OSDN Git Service

libhb: don't truncate last chapter of a stream
[handbrake-jp/handbrake-jp-git.git] / libhb / scan.c
index 6c97920..74cc52a 100644 (file)
@@ -1,13 +1,15 @@
 /* $Id: scan.c,v 1.52 2005/11/25 15:05:25 titer Exp $
 
    This file is part of the HandBrake source code.
-   Homepage: <http://handbrake.m0k.org/>.
+   Homepage: <http://handbrake.fr/>.
    It may be used under the terms of the GNU General Public License. */
 
 #include "hb.h"
 #include "a52dec/a52.h"
 #include "dca.h"
 
+#define HB_MAX_PREVIEWS 30 // 30 previews = every 5 minutes of a 2.5 hour video
+
 typedef struct
 {
     hb_handle_t * h;
@@ -18,16 +20,32 @@ typedef struct
 
     hb_dvd_t    * dvd;
        hb_stream_t * stream;
+       
+    int           preview_count;
+    int           store_previews;
 
 } hb_scan_t;
 
 static void ScanFunc( void * );
 static int  DecodePreviews( hb_scan_t *, hb_title_t * title );
-static void LookForAC3AndDCA( hb_title_t * title, hb_buffer_t * b );
-static int  AllAC3AndDCAOK( hb_title_t * title );
+static void LookForAudio( hb_title_t * title, hb_buffer_t * b );
+static int  AllAudioOK( hb_title_t * title );
+
+static const char *aspect_to_string( double aspect )
+{
+    switch ( (int)(aspect * 9.) )
+    {
+        case 9 * 4 / 3:    return "4:3";
+        case 9 * 16 / 9:   return "16:9";
+    }
+    static char arstr[32];
+    sprintf( arstr, aspect >= 1.? "%.2f:1" : "1:%.2f", aspect );
+    return arstr;
+}
 
 hb_thread_t * hb_scan_init( hb_handle_t * handle, const char * path,
-                            int title_index, hb_list_t * list_title )
+                            int title_index, hb_list_t * list_title,
+                            int preview_count, int store_previews )
 {
     hb_scan_t * data = calloc( sizeof( hb_scan_t ), 1 );
 
@@ -35,7 +53,10 @@ hb_thread_t * hb_scan_init( hb_handle_t * handle, const char * path,
     data->path         = strdup( path );
     data->title_index  = title_index;
     data->list_title   = list_title;
-
+    
+    data->preview_count  = preview_count;
+    data->store_previews = store_previews;
+    
     return hb_thread_init( "scan", ScanFunc, data, HB_NORMAL_PRIORITY );
 }
 
@@ -85,36 +106,9 @@ static void ScanFunc( void * _data )
         int j;
         hb_state_t state;
         hb_audio_t * audio;
-        hb_title_t * title_tmp = NULL;
 
         title = hb_list_item( data->list_title, i );
 
-        /* I've seen a DVD with strictly identical titles. Check this
-           here and ignore it if redundant */
-        for( j = 0; j < i; j++ )
-        {
-            title_tmp = hb_list_item( data->list_title, j );
-            if( title->vts         == title_tmp->vts &&
-                title->block_start == title_tmp->block_start &&
-                title->block_end   == title_tmp->block_end &&
-                title->block_count == title_tmp->block_count )
-            {
-                break;
-            }
-            else
-            {
-                title_tmp = NULL;
-            }
-        }
-        if( title_tmp )
-        {
-            hb_log( "scan: title %d is duplicate with title %d",
-                    title->index, title_tmp->index );
-            hb_list_rem( data->list_title, title );
-            free( title );      /* This _will_ leak! */
-            continue;
-        }
-
 #define p state.param.scanning
         /* Update the UI */
         state.state   = HB_STATE_SCANNING;
@@ -132,41 +126,14 @@ static void ScanFunc( void * _data )
             continue;
         }
 
-               if (data->stream)
-               {
-                       // Stream based processing uses PID's to handle the different audio options for a given title
-                       for( j = 0; j < hb_list_count( title->list_audio ); j++ )
-                       {
-                               audio = hb_list_item( title->list_audio, j );
-                               hb_stream_update_audio(data->stream, audio);
-                       }
-               }
-               else if (data->dvd)
-               {
-                       /* Make sure we found AC3 rates and bitrates */
-                       for( j = 0; j < hb_list_count( title->list_audio ); )
-                       {
-                               audio = hb_list_item( title->list_audio, j );
-                               if( audio->config.in.codec == HB_ACODEC_AC3 &&
-                    !audio->config.in.bitrate )
-                               {
-                                       hb_list_rem( title->list_audio, audio );
-                                       free( audio );
-                                       continue;
-                               }
-                               j++;
-                       }
-               }
-
-        /* Make sure we found AC3 / DCA rates and bitrates */
+        /* Make sure we found audio rates and bitrates */
         for( j = 0; j < hb_list_count( title->list_audio ); )
         {
             audio = hb_list_item( title->list_audio, j );
-            if( ( audio->config.in.codec == HB_ACODEC_AC3 || audio->config.in.codec == HB_ACODEC_DCA ) &&
-                !audio->config.in.bitrate )
+            if( !audio->config.in.bitrate )
             {
-                hb_log( "scan: removing audio with codec of 0x%x because of no bitrate",
-                        audio->config.in.codec );
+                hb_log( "scan: removing audio 0x%x because no bitrate found",
+                        audio->id );
                 hb_list_rem( title->list_audio, audio );
                 free( audio );
                 continue;
@@ -174,26 +141,6 @@ static void ScanFunc( void * _data )
             j++;
         }
 
