frame extraction from webm files

Makefile

@@ -1,5 +1,6 @@
 PROG =	oxframe
-SRC  =	oxframe.c
+SRC  =	src/oxframe.c
+NETEGGS = src/nestegg.o src/halloc/src/halloc.o
 
 PREFIX	?=	/usr/local
 BINDIR	?=	${PREFIX}/bin
@@ -8,6 +9,7 @@ MAN1DIR ?=	${PREFIX}/man/man1
 CC ?=	gcc
 CFLAGS ?=	-D_FILE_OFFSET_BITS=64
 CFLAGS +=	-Wall -ffast-math -fsigned-char
+CFLAGS +=	-I. -Isrc -Isrc/halloc
 
 INSTALL =	install
 
@@ -17,19 +19,22 @@ LINKFLAGS ?=    -L${PREFIX}/lib
 LINKFLAGS +=    `imlib2-config --libs`
 LINKFLAGS +=    -L/usr/local/lib /usr/local/lib/liboggplay.a 
 LINKFLAGS +=    -loggz -lfishsound -ltheora -lvorbisenc -lvorbis -lm -logg -lkate -lpthread
+LINKFLAGS +=    -lvpx
 
 
 all: ${PROG}
 
+src/nestegg.o: src/nestegg.h src/nestegg.c src/halloc/halloc.h
+src/halloc/src/halloc.o: src/halloc/src/halloc.c src/halloc/halloc.h src/halloc/src/align.h src/halloc/src/hlist.h src/halloc/src/macros.h
 
 
-${PROG}: ${SRC}
+${PROG}: ${SRC} ${NETEGGS}
-	${CC} -Wall -Wno-parentheses -O3 -fforce-addr -fomit-frame-pointer -finline-functions -funroll-loops  ${CFLAGS} ${INCLUDEFLAGS} -o ${PROG} $< ${LINKFLAGS}
+	${CC} -Wall -Wno-parentheses -O3 -fforce-addr -fomit-frame-pointer -finline-functions -funroll-loops  ${CFLAGS} ${INCLUDEFLAGS} -o ${PROG} $< ${NETEGGS} ${LINKFLAGS}
 
 install: ${PROG}
 	${INSTALL} -c -m 555 -o root -g bin ${PROG} ${BINDIR}
 	${INSTALL} -c -m 555 -o root -g bin oxposterframe ${BINDIR}
 
 clean:
-	-@rm -f ${PROG} *~ core *.core
+	-@rm -f ${PROG} *~ core *.core src/*.o src/halloc/src/*.o
 

src/oxframe.c

@@ -17,10 +17,19 @@
  * along with This program.  If not, see <http://www.gnu.org/licenses/>.
  */
 
-#include <oggplay/oggplay.h>
+#include <assert.h>
+#include <stdarg.h>
 #include <stdio.h>
 #include <stdlib.h>
 #include <getopt.h>
+#include <oggplay/oggplay.h>
+#include "nestegg.h"
+
+#define VPX_CODEC_DISABLE_COMPAT 1
+#include "vpx/vpx_decoder.h"
+#include "vpx/vp8dx.h"
+
+#define interface (&vpx_codec_vp8_dx_algo)
 
 #include <string.h>
 
@@ -185,16 +194,282 @@ void init_state(oxstate *state) {
   state->format = oxImageNotSet;
 }
 
