libmpcodecs/vf_detc.c

   1 #include <stdio.h>
   2 #include <stdlib.h>
   3 #include <string.h>
   4
   5 #include "config.h"
   6 #include "mp_msg.h"
   7
   8 #include "img_format.h"
   9 #include "mp_image.h"
  10 #include "vf.h"
  11
  12 #include "libvo/fastmemcpy.h"
  13
  14 struct metrics {
  15         int even;
  16         int odd;
  17         int noise;
  18         int temp;
  19 };
  20
  21 struct vf_priv_s {
  22         int frame;
  23         int drop, lastdrop;
  24         struct metrics pm;
  25         int thres[5];
  26         int inframes, outframes;
  27         int mode;
  28         int (*analyze)(struct vf_priv_s *, mp_image_t *, mp_image_t *);
  29         int needread;
  30 };
  31
  32 #define COMPE(a,b,e) (abs((a)-(b)) < (((a)+(b))>>(e)))
  33 #define COMPARABLE(a,b) COMPE((a),(b),2)
  34 #define VERYCLOSE(a,b) COMPE((a),(b),3)
  35
  36 #define OUTER_TC_NBHD(s) ( \
  37  COMPARABLE((s)[-1].m.even,(s)[-1].m.odd) && \
  38  COMPARABLE((s)[1].m.even,(s)[0].m.odd) && \
  39  COMPARABLE((s)[2].m.even,(s)[1].m.odd) && \
  40  COMPARABLE((s)[-1].m.noise,(s)[0].m.temp) && \
  41  COMPARABLE((s)[2].m.noise,(s)[2].m.temp) )
  42
  43 #define INNER_TC_NBHD(s,l,h) ( \
  44  COMPARABLE((s)[0].m.even,(l)) && \
  45  COMPARABLE((s)[2].m.odd,(l)) && ( \
  46  COMPARABLE((s)[0].m.noise,(h)) || \
  47  COMPARABLE((s)[1].m.noise,(h)) ) )
  48
  49 enum {
  50         TC_DROP,
  51         TC_PROG,
  52         TC_IL1,
  53         TC_IL2
  54 };
  55
  56 static void block_diffs(struct metrics *m, unsigned char *old, unsigned char *new, int os, int ns)
  57 {
  58         int x, y, even=0, odd=0, noise, temp;
  59         unsigned char *oldp, *newp;
  60         m->noise = m->temp = 0;
  61         for (x = 8; x; x--) {
  62                 oldp = old++;
  63                 newp = new++;
  64                 noise = temp = 0;
  65                 for (y = 4; y; y--) {
  66                         even += abs(newp[0]-oldp[0]);
  67                         odd += abs(newp[ns]-oldp[os]);
  68                         noise += newp[ns]-newp[0];
  69                         temp += oldp[os]-newp[0];
  70                         oldp += os<<1;
  71                         newp += ns<<1;
  72                 }
  73                 m->noise += abs(noise);
  74                 m->temp += abs(temp);
  75         }
  76         m->even = even;
  77         m->odd = odd;
  78 }
  79
  80 static void diff_planes(struct metrics *m, unsigned char *old, unsigned char *new, int w, int h, int os, int ns)
  81 {
  82         int x, y, me=0, mo=0, mn=0, mt=0;
  83         struct metrics l;
  84         for (y = 0; y < h-7; y += 8) {
  85                 for (x = 0; x < w-7; x += 8) {
  86                         block_diffs(&l, old+x+y*os, new+x+y*ns, os, ns);
  87                         if (l.even > me) me = l.even;
  88                         if (l.odd > mo) mo = l.odd;
  89                         if (l.noise > mn) mn = l.noise;
  90                         if (l.temp > mt) mt = l.temp;
  91                 }
  92         }
  93         m->even = me;
  94         m->odd = mo;
  95         m->noise = mn;
  96         m->temp = mt;
  97 }
  98
  99 static void diff_fields(struct metrics *metr, mp_image_t *old, mp_image_t *new)
 100 {
 101         struct metrics m, mu, mv;
 102         diff_planes(&m, old->planes[0], new->planes[0],
 103                 new->w, new->h, old->stride[0], new->stride[0]);
 104         if (new->flags & MP_IMGFLAG_PLANAR) {
 105                 diff_planes(&mu, old->planes[1], new->planes[1],
 106                         new->chroma_width, new->chroma_height,
 107                         old->stride[1], new->stride[1]);
 108                 diff_planes(&mv, old->planes[2], new->planes[2],
 109                         new->chroma_width, new->chroma_height,
 110                         old->stride[2], new->stride[2]);
 111                 if (mu.even > m.even) m.even = mu.even;
 112                 if (mu.odd > m.odd) m.odd = mu.odd;
 113                 if (mu.noise > m.noise) m.noise = mu.noise;
 114                 if (mu.temp > m.temp) m.temp = mu.temp;
 115                 if (mv.even > m.even) m.even = mv.even;
 116                 if (mv.odd > m.odd) m.odd = mv.odd;