-        /* Do we still have audio */
-        if( !hb_list_count( title->list_audio ) )
-        {
-            hb_list_rem( data->list_title, title );
-            free( title );
-            continue;
-        }
-
-        /* set a default input channel layout of stereo for LPCM or MPEG2 audio */
-        /* AC3 and DCA will already have had their layout set via DecodePreviews above, */
-        /* which calls LookForAC3AndDCA */
-        for( j = 0; j < hb_list_count( title->list_audio ); j++ )
-        {
-            audio = hb_list_item( title->list_audio, j );
-            if( audio->config.in.codec == HB_ACODEC_LPCM || audio->config.in.codec == HB_ACODEC_MPGA )
-            {
-                audio->config.in.channel_layout = HB_INPUT_CH_LAYOUT_STEREO;
-            }
-        }
-
         i++;
     }
 
@@ -215,19 +162,21 @@ static void ScanFunc( void * _data )
         /* Autocrop by default. Gnark gnark */
         memcpy( job->crop, title->crop, 4 * sizeof( int ) );
 
-        if( title->aspect == 16 )
+        /* Preserve a source's pixel aspect, if it's available. */
+        if( title->pixel_aspect_width && title->pixel_aspect_height )
         {
-            hb_reduce( &job->pixel_aspect_width, &job->pixel_aspect_height,
-                       16 * title->height, 9 * title->width );
+            job->anamorphic.par_width  = title->pixel_aspect_width;
+            job->anamorphic.par_height = title->pixel_aspect_height;
         }
-        else
+
+        if( title->aspect != 0 && title->aspect != 1. &&
+            !job->anamorphic.par_width && !job->anamorphic.par_height)
         {
-            hb_reduce( &job->pixel_aspect_width, &job->pixel_aspect_height,
-                       4 * title->height, 3 * title->width );
+            hb_reduce( &job->anamorphic.par_width, &job->anamorphic.par_height,
+                       (int)(title->aspect * title->height + 0.5), title->width );
         }
 
         job->width = title->width - job->crop[2] - job->crop[3];
-//        job->height = title->height - job->crop[0] - job->crop[1];
         hb_fix_aspect( job, HB_KEEP_WIDTH );
         if( job->height > title->height - job->crop[0] - job->crop[1] )
         {
@@ -235,8 +184,8 @@ static void ScanFunc( void * _data )
             hb_fix_aspect( job, HB_KEEP_HEIGHT );
         }
 
-    hb_log( "scan: title (%d) job->width:%d, job->height:%d",
-            i,job->width, job->height );
+        hb_log( "scan: title (%d) job->width:%d, job->height:%d",
+                i, job->width, job->height );
 
         job->keep_ratio = 1;
 
@@ -248,8 +197,7 @@ static void ScanFunc( void * _data )
         job->vrate_base = title->rate_base;
 
         job->list_audio = hb_list_init();
-
-        job->subtitle = -1;
+        job->list_subtitle = hb_list_init();
 
         job->mux = HB_MUX_MP4;
     }
@@ -267,6 +215,146 @@ static void ScanFunc( void * _data )
     _data = NULL;
 }
 
