Subversion Repositories Kolibri OS

Rev

Go to most recent revision | Blame | Last modification | View Log | RSS feed

  1. /*
  2.  * PGS subtitle decoder
  3.  * Copyright (c) 2009 Stephen Backway
  4.  *
  5.  * This file is part of FFmpeg.
  6.  *
  7.  * FFmpeg is free software; you can redistribute it and/or
  8.  * modify it under the terms of the GNU Lesser General Public
  9.  * License as published by the Free Software Foundation; either
  10.  * version 2.1 of the License, or (at your option) any later version.
  11.  *
  12.  * FFmpeg is distributed in the hope that it will be useful,
  13.  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14.  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  15.  * Lesser General Public License for more details.
  16.  *
  17.  * You should have received a copy of the GNU Lesser General Public
  18.  * License along with FFmpeg; if not, write to the Free Software
  19.  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  20.  */
  21.  
  22. /**
  23.  * @file
  24.  * PGS subtitle decoder
  25.  */
  26.  
  27. #include "avcodec.h"
  28. #include "dsputil.h"
  29. #include "bytestream.h"
  30. #include "libavutil/colorspace.h"
  31. #include "libavutil/imgutils.h"
  32. #include "libavutil/opt.h"
  33.  
  34. #define RGBA(r,g,b,a) (((a) << 24) | ((r) << 16) | ((g) << 8) | (b))
  35.  
  36. enum SegmentType {
  37.     PALETTE_SEGMENT      = 0x14,
  38.     PICTURE_SEGMENT      = 0x15,
  39.     PRESENTATION_SEGMENT = 0x16,
  40.     WINDOW_SEGMENT       = 0x17,
  41.     DISPLAY_SEGMENT      = 0x80,
  42. };
  43.  
  44. typedef struct PGSSubPictureReference {
  45.     int x;
  46.     int y;
  47.     int picture_id;
  48.     int composition;
  49. } PGSSubPictureReference;
  50.  
  51. typedef struct PGSSubPresentation {
  52.     int                    id_number;
  53.     int                    object_count;
  54.     PGSSubPictureReference *objects;
  55.     int64_t pts;
  56. } PGSSubPresentation;
  57.  
  58. typedef struct PGSSubPicture {
  59.     int          w;
  60.     int          h;
  61.     uint8_t      *rle;
  62.     unsigned int rle_buffer_size, rle_data_len;
  63.     unsigned int rle_remaining_len;
  64. } PGSSubPicture;
  65.  
  66. typedef struct PGSSubContext {
  67.     AVClass *class;
  68.     PGSSubPresentation presentation;
  69.     uint32_t           clut[256];
  70.     PGSSubPicture      pictures[UINT16_MAX];
  71.     int forced_subs_only;
  72. } PGSSubContext;
  73.  
  74. static av_cold int init_decoder(AVCodecContext *avctx)
  75. {
  76.     avctx->pix_fmt     = AV_PIX_FMT_PAL8;
  77.  
  78.     return 0;
  79. }
  80.  
  81. static av_cold int close_decoder(AVCodecContext *avctx)
  82. {
  83.     uint16_t picture;
  84.  
  85.     PGSSubContext *ctx = avctx->priv_data;
  86.  
  87.     av_freep(&ctx->presentation.objects);
  88.     ctx->presentation.object_count = 0;
  89.  
  90.     for (picture = 0; picture < UINT16_MAX; ++picture) {
  91.         av_freep(&ctx->pictures[picture].rle);
  92.         ctx->pictures[picture].rle_buffer_size = 0;
  93.     }
  94.  
  95.     return 0;
  96. }
  97.  
  98. /**
  99.  * Decode the RLE data.
  100.  *
  101.  * The subtitle is stored as a Run Length Encoded image.
  102.  *
  103.  * @param avctx contains the current codec context
  104.  * @param sub pointer to the processed subtitle data
  105.  * @param buf pointer to the RLE data to process
  106.  * @param buf_size size of the RLE data to process
  107.  */
  108. static int decode_rle(AVCodecContext *avctx, AVSubtitle *sub, int rect,
  109.                       const uint8_t *buf, unsigned int buf_size)
  110. {
  111.     const uint8_t *rle_bitmap_end;
  112.     int pixel_count, line_count;
  113.  
  114.     rle_bitmap_end = buf + buf_size;
  115.  
  116.     sub->rects[rect]->pict.data[0] = av_malloc(sub->rects[rect]->w * sub->rects[rect]->h);
  117.  
  118.     if (!sub->rects[rect]->pict.data[0])
  119.         return -1;
  120.  
  121.     pixel_count = 0;
  122.     line_count  = 0;
  123.  
  124.     while (buf < rle_bitmap_end && line_count < sub->rects[rect]->h) {
  125.         uint8_t flags, color;
  126.         int run;
  127.  
  128.         color = bytestream_get_byte(&buf);
  129.         run   = 1;
  130.  