-int main (int argc, char * argv[]) {
 
-  int c,long_option_index;
+//WebM
-  OggPlay       * player;
+static int
-  OggPlayReader * reader = NULL;
+stdio_read(void * p, size_t length, void * fp)
+{
+  size_t r;
+
+  r = fread(p, length, 1, fp);
+  if (r == 0 && feof(fp)) {
+    return 0;
+  }
+  return r == 0 ? -1 : 1;
+}
+
+static int
+stdio_seek(int64_t offset, int whence, void * fp)
+{
+  return fseek(fp, offset, whence);
+}
+
+static int64_t
+stdio_tell(void * fp)
+{
+  return ftell(fp);
+}
+
+static void
+log_callback(nestegg * ctx, unsigned int severity, char const * fmt, ...)
+{
+  va_list ap;
+  char const * sev = NULL;
+
+#ifndef DEBUG
+  if (severity < NESTEGG_LOG_WARNING)
+    return;
+#endif
+
+  switch (severity) {
+  case NESTEGG_LOG_DEBUG:
+    sev = "debug:   ";
+    break;
+  case NESTEGG_LOG_WARNING:
+    sev = "warning: ";
+    break;
+  case NESTEGG_LOG_CRITICAL:
+    sev = "critical:";
+    break;
+  default:
+    sev = "unknown: ";
+  }
+
+  fprintf(stderr, "%p %s ", (void *) ctx, sev);
+
+  va_start(ap, fmt);
+  vfprintf(stderr, fmt, ap);
+  va_end(ap);
+
+  fprintf(stderr, "\n");
+}
+static void die_codec(vpx_codec_ctx_t *ctx, const char *s) {
+    const char *detail = vpx_codec_error_detail(ctx);
+ 
+    printf("%s: %s\n", s, vpx_codec_error(ctx));
+    if(detail)
+        printf("    %s\n",detail);
+    exit(EXIT_FAILURE);
+}
+
+int extract_frame_ogv(oxstate *state) {
   int             i;
   int             fps_num = 25;
   int             fps_denom = 1;
   int             granuleshift = 6;
   long            max_num, offset;
+
+  OggPlay       * player;
+  OggPlayReader * reader = NULL;
+
+  reader = oggplay_file_reader_new(state->input);
+  player = oggplay_open_with_reader(reader);
+
+  if (player == NULL) {
+    fprintf (stderr, "could not initialise oggplay with this file\n");
+    exit (1);
+  }
+
+  for (i = 0; i < oggplay_get_num_tracks (player); i++) {
+    if (oggplay_get_track_type (player, i) == OGGZ_CONTENT_THEORA) {
+      oggplay_set_callback_num_frames (player, i, 1);
+      oggplay_get_video_fps(player, i, &fps_denom, &fps_num);
+    }
+    oggplay_set_track_active(player, i);
+  }
+  oggplay_set_data_callback(player, dump_frame_callback, state);
+
+  max_num = 1 << granuleshift;
+  offset = (1000 * max_num * fps_denom) / fps_num;
+
+  state->duration = oggplay_get_duration(player);
+
+  /*
+  if (frame_pos > state->duration) {
+    fprintf (stderr, "can not seek to frame later than duration\n");
+    exit (1);
+  }
+  */
+
+  if(state->frame_pos - offset > 0) {
+    if (oggplay_seek(player, state->frame_pos - offset) == E_OGGPLAY_CANT_SEEK) {
+      fprintf (stderr, "failed to seeek to %ld\n", state->frame_pos);
+      exit (1);
+    }
+  }
+  oggplay_start_decoding(player);
+  
+  oggplay_close (player);
+  return 0;
+}
+
+int extract_frame_webm(oxstate *state) {
+    FILE * fp;
+    int r, type, codec_id;
+    nestegg * ctx;
+    nestegg_packet * pkt;
+    nestegg_video_params vparams;
+    uint64_t duration, pkt_tstamp;
+    unsigned int i, tracks;
+    int flags = 0;
+    int done = 0;
+    vpx_codec_ctx_t   codec;
+    vpx_image_t      *img;
+
+    //in nanoseconds
+    uint64_t  seek_tstamp = (uint64_t)state->frame_pos*1000000;
+
+    nestegg_io io = {
+      stdio_read,
+      stdio_seek,
+      stdio_tell,
+      NULL
+    };
+
+    fp = fopen(state->input, "rb");
+    if (!fp) {
+        fprintf (stderr, "could not open input file\n");
+        exit (1);
+    }
+    io.userdata = fp;
+    ctx = NULL;
+    r = nestegg_init(&ctx, io, log_callback);
+    if (r != 0)
+      return EXIT_FAILURE;
+
+    nestegg_track_count(ctx, &tracks);
+    nestegg_duration(ctx, &duration);
+
+    if (seek_tstamp > duration) {
+      fprintf (stderr, "can not seek to frame later than duration\n");
+      exit (1);
+    }
+
+    /* Initialize codec */
+    if(vpx_codec_dec_init(&codec, interface, NULL, flags))
+        die_codec(&codec, "Failed to initialize decoder");
+    
+    for (i = 0; i < tracks; ++i) {
+      type = nestegg_track_type(ctx, i);
+      codec_id = nestegg_track_codec_id(ctx, i);
+
+      if (type == NESTEGG_TRACK_VIDEO && codec_id == NESTEGG_CODEC_VP8) {
+        nestegg_track_video_params(ctx, i, &vparams);
+
+        if(!nestegg_track_seek(ctx, i, seek_tstamp)) {
+
+          while (!done && (r = nestegg_read_packet(ctx, &pkt)) > 0) {
+            unsigned int track;
+
+            nestegg_packet_track(pkt, &track);
+            if(nestegg_packet_tstamp(pkt, &pkt_tstamp) < 0) {
+              fprintf (stderr, "faild to get timestamp\n");
+            }
+
+            // only look for video track
+            if (track == i) {
+              unsigned int chunk, chunks;
+
+              nestegg_packet_count(pkt, &chunks);
+
+              // Decode each chunk of data.
+              for (chunk = 0; chunk < chunks; ++chunk) {
+                vpx_codec_iter_t  iter = NULL;
+                unsigned char * data;
+                size_t data_size;
+
+                nestegg_packet_data(pkt, chunk, &data, &data_size);
+
+                /* Decode the frame */
+                if(vpx_codec_decode(&codec, data, data_size, NULL, 0))
+                    die_codec(&codec, "Failed to decode frame");
+
+                while((img = vpx_codec_get_frame(&codec, &iter))) {
+                  //we got a frame...
+                  if(seek_tstamp-pkt_tstamp<=0) {
+                    unsigned int y;
+                    unsigned char *q, *p, *q2, *p2;
+                    OggPlayYUVChannels  from;
+                    OggPlayRGBChannels  to;
+
+                    done = 1;
+
+                    from.y_width = img->d_w;
+                    from.y_height = img->d_h;
+                    from.uv_width = (1 + img->d_w) / 2;
+                    from.uv_height = (1 + img->d_h) / 2;
+
+                    from.ptry = malloc(from.y_width * from.y_height);
+                    from.ptru = malloc(from.uv_width * from.uv_height);
+                    from.ptrv = malloc(from.uv_width * from.uv_height);
+
+                    q =img->planes[PLANE_Y];
+                    p = from.ptry;
+                    for(y=0; y<img->d_h; y++) {
+                      memcpy(p, q, img->d_w);
+                      p += img->d_w;
+                      q += img->stride[PLANE_Y];
+                    }
+                    q =img->planes[PLANE_U];
+                    p = from.ptru;
+                    q2 =img->planes[PLANE_V];
+                    p2 = from.ptrv;
+                    for(y=0; y<(1 + img->d_h) / 2; y++) {
+                      memcpy(p, q, (1 + img->d_w) / 2);
+                      memcpy(p2, q2, (1 + img->d_w) / 2);
+                      p += (1 + img->d_w) / 2;
+                      q += img->stride[PLANE_U];
+                      p2 += (1 + img->d_w) / 2;
+                      q2 += img->stride[PLANE_V];
+                    }
+
+                    to.ptro = malloc(from.y_width * from.y_height * 4);
+                    to.rgb_width = from.y_width;
+                    to.rgb_height = from.y_height;
+                   
+                    oggplay_yuv2bgra (&from, &to);
+
+                    write_image_file(&to, state);
+                    free(from.ptry);
+                    free(from.ptru);
+                    free(from.ptrv);
+                    free(to.ptro);
+                  }
+                }
+
+              }
+            }
+            nestegg_free_packet(pkt);
+          }
+
+          if(vpx_codec_destroy(&codec))
+              die_codec(&codec, "Failed to destroy codec");
+
+          nestegg_destroy(ctx);
+
+        } else {
+          fprintf (stderr, "failed to seek\n");
+          exit (1);
+        }
+
+      }
+    }
+    return 0;
+}
+
+int main (int argc, char * argv[]) {
+
+  int c,long_option_index;
   oxstate         state;
 