+// -----------------------------------------------
+// stuff related to cropping
+
+#define DARK 32
+
+static inline int absdiff( int x, int y )
+{
+    return x < y ? y - x : x - y;
+}
+
+static inline int clampBlack( int x ) 
+{
+    // luma 'black' is 16 and anything less should be clamped at 16
+    return x < 16 ? 16 : x;
+}
+
+static int row_all_dark( hb_title_t *title, uint8_t* luma, int row )
+{
+    luma += title->width * row;
+
+    // compute the average luma value of the row
+    int i, avg = 0;
+    for ( i = 0; i < title->width; ++i )
+    {
+        avg += clampBlack( luma[i] );
+    }
+    avg /= title->width;
+    if ( avg >= DARK )
+        return 0;
+
+    // since we're trying to detect smooth borders, only take the row if
+    // all pixels are within +-16 of the average (this range is fairly coarse
+    // but there's a lot of quantization noise for luma values near black
+    // so anything less will fail to crop because of the noise).
+    for ( i = 0; i < title->width; ++i )
+    {
+        if ( absdiff( avg, clampBlack( luma[i] ) ) > 16 )
+            return 0;
+    }
+    return 1;
+}
+
+static int column_all_dark( hb_title_t *title, uint8_t* luma, int top, int bottom,
+                            int col )
+{
+    int stride = title->width;
+    int height = title->height - top - bottom;
+    luma += stride * top + col;
+
+    // compute the average value of the column
+    int i = height, avg = 0, row = 0;
+    for ( ; --i >= 0; row += stride )
+    {
+        avg += clampBlack( luma[row] );
+    }
+    avg /= height;
+    if ( avg >= DARK )
+        return 0;
+
+    // since we're trying to detect smooth borders, only take the column if
+    // all pixels are within +-16 of the average.
+    i = height, row = 0;
+    for ( ; --i >= 0; row += stride )
+    {
+        if ( absdiff( avg, clampBlack( luma[row] ) ) > 16 )
+            return 0;
+    }
+    return 1;
+}
+#undef DARK
+
+typedef struct {
+    int n;
+    int t[HB_MAX_PREVIEWS];
+    int b[HB_MAX_PREVIEWS];
+    int l[HB_MAX_PREVIEWS];
+    int r[HB_MAX_PREVIEWS];
+} crop_record_t;
+
+static void record_crop( crop_record_t *crops, int t, int b, int l, int r )
+{
+    crops->t[crops->n] = t;
+    crops->b[crops->n] = b;
+    crops->l[crops->n] = l;
+    crops->r[crops->n] = r;
+    ++crops->n;
+}
+
+static int compare_int( const void *a, const void *b )
+{
+    return *(const int *)a - *(const int *)b;
+}
+
+static void sort_crops( crop_record_t *crops )
+{
+    qsort( crops->t, crops->n, sizeof(crops->t[0]), compare_int );
+    qsort( crops->b, crops->n, sizeof(crops->t[0]), compare_int );
+    qsort( crops->l, crops->n, sizeof(crops->t[0]), compare_int );
+    qsort( crops->r, crops->n, sizeof(crops->t[0]), compare_int );
+}
+
+// -----------------------------------------------
+// stuff related to title width/height/aspect info
+
+typedef struct {
+    int count;              /* number of times we've seen this info entry */
+    hb_work_info_t info;    /* copy of info entry */
+} info_list_t;
+
+static void remember_info( info_list_t *info_list, hb_work_info_t *info )
+{
+    for ( ; info_list->count; ++info_list )
+    {
+        if ( memcmp( &info_list->info, info, sizeof(*info) ) == 0 )
+        {
+            // we found a match - bump its count
+            ++info_list->count;
+            return;
+        }
+    }
+    // no match found - add new entry to list (info_list points to
+    // the first free slot). NB - we assume that info_list was allocated
+    // so that it's big enough even if there are no dups. I.e., 10 slots
+    // allocated if there are 10 previews.
+    info_list->count = 1;
+    info_list->info = *info;
+}
+
+static void most_common_info( info_list_t *info_list, hb_work_info_t *info )
+{
+    int i, biggest = 0;
+    for ( i = 1; info_list[i].count; ++i )
+    {
+        if ( info_list[i].count > info_list[biggest].count )
+            biggest = i;
+    }
+    *info = info_list[biggest].info;
+    free( info_list );
+}
+
 /***********************************************************************
  * DecodePreviews
  ***********************************************************************
@@ -277,49 +365,72 @@ static void ScanFunc( void * _data )
 static int DecodePreviews( hb_scan_t * data, hb_title_t * title )
 {
     int             i, npreviews = 0;
-    hb_buffer_t   * buf_ps, * buf_es, * buf_raw;
-    hb_list_t     * list_es, * list_raw;
-    hb_libmpeg2_t * mpeg2;
+    hb_buffer_t   * buf_ps, * buf_es;
+    hb_list_t     * list_es;
     int progressive_count = 0;
     int interlaced_preview_count = 0;
-
-    int ar16_count = 0, ar4_count = 0;
+    info_list_t * info_list = calloc( data->preview_count+1, sizeof(*info_list) );
+    crop_record_t *crops = calloc( 1, sizeof(*crops) );
 
     buf_ps   = hb_buffer_init( HB_DVD_READ_BUFFER_SIZE );
     list_es  = hb_list_init();
-    list_raw = hb_list_init();
 
     hb_log( "scan: decoding previews for title %d", title->index );
 
     if (data->dvd)
-      hb_dvd_start( data->dvd, title->index, 1 );
+    {
+      hb_dvd_start( data->dvd, title, 1 );
+      title->angle_count = hb_dvd_angle_count( data->dvd );
+      hb_log( "scan: title angle(s) %d", title->angle_count );
+    }
 
-    for( i = 0; i < 10; i++ )
+    for( i = 0; i < data->preview_count; i++ )
     {
-        int j, k;
+        int j;
         FILE * file_preview;
         char   filename[1024];
 
-        //hb_log("Seeking to: %f", (float) ( i + 1 ) / 11.0 );
-
         if (data->dvd)
         {
-          if( !hb_dvd_seek( data->dvd, (float) ( i + 1 ) / 11.0 ) )
+          if( !hb_dvd_seek( data->dvd, (float) ( i + 1 ) / ( data->preview_count + 1.0 ) ) )
           {
-              goto error;
+              continue;
           }
         }
         else if (data->stream)
         {
-          if (!hb_stream_seek(data->stream, (float) ( i + 1 ) / 11.0 ) )
+          /* we start reading streams at zero rather than 1/11 because
+           * short streams may have only one sequence header in the entire
+           * file and we need it to decode any previews. */
+          if (!hb_stream_seek(data->stream, (float) i / ( data->preview_count + 1.0 ) ) )
           {
-            goto error;
+              continue;
           }
         }
 
-        hb_log( "scan: preview %d", i + 1 );
+        hb_deep_log( 2, "scan: preview %d", i + 1 );
 
-        mpeg2 = hb_libmpeg2_init();
+        int vcodec = title->video_codec? title->video_codec : WORK_DECMPEG2;
+        hb_work_object_t *vid_decoder = hb_get_work( vcodec );
+        vid_decoder->codec_param = title->video_codec_param;
+        vid_decoder->title = title;
+        vid_decoder->init( vid_decoder, NULL );
+        hb_buffer_t * vid_buf = NULL;
+        int vidskip = 0;
+
+        if ( title->flags & HBTF_NO_IDR )
+        {
+            // title doesn't have IDR frames so we decode but drop one second's
+            // worth of frames to allow the decoder to converge.
+            if ( ! title->rate_base )
+            {
+                vidskip = 30;
+            }
+            else
+            {
+                vidskip = (double)title->rate / (double)title->rate_base + 0.5;
+            }
+        }
 