 117                 if (mv.noise > m.noise) m.noise = mv.noise;
 118                 if (mv.temp > m.temp) m.temp = mv.temp;
 119         }
 120         *metr = m;
 121 }
 122
 123 static void status(int f, struct metrics *m)
 124 {
 125         mp_msg(MSGT_VFILTER, MSGL_V, "frame %d: e=%d o=%d n=%d t=%d\n",
 126                 f, m->even, m->odd, m->noise, m->temp);
 127 }
 128
 129 static int analyze_fixed_pattern(struct vf_priv_s *p, mp_image_t *new, mp_image_t *old)
 130 {
 131         if (p->frame >= 0) p->frame = (p->frame+1)%5;
 132         mp_msg(MSGT_VFILTER, MSGL_V, "frame %d\n", p->frame);
 133         switch (p->frame) {
 134         case -1: case 0: case 1: case 2:
 135                 return TC_PROG;
 136         case 3:
 137                 return TC_IL1;
 138         case 4:
 139                 return TC_IL2;
 140         }
 141         return 0;
 142 }
 143
 144 static int analyze_aggressive(struct vf_priv_s *p, mp_image_t *new, mp_image_t *old)
 145 {
 146         struct metrics m, pm;
 147
 148         if (p->frame >= 0) p->frame = (p->frame+1)%5;
 149
 150         diff_fields(&m, old, new);
 151
 152         status(p->frame, &m);
 153
 154         pm = p->pm;
 155         p->pm = m;
 156
 157         if (p->frame == 4) {
 158                 /* We need to break at scene changes, but is this a valid test? */
 159                 if ((m.even > p->thres[2]) && (m.odd > p->thres[2]) && (m.temp > p->thres[3])
 160                         && (m.temp > 5*pm.temp) && (m.temp*2 > m.noise)) {
 161                         mp_msg(MSGT_VFILTER, MSGL_V, "scene change breaking telecine!\n");
 162                         p->frame = -1;
 163                         return TC_DROP;
 164                 }
 165                 /* Thres. is to compensate for quantization errors when noise is low */
 166                 if (m.noise - m.temp > -p->thres[4]) {
 167                         if (COMPARABLE(m.even, pm.odd)) {
 168                                 //mp_msg(MSGT_VFILTER, MSGL_V, "confirmed field match!\n");
 169                                 return TC_IL2;
 170                         } else if ((m.even < p->thres[0]) && (m.odd < p->thres[0]) && VERYCLOSE(m.even, m.odd)
 171                                 && VERYCLOSE(m.noise,m.temp) && VERYCLOSE(m.noise,pm.noise)) {
 172                                 mp_msg(MSGT_VFILTER, MSGL_V, "interlaced frame appears in duplicate!!!\n");
 173                                 p->pm = pm; /* hack :) */
 174                                 p->frame = 3;
 175                                 return TC_IL1;
 176                         }
 177                 } else {
 178                         mp_msg(MSGT_VFILTER, MSGL_V, "mismatched telecine fields!\n");
 179                         p->frame = -1;
 180                 }
 181         }
 182
 183         if (2*m.even*m.temp < m.odd*m.noise) {
 184                 mp_msg(MSGT_VFILTER, MSGL_V, "caught telecine sync!\n");
 185                 p->frame = 3;
 186                 return TC_IL1;
 187         }
 188
 189         if (p->frame < 3) {
 190                 if (m.noise > p->thres[3]) {
 191                         if (m.noise > 2*m.temp) {
 192                                 mp_msg(MSGT_VFILTER, MSGL_V, "merging fields out of sequence!\n");
 193                                 return TC_IL2;
 194                         }
 195                         if ((m.noise > 2*pm.noise) && (m.even > p->thres[2]) && (m.odd > p->thres[2])) {
 196                                 mp_msg(MSGT_VFILTER, MSGL_V, "dropping horrible interlaced frame!\n");
 197                                 return TC_DROP;
 198                         }
 199                 }
 200         }
 201
 202         switch (p->frame) {
 203         case -1:
 204                 if (4*m.noise > 5*m.temp) {
 205                         mp_msg(MSGT_VFILTER, MSGL_V, "merging fields out of sequence!\n");
 206                         return TC_IL2;
 207                 }
 208         case 0:
 209         case 1:
 210         case 2:
 211                 return TC_PROG;
 212         case 3:
 213                 if ((m.even > p->thres[1]) && (m.even > m.odd) && (m.temp > m.noise)) {
 214                         mp_msg(MSGT_VFILTER, MSGL_V, "lost telecine tracking!\n");
 215                         p->frame = -1;