   const char *optstring = "h:x:y:p:i:o:f:";
@@ -232,7 +507,6 @@ int main (int argc, char * argv[]) {
         break;
       case 'i':
         state.input = optarg;
-        reader = oggplay_file_reader_new(state.input);
         break;
       case 'o':
         state.output = optarg;
@@ -256,47 +530,17 @@ int main (int argc, char * argv[]) {
       state.format = oxJPG;
   }
 
-  player = oggplay_open_with_reader(reader);
-
   if (state.input == NULL) {
     fprintf (stderr, "please provide input file\n");
     exit (1);
   }
-  if (player == NULL) {
+
-    fprintf (stderr, "could not initialise oggplay with this file\n");
+  if (strstr(&(state.input[strlen(state.input)-5]), ".webm") == NULL) { //ogv
-    exit (1);
+    extract_frame_ogv(&state);
   }
-
+  else { // .webm using nestegg + libvpx
-  for (i = 0; i < oggplay_get_num_tracks (player); i++) {
+    extract_frame_webm(&state);
-    if (oggplay_get_track_type (player, i) == OGGZ_CONTENT_THEORA) {
-      oggplay_set_callback_num_frames (player, i, 1);
-      oggplay_get_video_fps(player, i, &fps_denom, &fps_num);
-    }
-    oggplay_set_track_active(player, i);
   }
-  oggplay_set_data_callback(player, dump_frame_callback, &state);
-
-  max_num = 1 << granuleshift;
-  offset = (1000 * max_num * fps_denom) / fps_num;
-
-  state.duration = oggplay_get_duration(player);
-
-  /*
-  if (frame_pos > duration) {
-    fprintf (stderr, "can not seek to frame later than duration\n");
-    exit (1);
-  }
-  */
-
-  if(state.frame_pos - offset > 0) {
-      if (oggplay_seek(player, state.frame_pos - offset) == E_OGGPLAY_CANT_SEEK) {
-        fprintf (stderr, "failed to seeek to %ld\n", state.frame_pos);
-        exit (1);
-      }
-  }
-  oggplay_start_decoding(player);
-  
-  oggplay_close (player);
 
   return 0;
 }