         for( j = 0; j < 10240 ; j++ )
         {
@@ -327,6 +438,10 @@ static int DecodePreviews( hb_scan_t * data, hb_title_t * title )
             {
               if( !hb_dvd_read( data->dvd, buf_ps ) )
               {
+                  if ( vid_buf )
+                  {
+                    break;
+                  }
                   hb_log( "Warning: Could not read data for preview %d, skipped", i + 1 );
                   goto skip_preview;
               }
@@ -335,62 +450,72 @@ static int DecodePreviews( hb_scan_t * data, hb_title_t * title )
             {
               if ( !hb_stream_read(data->stream,buf_ps) )
               {
+                  if ( vid_buf )
+                  {
+                    break;
+                  }
                   hb_log( "Warning: Could not read data for preview %d, skipped", i + 1 );
                   goto skip_preview;
               }
             }
-            hb_demux_ps( buf_ps, list_es );
+            (hb_demux[title->demuxer])(buf_ps, list_es, 0 );
 
             while( ( buf_es = hb_list_item( list_es, 0 ) ) )
             {
                 hb_list_rem( list_es, buf_es );
-                if( buf_es->id == 0xE0 && !hb_list_count( list_raw ) )
+                if( buf_es->id == title->video_id && vid_buf == NULL )
                 {
-                    hb_libmpeg2_decode( mpeg2, buf_es, list_raw );
-                    int ar = hb_libmpeg2_clear_aspect_ratio( mpeg2 );
-                    if ( ar != 0 )
+                    vid_decoder->work( vid_decoder, &buf_es, &vid_buf );
+                    if ( vid_buf && vidskip && --vidskip > 0 )
                     {
-                        ( ar == (HB_ASPECT_BASE * 4 / 3) ) ?
-                            ++ar4_count : ++ar16_count ;
+                        // we're dropping frames to get the video decoder in sync
+                        // when the video stream doesn't contain IDR frames
+                        hb_buffer_close( &vid_buf );
+                        vid_buf = NULL;
                     }
                 }
-                else if( !i )
+                else if( ! AllAudioOK( title ) )
                 {
-                    LookForAC3AndDCA( title, buf_es );
-                }
-                hb_buffer_close( &buf_es );
-
-                if( hb_list_count( list_raw ) &&
-                    ( i || AllAC3AndDCAOK( title ) ) )
-                {
-                    /* We got a picture */
-                    break;
+                    LookForAudio( title, buf_es );
                 }
+                if ( buf_es )
+                    hb_buffer_close( &buf_es );
             }
 
-            if( hb_list_count( list_raw ) &&
-                ( i || AllAC3AndDCAOK( title ) ) )
-            {
+            if( vid_buf && AllAudioOK( title ) )
                 break;
-            }
         }
 
-        if( !hb_list_count( list_raw ) )
+        if( ! vid_buf )
         {
             hb_log( "scan: could not get a decoded picture" );
-            goto error;
+            continue;
         }
 
         /* Get size and rate infos */
-        title->rate = 27000000;
-        int ar;
-        hb_libmpeg2_info( mpeg2, &title->width, &title->height,
-                          &title->rate_base, &ar );
 
-        /* if we found mostly 4:3 previews use that as the aspect ratio otherwise
-           use 16:9 */
-        title->aspect = ar4_count > ar16_count ?
-                            HB_ASPECT_BASE * 4 / 3 : HB_ASPECT_BASE * 16 / 9;
+        hb_work_info_t vid_info;
+        if( !vid_decoder->info( vid_decoder, &vid_info ) )
+        {
+            /*
+             * Could not fill vid_info, don't continue and try to use vid_info
+             * in this case.
+             */
+            vid_decoder->close( vid_decoder );
+            free( vid_decoder );
+            continue;
+        }
+        vid_decoder->close( vid_decoder );
+        free( vid_decoder );
+
+        remember_info( info_list, &vid_info );
+
+        title->video_codec_name = strdup( vid_info.name );
+        title->width = vid_info.width;
+        title->height = vid_info.height;
+        title->rate = vid_info.rate;
+        title->rate_base = vid_info.rate_base;
+        title->video_bitrate = vid_info.bitrate;
 
         if( title->rate_base == 1126125 )
         {
@@ -416,7 +541,7 @@ static int DecodePreviews( hb_scan_t * data, hb_title_t * title )
                 */
                 if( progressive_count == 6 )
                 {
-                    hb_log("Title's mostly NTSC Film, setting fps to 23.976");
+                    hb_deep_log( 2, "Title's mostly NTSC Film, setting fps to 23.976");
                 }
                 title->rate_base = 1126125;
             }
@@ -430,121 +555,189 @@ static int DecodePreviews( hb_scan_t * data, hb_title_t * title )
             title->rate_base = 1126125;
         }
 
-        // start from third frame to skip opening logos
-        if( i == 2)
-        {
-            title->crop[0] = title->crop[1] = title->height / 2;
-            title->crop[2] = title->crop[3] = title->width / 2;
-        }
-
-        hb_libmpeg2_close( &mpeg2 );
-
         while( ( buf_es = hb_list_item( list_es, 0 ) ) )
         {
             hb_list_rem( list_es, buf_es );
             hb_buffer_close( &buf_es );
         }
 
-        buf_raw = hb_list_item( list_raw, 0 );
-
         /* Check preview for interlacing artifacts */
-        if( hb_detect_comb( buf_raw, title->width, title->height, 10, 30, 9 ) )
+        if( hb_detect_comb( vid_buf, title->width, title->height, 10, 30, 9, 10, 30, 9 ) )
         {
-            hb_log("Interlacing detected in preview frame %i", i);
+            hb_deep_log( 2, "Interlacing detected in preview frame %i", i+1);
             interlaced_preview_count++;
         }
-
-        hb_get_tempory_filename( data->h, filename, "%x%d",
-                                 (intptr_t)title, i );
-
-        file_preview = fopen( filename, "w" );
-        if( file_preview )
+        
+        if( data->store_previews )
         {
-            fwrite( buf_raw->data, title->width * title->height * 3 / 2,
-                    1, file_preview );
-            fclose( file_preview );
-        }
-        else
-        {
-            hb_log( "scan: fopen failed (%s)", filename );
-        }
+            hb_get_tempory_filename( data->h, filename, "%" PRIxPTR "%d",
+                                     (intptr_t)title, i );
 