 216                         return TC_PROG;
 217                 }
 218                 return TC_IL1;
 219         case 4:
 220                 return TC_IL2;
 221         }
 222         return 0;
 223 }
 224
 225 static void copy_image(mp_image_t *dmpi, mp_image_t *mpi, int field)
 226 {
 227         switch (field) {
 228         case 0:
 229                 my_memcpy_pic(dmpi->planes[0], mpi->planes[0], mpi->w, mpi->h/2,
 230                         dmpi->stride[0]*2, mpi->stride[0]*2);
 231                 if (mpi->flags & MP_IMGFLAG_PLANAR) {
 232                         my_memcpy_pic(dmpi->planes[1], mpi->planes[1],
 233                                 mpi->chroma_width, mpi->chroma_height/2,
 234                                 dmpi->stride[1]*2, mpi->stride[1]*2);
 235                         my_memcpy_pic(dmpi->planes[2], mpi->planes[2],
 236                                 mpi->chroma_width, mpi->chroma_height/2,
 237                                 dmpi->stride[2]*2, mpi->stride[2]*2);
 238                 }
 239                 break;
 240         case 1:
 241                 my_memcpy_pic(dmpi->planes[0]+dmpi->stride[0],
 242                         mpi->planes[0]+mpi->stride[0], mpi->w, mpi->h/2,
 243                         dmpi->stride[0]*2, mpi->stride[0]*2);
 244                 if (mpi->flags & MP_IMGFLAG_PLANAR) {
 245                         my_memcpy_pic(dmpi->planes[1]+dmpi->stride[1],
 246                                 mpi->planes[1]+mpi->stride[1],
 247                                 mpi->chroma_width, mpi->chroma_height/2,
 248                                 dmpi->stride[1]*2, mpi->stride[1]*2);
 249                         my_memcpy_pic(dmpi->planes[2]+dmpi->stride[2],
 250                                 mpi->planes[2]+mpi->stride[2],
 251                                 mpi->chroma_width, mpi->chroma_height/2,
 252                                 dmpi->stride[2]*2, mpi->stride[2]*2);
 253                 }
 254                 break;
 255         case 2:
 256                 memcpy_pic(dmpi->planes[0], mpi->planes[0], mpi->w, mpi->h,
 257                         dmpi->stride[0], mpi->stride[0]);
 258                 if (mpi->flags & MP_IMGFLAG_PLANAR) {
 259                         memcpy_pic(dmpi->planes[1], mpi->planes[1],
 260                                 mpi->chroma_width, mpi->chroma_height,
 261                                 dmpi->stride[1], mpi->stride[1]);
 262                         memcpy_pic(dmpi->planes[2], mpi->planes[2],
 263                                 mpi->chroma_width, mpi->chroma_height,
 264                                 dmpi->stride[2], mpi->stride[2]);
 265                 }
 266                 break;
 267         }
 268 }
 269
 270 static int do_put_image(struct vf_instance* vf, mp_image_t *dmpi)
 271 {
 272         struct vf_priv_s *p = vf->priv;
 273         int dropflag;
 274
 275         switch (p->drop) {
 276         default:
 277                 dropflag = 0;
 278                 break;
 279         case 1:
 280                 dropflag = (++p->lastdrop >= 5);
 281                 break;
 282         case 2:
 283                 dropflag = (++p->lastdrop >= 5) && (4*p->inframes <= 5*p->outframes);
 284                 break;
 285         }
 286
 287         if (dropflag) {
 288                 mp_msg(MSGT_VFILTER, MSGL_V, "drop! [%d/%d=%g]\n",
 289                         p->outframes, p->inframes, (float)p->outframes/p->inframes);
 290                 p->lastdrop = 0;
 291                 return 0;
 292         }
 293
 294         p->outframes++;
 295         return vf_next_put_image(vf, dmpi, MP_NOPTS_VALUE);
 296 }
 297
 298 static int put_image(struct vf_instance* vf, mp_image_t *mpi, double pts)
 299 {
 300         int ret=0;
 301         mp_image_t *dmpi;
 302         struct vf_priv_s *p = vf->priv;
 303
 304         p->inframes++;
 305
 306         if (p->needread) dmpi = vf_get_image(vf->next, mpi->imgfmt,
 307                 MP_IMGTYPE_STATIC, MP_IMGFLAG_ACCEPT_STRIDE |
 308                 MP_IMGFLAG_PRESERVE | MP_IMGFLAG_READABLE,
 309                 mpi->width, mpi->height);
 310         /* FIXME: is there a good way to get rid of static type? */
 311         else dmpi = vf_get_image(vf->next, mpi->imgfmt,
