kernel, dilate->nb_iterations);
 301 }
 302
 303 typedef struct {
 304     const char *name;
 305     size_t priv_size;
 306     int  (*init)(AVFilterContext *ctx, const char *args);
 307     void (*uninit)(AVFilterContext *ctx);
 308     void (*end_frame_filter)(AVFilterContext *ctx, IplImage *inimg, IplImage *outimg);
 309 } OCVFilterEntry;
 310
 311 static OCVFilterEntry ocv_filter_entries[] = {
 312     { "dilate", sizeof(DilateContext), dilate_init, dilate_uninit, dilate_end_frame_filter },
 313     { "erode",  sizeof(DilateContext), dilate_init, dilate_uninit, erode_end_frame_filter  },
 314     { "smooth", sizeof(SmoothContext), smooth_init, NULL, smooth_end_frame_filter },
 315 };
 316
 317 static av_cold int init(AVFilterContext *ctx, const char *args)
 318 {
 319     OCVContext *ocv = ctx->priv;
 320     char name[128], priv_args[1024];
 321     int i;
 322     char c;
 323
 324     sscanf(args, "%127[^=:]%c%1023s", name, &c, priv_args);
 325
 326     for (i = 0; i < FF_ARRAY_ELEMS(ocv_filter_entries); i++) {
 327         OCVFilterEntry *entry = &ocv_filter_entries[i];
 328         if (!strcmp(name, entry->name)) {
 329             ocv->name             = entry->name;
 330             ocv->init             = entry->init;
 331             ocv->uninit           = entry->uninit;
 332             ocv->end_frame_filter = entry->end_frame_filter;
 333
 334             if (!(ocv->priv = av_mallocz(entry->priv_size)))
 335                 return AVERROR(ENOMEM);
 336             return ocv->init(ctx, priv_args);
 337         }
 338     }
 339
 340     av_log(ctx, AV_LOG_ERROR, "No libopencv filter named '%s'\n", name);
 341     return AVERROR(EINVAL);
 342 }
 343
 344 static av_cold void uninit(AVFilterContext *ctx)
 345 {
 346     OCVContext *ocv = ctx->priv;
 347
 348     if (ocv->uninit)
 349         ocv->uninit(ctx);
 350     av_free(ocv->priv);
 351     memset(ocv, 0, sizeof(*ocv));
 352 }
 353
 354 static int filter_frame(AVFilterLink *inlink, AVFilterBufferRef *in)
 355 {
 356     AVFilterContext *ctx = inlink->dst;
 357     OCVContext *ocv = ctx->priv;
 358     AVFilterLink *outlink= inlink->dst->outputs[0];
 359     AVFilterBufferRef *out;
 360     IplImage inimg, outimg;
 361
 362     out = ff_get_video_buffer(outlink, AV_PERM_WRITE, outlink->w, outlink->h);
 363     if (!out) {
 364         avfilter_unref_bufferp(&in);
 365         return AVERROR(ENOMEM);
 366     }
 367     avfilter_copy_buffer_ref_props(out, in);
 368
 369     fill_iplimage_from_picref(&inimg , in , inlink->format);
 370     fill_iplimage_from_picref(&outimg, out, inlink->format);
 371     ocv->end_frame_filter(ctx, &inimg, &outimg);
 372     fill_picref_from_iplimage(out, &outimg, inlink->format);
 373
 374     avfilter_unref_bufferp(&in);
 375
 376     return ff_filter_frame(outlink, out);
 377 }
 378
 379 static const AVFilterPad avfilter_vf_ocv_inputs[] = {
 380     {
 381         .name       = "default",
 382         .type       = AVMEDIA_TYPE_VIDEO,
 383         .filter_frame = filter_frame,
 384         .min_perms  = AV_PERM_READ
 385     },
 386     { NULL }
 387 };
 388
 389 static const AVFilterPad avfilter_vf_ocv_outputs[] = {
 390     {
 391         .name = "default",
 392         .type = AVMEDIA_TYPE_VIDEO,
 393     },
 394     { NULL }
 395 };
 396
 397 AVFilter avfilter_vf_ocv = {
 398     .name        = "ocv",
 399     .description = NULL_IF_CONFIG_SMALL("Apply transform using libopencv."),
 400
 401     .priv_size = sizeof(OCVContext),
 402
 403     .query_formats = query_formats,
 404     .init = init,
 405     .uninit = uninit,
 406
 407     .inputs    = avfilter_vf_ocv_inputs,
 408
 409     .outputs   = avfilter_vf_ocv_outputs,
 410 };