-#define Y    buf_raw->data
-#define DARK 64
+            file_preview = fopen( filename, "wb" );
+            if( file_preview )
+            {
+                fwrite( vid_buf->data, title->width * title->height * 3 / 2,
+                        1, file_preview );
+                fclose( file_preview );
+            }
+            else
+            {
+                hb_log( "scan: fopen failed (%s)", filename );
+            }
+        }
 
         /* Detect black borders */
 
-        for( j = 0; j < title->width; j++ )
+#define Y    vid_buf->data
+        int top, bottom, left, right;
+        int h4 = title->height / 4, w4 = title->width / 4;
+
+        // When widescreen content is matted to 16:9 or 4:3 there's sometimes
+        // a thin border on the outer edge of the matte. On TV content it can be
+        // "line 21" VBI data that's normally hidden in the overscan. For HD
+        // content it can just be a diagnostic added in post production so that
+        // the frame borders are visible. We try to ignore these borders so
+        // we can crop the matte. The border width depends on the resolution
+        // (12 pixels on 1080i looks visually the same as 4 pixels on 480i)
+        // so we allow the border to be up to 1% of the frame height.
+        const int border = title->height / 100;
+
+        for ( top = border; top < h4; ++top )
         {
-            for( k = 0; k < title->crop[0]; k++ )
-                if( Y[ k * title->width + j ] > DARK )
-                {
-                    title->crop[0] = k;
-                    break;
-                }
-            for( k = 0; k < title->crop[1]; k++ )
-                if( Y[ ( title->height - k - 1 ) *
-                       title->width + j ] > DARK )
-                {
-                    title->crop[1] = k;
-                    break;
-                }
+            if ( ! row_all_dark( title, Y, top ) )
+                break;
         }
-        for( j = 0; j < title->height; j++ )
+        if ( top <= border )
         {
-            for( k = 0; k < title->crop[2]; k++ )
-                if( Y[ j * title->width + k ] > DARK )
-                {
-                    title->crop[2] = k;
+            // we never made it past the border region - see if the rows we
+            // didn't check are dark or if we shouldn't crop at all.
+            for ( top = 0; top < border; ++top )
+            {
+                if ( ! row_all_dark( title, Y, top ) )
                     break;
-                }
-            for( k = 0; k < title->crop[3]; k++ )
-                if( Y[ j * title->width +
-                        title->width - k - 1 ] > DARK )
-                {
-                    title->crop[3] = k;
+            }
+            if ( top >= border )
+            {
+                top = 0;
+            }
+        }
+        for ( bottom = border; bottom < h4; ++bottom )
+        {
+            if ( ! row_all_dark( title, Y, title->height - 1 - bottom ) )
+                break;
+        }
+        if ( bottom <= border )
+        {
+            for ( bottom = 0; bottom < border; ++bottom )
+            {
+                if ( ! row_all_dark( title, Y, title->height - 1 - bottom ) )
                     break;
-                }
+            }
+            if ( bottom >= border )
+            {
+                bottom = 0;
+            }
+        }
+        for ( left = 0; left < w4; ++left )
+        {
+            if ( ! column_all_dark( title, Y, top, bottom, left ) )
+                break;
+        }
+        for ( right = 0; right < w4; ++right )
+        {
+            if ( ! column_all_dark( title, Y, top, bottom, title->width - 1 - right ) )
+                break;
+        }
+
+        // only record the result if all the crops are less than a quarter of
+        // the frame otherwise we can get fooled by frames with a lot of black
+        // like titles, credits & fade-thru-black transitions.
+        if ( top < h4 && bottom < h4 && left < w4 && right < w4 )
+        {
+            record_crop( crops, top, bottom, left, right );
         }
         ++npreviews;
 
 skip_preview:
-        while( ( buf_raw = hb_list_item( list_raw, 0 ) ) )
+        if ( vid_buf )
+            hb_buffer_close( &vid_buf );
+    }
+
+    if ( npreviews )
+    {
+        // use the most common frame info for our final title dimensions
+        hb_work_info_t vid_info;
+        most_common_info( info_list, &vid_info );
+
+        title->width = vid_info.width;
+        title->height = vid_info.height;
+        title->pixel_aspect_width = vid_info.pixel_aspect_width;
+        title->pixel_aspect_height = vid_info.pixel_aspect_height;
+
+        // compute the aspect ratio based on the storage dimensions and the
+        // pixel aspect ratio (if supplied) or just storage dimensions if no PAR.
+        title->aspect = (double)title->width / (double)title->height;
+        if( title->pixel_aspect_width && title->pixel_aspect_height )
         {
-            hb_list_rem( list_raw, buf_raw );
-            hb_buffer_close( &buf_raw );
+            title->aspect *= (double)title->pixel_aspect_width /
+                             (double)title->pixel_aspect_height;
+
+            // For unknown reasons some French PAL DVDs put the original
+            // content's aspect ratio into the mpeg PAR even though it's
+            // the wrong PAR for the DVD. Apparently they rely on the fact
+            // that DVD players ignore the content PAR and just use the
+            // aspect ratio from the DVD metadata. So, if the aspect computed
+            // from the PAR is different from the container's aspect we use
+            // the container's aspect & recompute the PAR from it.
+            if( title->container_aspect && (int)(title->aspect * 9) != (int)(title->container_aspect * 9) )
+            {
+                hb_log("scan: content PAR gives wrong aspect %.2f; "
+                       "using container aspect %.2f", title->aspect,
+                       title->container_aspect );
+                title->aspect = title->container_aspect;
+                hb_reduce( &title->pixel_aspect_width, &title->pixel_aspect_height,
+                           (int)(title->aspect * title->height + 0.5), title->width );
+            }
         }
-    }
 