  131.         if (color == 0x00) {
  132.             flags = bytestream_get_byte(&buf);
  133.             run   = flags & 0x3f;
  134.             if (flags & 0x40)
  135.                 run = (run << 8) + bytestream_get_byte(&buf);
  136.             color = flags & 0x80 ? bytestream_get_byte(&buf) : 0;
  137.         }
  138.  
  139.         if (run > 0 && pixel_count + run <= sub->rects[rect]->w * sub->rects[rect]->h) {
  140.             memset(sub->rects[rect]->pict.data[0] + pixel_count, color, run);
  141.             pixel_count += run;
  142.         } else if (!run) {
  143.             /*
  144.              * New Line. Check if correct pixels decoded, if not display warning
  145.              * and adjust bitmap pointer to correct new line position.
  146.              */
  147.             if (pixel_count % sub->rects[rect]->w > 0)
  148.                 av_log(avctx, AV_LOG_ERROR, "Decoded %d pixels, when line should be %d pixels\n",
  149.                        pixel_count % sub->rects[rect]->w, sub->rects[rect]->w);
  150.             line_count++;
  151.         }
  152.     }
  153.  
  154.     if (pixel_count < sub->rects[rect]->w * sub->rects[rect]->h) {
  155.         av_log(avctx, AV_LOG_ERROR, "Insufficient RLE data for subtitle\n");
  156.         return -1;
  157.     }
  158.  
  159.     av_dlog(avctx, "Pixel Count = %d, Area = %d\n", pixel_count, sub->rects[rect]->w * sub->rects[rect]->h);
  160.  
  161.     return 0;
  162. }
  163.  
  164. /**
  165.  * Parse the picture segment packet.
  166.  *
  167.  * The picture segment contains details on the sequence id,
  168.  * width, height and Run Length Encoded (RLE) bitmap data.
  169.  *
  170.  * @param avctx contains the current codec context
  171.  * @param buf pointer to the packet to process
  172.  * @param buf_size size of packet to process
  173.  * @todo TODO: Enable support for RLE data over multiple packets
  174.  */
  175. static int parse_picture_segment(AVCodecContext *avctx,
  176.                                   const uint8_t *buf, int buf_size)
  177. {
  178.     PGSSubContext *ctx = avctx->priv_data;
  179.  
  180.     uint8_t sequence_desc;
  181.     unsigned int rle_bitmap_len, width, height;
  182.     uint16_t picture_id;
  183.  
  184.     if (buf_size <= 4)
  185.         return -1;
  186.     buf_size -= 4;
  187.  
  188.     picture_id = bytestream_get_be16(&buf);
  189.  
  190.     /* skip 1 unknown byte: Version Number */
  191.     buf++;
  192.  
  193.     /* Read the Sequence Description to determine if start of RLE data or appended to previous RLE */
  194.     sequence_desc = bytestream_get_byte(&buf);
  195.  
  196.     if (!(sequence_desc & 0x80)) {
  197.         /* Additional RLE data */
  198.         if (buf_size > ctx->pictures[picture_id].rle_remaining_len)
  199.             return -1;
  200.  
  201.         memcpy(ctx->pictures[picture_id].rle + ctx->pictures[picture_id].rle_data_len, buf, buf_size);
  202.         ctx->pictures[picture_id].rle_data_len += buf_size;
  203.         ctx->pictures[picture_id].rle_remaining_len -= buf_size;
  204.  
  205.         return 0;
  206.     }
  207.  
  208.     if (buf_size <= 7)
  209.         return -1;
  210.     buf_size -= 7;
  211.  
  212.     /* Decode rle bitmap length, stored size includes width/height data */
  213.     rle_bitmap_len = bytestream_get_be24(&buf) - 2*2;
  214.  
  215.     /* Get bitmap dimensions from data */
  216.     width  = bytestream_get_be16(&buf);
  217.     height = bytestream_get_be16(&buf);
  218.  
  219.     /* Make sure the bitmap is not too large */
  220.     if (avctx->width < width || avctx->height < height) {
  221.         av_log(avctx, AV_LOG_ERROR, "Bitmap dimensions larger than video.\n");
  222.         return -1;
  223.     }
  224.  
  225.     if (buf_size > rle_bitmap_len) {
  226.         av_log(avctx, AV_LOG_ERROR, "too much RLE data\n");
  227.         return AVERROR_INVALIDDATA;
  228.     }
  229.  
  230.     ctx->pictures[picture_id].w = width;
  231.     ctx->pictures[picture_id].h = height;
  232.  
  233.     av_fast_malloc(&ctx->pictures[picture_id].rle, &ctx->pictures[picture_id].rle_buffer_size, rle_bitmap_len);
  234.  
  235.     if (!ctx->pictures[picture_id].rle)
  236.         return -1;
  237.  