 312                 MP_IMGTYPE_STATIC, MP_IMGFLAG_ACCEPT_STRIDE |
 313                 MP_IMGFLAG_PRESERVE, mpi->width, mpi->height);
 314
 315         switch (p->analyze(p, mpi, dmpi)) {
 316         case TC_DROP:
 317                 /* Don't copy anything unless we'll need to read it. */
 318                 if (p->needread) copy_image(dmpi, mpi, 2);
 319                 p->lastdrop = 0;
 320                 break;
 321         case TC_PROG:
 322                 /* Copy and display the whole frame. */
 323                 copy_image(dmpi, mpi, 2);
 324                 ret = do_put_image(vf, dmpi);
 325                 break;
 326         case TC_IL1:
 327                 /* Only copy bottom field unless we need to read. */
 328                 if (p->needread) copy_image(dmpi, mpi, 2);
 329                 else copy_image(dmpi, mpi, 1);
 330                 p->lastdrop = 0;
 331                 break;
 332         case TC_IL2:
 333                 /* Copy top field and show frame, then copy bottom if needed. */
 334                 copy_image(dmpi, mpi, 0);
 335                 ret = do_put_image(vf, dmpi);
 336                 if (p->needread) copy_image(dmpi, mpi, 1);
 337                 break;
 338         }
 339         return ret;
 340 }
 341
 342 static int query_format(struct vf_instance* vf, unsigned int fmt)
 343 {
 344         /* FIXME - figure out which other formats work */
 345         switch (fmt) {
 346         case IMGFMT_YV12:
 347         case IMGFMT_IYUV:
 348         case IMGFMT_I420:
 349                 return vf_next_query_format(vf, fmt);
 350         }
 351         return 0;
 352 }
 353
 354 static int config(struct vf_instance* vf,
 355         int width, int height, int d_width, int d_height,
 356         unsigned int flags, unsigned int outfmt)
 357 {
 358         return vf_next_config(vf,width,height,d_width,d_height,flags,outfmt);
 359 }
 360
 361 static void uninit(struct vf_instance* vf)
 362 {
 363         free(vf->priv);
 364 }
 365
 366 static struct {
 367         char *name;
 368         int (*func)(struct vf_priv_s *p, mp_image_t *new, mp_image_t *old);
 369         int needread;
 370 } anal_funcs[] = {
 371         { "fixed", analyze_fixed_pattern, 0 },
 372         { "aggressive", analyze_aggressive, 1 },
 373         { NULL, NULL, 0 }
 374 };
 375
 376 #define STARTVARS if (0)
 377 #define GETVAR(str, name, out, func) \
 378  else if (!strncmp((str), name "=", sizeof(name))) \
 379  (out) = (func)((str) + sizeof(name))
 380
 381 static void parse_var(struct vf_priv_s *p, char *var)
 382 {
 383         STARTVARS;
 384         GETVAR(var, "dr", p->drop, atoi);
 385         GETVAR(var, "t0", p->thres[0], atoi);
 386         GETVAR(var, "t1", p->thres[1], atoi);
 387         GETVAR(var, "t2", p->thres[2], atoi);
 388         GETVAR(var, "t3", p->thres[3], atoi);
 389         GETVAR(var, "t4", p->thres[4], atoi);
 390         GETVAR(var, "fr", p->frame, atoi);
 391         GETVAR(var, "am", p->mode, atoi);
 392 }
 393
 394 static void parse_args(struct vf_priv_s *p, char *args)
 395 {
 396         char *next, *orig;
 397         for (args=orig=strdup(args); args; args=next) {
 398                 next = strchr(args, ':');
 399                 if (next) *next++ = 0;
 400                 parse_var(p, args);
 401         }
 402         free(orig);
 403 }
 404
 405 static int open(vf_instance_t *vf, char* args)
 406 {
 407         struct vf_priv_s *p;
 408         vf->config = config;
 409         vf->put_image = put_image;
 410         vf->query_format = query_format;
 411         vf->uninit = uninit;
 412         vf->default_reqs = VFCAP_ACCEPT_STRIDE;
 413         vf->priv = p = calloc(1, sizeof(struct vf_priv_s));
 414         p->frame = -1;
 415         p->thres[0] = 440;
 416         p->thres[1] = 720;
 417         p->thres[2] = 2500;
 418         p->thres[3] = 2500;
 419         p->thres[4] = 800;
 420         p->drop = 0;
 421         p->mode = 1;
 422         if (args) parse_args(p, args);
 423         p->analyze = anal_funcs[p->mode].func;
 424         p->needread = anal_funcs[p->mode].needread;
 425         return 1;
 426 }
 427
 428 const vf_info_t vf_info_detc = {
 429     "de-telecine filter",
 430     "detc",
 431     "Rich Felker",
 432     "",
 433     open,
 434     NULL
 435 };