-    title->crop[0] = EVEN( title->crop[0] );
-    title->crop[1] = EVEN( title->crop[1] );
-    title->crop[2] = EVEN( title->crop[2] );
-    title->crop[3] = EVEN( title->crop[3] );
+        // don't try to crop unless we got at least 3 previews
+        if ( crops->n > 2 )
+        {
+            sort_crops( crops );
+            // The next line selects median cropping - at least
+            // 50% of the frames will have their borders removed.
+            // Other possible choices are loose cropping (i = 0) where 
+            // no non-black pixels will be cropped from any frame and a
+            // tight cropping (i = crops->n - (crops->n >> 2)) where at
+            // least 75% of the frames will have their borders removed.
+            i = crops->n >> 1;
+            title->crop[0] = EVEN( crops->t[i] );
+            title->crop[1] = EVEN( crops->b[i] );
+            title->crop[2] = EVEN( crops->l[i] );
+            title->crop[3] = EVEN( crops->r[i] );
+        }
+        free( crops );
 
-    hb_log( "scan: %d previews, %dx%d, %.3f fps, autocrop = %d/%d/%d/%d, aspect %s",
-            npreviews, title->width, title->height, (float) title->rate /
-            (float) title->rate_base, title->crop[0], title->crop[1],
-            title->crop[2], title->crop[3],
-            title->aspect == HB_ASPECT_BASE * 16 / 9 ? "16:9" :
-                title->aspect == HB_ASPECT_BASE * 4 / 3 ? "4:3" : "none" );
+        hb_log( "scan: %d previews, %dx%d, %.3f fps, autocrop = %d/%d/%d/%d, "
+                "aspect %s, PAR %d:%d",
+                npreviews, title->width, title->height, (float) title->rate /
+                (float) title->rate_base,
+                title->crop[0], title->crop[1], title->crop[2], title->crop[3],
+                aspect_to_string( title->aspect ), title->pixel_aspect_width,
+                title->pixel_aspect_height );
 
-    if( interlaced_preview_count >= ( npreviews / 2 ) )
-    {
-        hb_log("Title is likely interlaced or telecined (%i out of %i previews). You should do something about that.",
-               interlaced_preview_count, npreviews);
-        title->detected_interlacing = 1;
-    }
-    else
-    {
-        title->detected_interlacing = 0;
+        if( interlaced_preview_count >= ( npreviews / 2 ) )
+        {
+            hb_log("Title is likely interlaced or telecined (%i out of %i previews). You should do something about that.",
+                   interlaced_preview_count, npreviews);
+            title->detected_interlacing = 1;
+        }
+        else
+        {
+            title->detected_interlacing = 0;
+        }
     }
 
-    goto cleanup;
-
-error:
-    npreviews = 0;
-
-cleanup:
     hb_buffer_close( &buf_ps );
     while( ( buf_es = hb_list_item( list_es, 0 ) ) )
     {
@@ -552,35 +745,36 @@ cleanup:
         hb_buffer_close( &buf_es );
     }
     hb_list_close( &list_es );
-    while( ( buf_raw = hb_list_item( list_raw, 0 ) ) )
-    {
-        hb_list_rem( list_raw, buf_raw );
-        hb_buffer_close( &buf_raw );
-    }
-    hb_list_close( &list_raw );
     if (data->dvd)
       hb_dvd_stop( data->dvd );
 
     return npreviews;
 }
 
-static void LookForAC3AndDCA( hb_title_t * title, hb_buffer_t * b )
+/*
+ * This routine is called for every frame from a non-video elementary stream.
+ * These are a mix of audio & subtitle streams, some of which we want & some
+ * we're ignoring. This routine checks the frame against all our audio streams
+ * to see if it's one we want and haven't identified yet. If yes, it passes the
+ * frame to a codec-specific id routine which is responsible for filling in
+ * the sample rate, bit rate, channels & other audio parameters.
+ *
+ * Since a sample rate is essential for further audio processing, any audio
+ * stream which isn't successfully id'd by is deleted at the end of the scan.
+ * This is necessary to avoid ambiguities where things that might be audio
+ * aren't (e.g., some European DVD Teletext streams use the same IDs as US ATSC
+ * AC-3 audio).
+ */
+static void LookForAudio( hb_title_t * title, hb_buffer_t * b )
 {
     int i;
-    int flags;
-    int rate;
-    int bitrate;
-    int frame_length;
-    dca_state_t * state;
 
-    /* Figure out if this is a AC3 or DCA buffer for a known track */
     hb_audio_t * audio = NULL;
     for( i = 0; i < hb_list_count( title->list_audio ); i++ )
     {
         audio = hb_list_item( title->list_audio, i );
-        /* check if we have an AC3 or DCA which we recognise */
-        if( ( audio->config.in.codec == HB_ACODEC_AC3 || audio->config.in.codec == HB_ACODEC_DCA ) &&
-            audio->id    == b->id )
+        /* check if this elementary stream is one we want */
+        if ( audio->id == b->id )
         {
             break;
         }
@@ -589,181 +783,85 @@ static void LookForAC3AndDCA( hb_title_t * title, hb_buffer_t * b )
             audio = NULL;
         }
     }
-    if( !audio )
+    if( !audio || audio->config.in.bitrate != 0 )
     {
+        /* not found or already done */
         return;
     }
 
-    if( audio->config.in.bitrate )
+    hb_work_object_t *w = hb_codec_decoder( audio->config.in.codec );
+
+    if ( w == NULL || w->bsinfo == NULL )
     {
-        /* Already done for this track */
-        return;
+        hb_log( "Internal error in scan: unhandled audio type %d for id 0x%x",
+                audio->config.in.codec, audio->id );
+        goto drop_audio;
     }
 
-    for( i = 0; i < b->size - 7; i++ )
+    hb_work_info_t info;
+    w->audio = audio;
+    w->codec_param = audio->config.in.codec_param;
+    int ret = w->bsinfo( w, b, &info );
+    if ( ret < 0 )
     {
-
-        if ( audio->config.in.codec == HB_ACODEC_AC3 )
-        {
-
-            /* check for a52 */
-            if( a52_syncinfo( &b->data[i], &flags, &rate, &bitrate ) )
-            {
-                hb_log( "scan: AC3, rate=%dHz, bitrate=%d", rate, bitrate );
-                audio->config.in.samplerate = rate;
-                audio->config.in.bitrate = bitrate;
-                switch( flags & A52_CHANNEL_MASK )
-                {
-                    /* mono sources */
-                    case A52_MONO:
-                    case A52_CHANNEL1:
-                    case A52_CHANNEL2:
-                        audio->config.in.channel_layout = HB_INPUT_CH_LAYOUT_MONO;
-                        break;
-                    /* stereo input */
-                    case A52_CHANNEL:
-                    case A52_STEREO:
-                        audio->config.in.channel_layout = HB_INPUT_CH_LAYOUT_STEREO;
-                        break;
-                    /* dolby (DPL1 aka Dolby Surround = 4.0 matrix-encoded) input */
-                    case A52_DOLBY:
-                        audio->config.in.channel_layout = HB_INPUT_CH_LAYOUT_DOLBY;
-                        break;
-                    /* 3F/2R input */
-                    case A52_3F2R:
-                        audio->config.in.channel_layout = HB_INPUT_CH_LAYOUT_3F2R;
-                        break;
-                    /* 3F/1R input */
-                    case A52_3F1R:
-                        audio->config.in.channel_layout = HB_INPUT_CH_LAYOUT_3F1R;
-                        break;
-                    /* other inputs */
-                    case A52_3F:
-                        audio->config.in.channel_layout = HB_INPUT_CH_LAYOUT_3F;
-                        break;
-                    case A52_2F1R:
-                        audio->config.in.channel_layout = HB_INPUT_CH_LAYOUT_2F1R;
-                        break;
-                    case A52_2F2R:
-                        audio->config.in.channel_layout = HB_INPUT_CH_LAYOUT_2F2R;
-                        break;
-                    /* unknown */
-                    default:
-                        audio->config.in.channel_layout = HB_INPUT_CH_LAYOUT_STEREO;
-                }
-
-                /* add in our own LFE flag if the source has LFE */
-                if (flags & A52_LFE)
-                {
-                    audio->config.in.channel_layout = audio->config.in.channel_layout | HB_INPUT_CH_LAYOUT_HAS_LFE;
-                }
-
-                /* store the AC3 flags for future reference
-                 * This enables us to find out if we had a stereo or Dolby source later on
-                 * Store the ac3 flags in the public ac3flags property too, so we can access it from the GUI
-                 */
-                audio->config.flags.ac3 = audio->priv.config.a52.ac3flags = flags;
-
-                /* XXX */
-                if ( (flags & A52_CHANNEL_MASK) == A52_DOLBY ) {
-                    sprintf( audio->config.lang.description + strlen( audio->config.lang.description ),
-                         " (Dolby Surround)" );
-                } else {
-                    sprintf( audio->config.lang.description + strlen( audio->config.lang.description ),
-                         " (%d.%d ch)",
-                        HB_INPUT_CH_LAYOUT_GET_DISCRETE_FRONT_COUNT(audio->config.in.channel_layout) +
-                        HB_INPUT_CH_LAYOUT_GET_DISCRETE_REAR_COUNT(audio->config.in.channel_layout),
-                        HB_INPUT_CH_LAYOUT_GET_DISCRETE_LFE_COUNT(audio->config.in.channel_layout));
-                }
-
-                break;
-
-            }
-
-        }
-        else if ( audio->config.in.codec == HB_ACODEC_DCA )
-        {
-
-            hb_log( "scan: checking for DCA syncinfo" );
-
-            /* check for dca */
-            state = dca_init( 0 );
-            if( dca_syncinfo( state, &b->data[i], &flags, &rate, &bitrate, &frame_length ) )
-            {
-                hb_log( "scan: DCA, rate=%dHz, bitrate=%d", rate, bitrate );
-                audio->config.in.samplerate    = rate;
-                audio->config.in.bitrate = bitrate;
-                switch( flags & DCA_CHANNEL_MASK )
-                {
-                    /* mono sources */
-                    case DCA_MONO:
-                        audio->config.in.channel_layout = HB_INPUT_CH_LAYOUT_MONO;
-                        break;
-                    /* stereo input */
-                    case DCA_CHANNEL:
-                    case DCA_STEREO:
-                    case DCA_STEREO_SUMDIFF:
-                    case DCA_STEREO_TOTAL:
-                        audio->config.in.channel_layout = HB_INPUT_CH_LAYOUT_STEREO;
-                        break;
-                    /* 3F/2R input */
-                    case DCA_3F2R:
-                        audio->config.in.channel_layout = HB_INPUT_CH_LAYOUT_3F2R;
-                        break;
-                    /* 3F/1R input */
-                    case DCA_3F1R:
-                        audio->config.in.channel_layout = HB_INPUT_CH_LAYOUT_3F1R;
-                        break;
-                    /* other inputs */
-                    case DCA_3F:
-                        audio->config.in.channel_layout = HB_INPUT_CH_LAYOUT_3F;
-                        break;
-                    case DCA_2F1R:
-                        audio->config.in.channel_layout = HB_INPUT_CH_LAYOUT_2F1R;
-                        break;
-                    case DCA_2F2R:
-                        audio->config.in.channel_layout = HB_INPUT_CH_LAYOUT_2F2R;
-                        break;
-                    case DCA_4F2R:
-                        audio->config.in.channel_layout = HB_INPUT_CH_LAYOUT_4F2R;
-                        break;
-                    /* unknown */
-                    default:
-                        audio->config.in.channel_layout = HB_INPUT_CH_LAYOUT_STEREO;
-                }
-
-                /* add in our own LFE flag if the source has LFE */
-                if (flags & DCA_LFE)
-                {
-                    audio->config.in.channel_layout = audio->config.in.channel_layout | HB_INPUT_CH_LAYOUT_HAS_LFE;
-                }
-
-                /* store the DCA flags for future reference
-                 * This enables us to find out if we had a stereo or Dolby source later on
-                 * store the dca flags in the public dcaflags property too, so we can access it from the GUI
-                 */
-                audio->config.flags.dca = audio->priv.config.dca.dcaflags = flags;
-
-                /* XXX */
-                if ( (flags & DCA_CHANNEL_MASK) == DCA_DOLBY ) {
-                    sprintf( audio->config.lang.description + strlen( audio->config.lang.description ),
-                         " (Dolby Surround)" );
-                } else {
-                    sprintf( audio->config.lang.description + strlen( audio->config.lang.description ),
-                         " (%d.%d ch)",
-                        HB_INPUT_CH_LAYOUT_GET_DISCRETE_FRONT_COUNT(audio->config.in.channel_layout) +
-                        HB_INPUT_CH_LAYOUT_GET_DISCRETE_REAR_COUNT(audio->config.in.channel_layout),
-                        HB_INPUT_CH_LAYOUT_GET_DISCRETE_LFE_COUNT(audio->config.in.channel_layout));
-                }
-
-                break;
-            }
-        }
+        hb_log( "no info on audio type %d/0x%x for id 0x%x",
+                audio->config.in.codec, audio->config.in.codec_param,
+                audio->id );
+        goto drop_audio;
+    }
+    if ( !info.bitrate )
+    {
+        /* didn't find any info */
+        return;
+    }
+    audio->config.in.samplerate = info.rate;
+    audio->config.in.bitrate = info.bitrate;
+    audio->config.in.channel_layout = info.channel_layout;
+    audio->config.in.version = info.version;
+    audio->config.in.mode = info.mode;
+    audio->config.flags.ac3 = info.flags;
+
+    // update the audio description string based on the info we found
+    if ( audio->config.flags.ac3 & AUDIO_F_DOLBY )
+    {
+        strcat( audio->config.lang.description, " (Dolby Surround)" );
+    }
+    else
+    {
+        int layout = audio->config.in.channel_layout;
+        char *desc = audio->config.lang.description +
+                        strlen( audio->config.lang.description );
+        sprintf( desc, " (%d.%d ch)",
+                 HB_INPUT_CH_LAYOUT_GET_DISCRETE_FRONT_COUNT(layout) +
+                     HB_INPUT_CH_LAYOUT_GET_DISCRETE_REAR_COUNT(layout),
+                 HB_INPUT_CH_LAYOUT_GET_DISCRETE_LFE_COUNT(layout) );
     }
 
+    hb_log( "scan: audio 0x%x: %s, rate=%dHz, bitrate=%d %s", audio->id,
+            info.name, audio->config.in.samplerate, audio->config.in.bitrate,
+            audio->config.lang.description );
+    free( w );
+    return;
+
+    // We get here if there's no hope of finding info on an audio bitstream,
+    // either because we don't have a decoder (or a decoder with a bitstream
+    // info proc) or because the decoder's info proc said that the stream
+    // wasn't something it could handle. Delete the item from the title's
+    // audio list so we won't keep reading packets while trying to get its
+    // bitstream info.
+ drop_audio:
+    if ( w )
+        free( w );
+
+    hb_list_rem( title->list_audio, audio );
 }
 
-static int  AllAC3AndDCAOK( hb_title_t * title )
+/*
+ * This routine checks to see if we've ID'd all the audio streams associated
+ * with a title. It returns 0 if there are more to ID & 1 if all are done.
+ */
+static int  AllAudioOK( hb_title_t * title )
 {
     int i;
     hb_audio_t * audio;
@@ -771,12 +869,10 @@ static int  AllAC3AndDCAOK( hb_title_t * title )
     for( i = 0; i < hb_list_count( title->list_audio ); i++ )
     {
         audio = hb_list_item( title->list_audio, i );
-        if( ( audio->config.in.codec == HB_ACODEC_AC3 || audio->config.in.codec == HB_ACODEC_DCA ) &&
-            !audio->config.in.bitrate )
+        if( audio->config.in.bitrate == 0 )
         {
             return 0;
         }
     }
-
     return 1;
 }