  238.     memcpy(ctx->pictures[picture_id].rle, buf, buf_size);
  239.     ctx->pictures[picture_id].rle_data_len      = buf_size;
  240.     ctx->pictures[picture_id].rle_remaining_len = rle_bitmap_len - buf_size;
  241.  
  242.     return 0;
  243. }
  244.  
  245. /**
  246.  * Parse the palette segment packet.
  247.  *
  248.  * The palette segment contains details of the palette,
  249.  * a maximum of 256 colors can be defined.
  250.  *
  251.  * @param avctx contains the current codec context
  252.  * @param buf pointer to the packet to process
  253.  * @param buf_size size of packet to process
  254.  */
  255. static void parse_palette_segment(AVCodecContext *avctx,
  256.                                   const uint8_t *buf, int buf_size)
  257. {
  258.     PGSSubContext *ctx = avctx->priv_data;
  259.  
  260.     const uint8_t *buf_end = buf + buf_size;
  261.     const uint8_t *cm      = ff_cropTbl + MAX_NEG_CROP;
  262.     int color_id;
  263.     int y, cb, cr, alpha;
  264.     int r, g, b, r_add, g_add, b_add;
  265.  
  266.     /* Skip two null bytes */
  267.     buf += 2;
  268.  
  269.     while (buf < buf_end) {
  270.         color_id  = bytestream_get_byte(&buf);
  271.         y         = bytestream_get_byte(&buf);
  272.         cr        = bytestream_get_byte(&buf);
  273.         cb        = bytestream_get_byte(&buf);
  274.         alpha     = bytestream_get_byte(&buf);
  275.  
  276.         YUV_TO_RGB1(cb, cr);
  277.         YUV_TO_RGB2(r, g, b, y);
  278.  
  279.         av_dlog(avctx, "Color %d := (%d,%d,%d,%d)\n", color_id, r, g, b, alpha);
  280.  
  281.         /* Store color in palette */
  282.         ctx->clut[color_id] = RGBA(r,g,b,alpha);
  283.     }
  284. }
  285.  
  286. /**
  287.  * Parse the presentation segment packet.
  288.  *
  289.  * The presentation segment contains details on the video
  290.  * width, video height, x & y subtitle position.
  291.  *
  292.  * @param avctx contains the current codec context
  293.  * @param buf pointer to the packet to process
  294.  * @param buf_size size of packet to process
  295.  * @todo TODO: Implement cropping
  296.  */
  297. static void parse_presentation_segment(AVCodecContext *avctx,
  298.                                        const uint8_t *buf, int buf_size,
  299.                                        int64_t pts)
  300. {
  301.     PGSSubContext *ctx = avctx->priv_data;
  302.  
  303.     int w = bytestream_get_be16(&buf);
  304.     int h = bytestream_get_be16(&buf);
  305.  
  306.     uint16_t object_index;
  307.  
  308.     ctx->presentation.pts = pts;
  309.  
  310.     av_dlog(avctx, "Video Dimensions %dx%d\n",
  311.             w, h);
  312.     if (av_image_check_size(w, h, 0, avctx) >= 0)
  313.         avcodec_set_dimensions(avctx, w, h);
  314.  
  315.     /* Skip 1 bytes of unknown, frame rate? */
  316.     buf++;
  317.  
  318.     ctx->presentation.id_number = bytestream_get_be16(&buf);
  319.  
  320.     /*
  321.      * Skip 3 bytes of unknown:
  322.      *     state
  323.      *     palette_update_flag (0x80),
  324.      *     palette_id_to_use,
  325.      */
  326.     buf += 3;
  327.  
  328.     ctx->presentation.object_count = bytestream_get_byte(&buf);
  329.     if (!ctx->presentation.object_count)
  330.         return;
  331.  
  332.     /* Verify that enough bytes are remaining for all of the objects. */
  333.     buf_size -= 11;
  334.     if (buf_size < ctx->presentation.object_count * 8) {
  335.         ctx->presentation.object_count = 0;
  336.         return;
  337.     }
  338.  
  339.     av_freep(&ctx->presentation.objects);
  340.     ctx->presentation.objects = av_malloc(sizeof(PGSSubPictureReference) * ctx->presentation.object_count);
  341.     if (!ctx->presentation.objects) {
  342.         ctx->presentation.object_count = 0;
  343.         return;
  344.     }
  345.  
  346.     for (object_index = 0; object_index < ctx->presentation.object_count; ++object_index) {
  347.         PGSSubPictureReference *reference = &ctx->presentation.objects[object_index];
  348.         reference->picture_id             = bytestream_get_be16(&buf);
  349.  
  350.         /* Skip window_id_ref */
  351.         buf++;
  352.         /* composition_flag (0x80 - object cropped, 0x40 - object forced) */
  353.         reference->composition = bytestream_get_byte(&buf);
  354.  
  355.         reference->x = bytestream_get_be16(&buf);
  356.         reference->y = bytestream_get_be16(&buf);
  357.  
  358.         /* TODO If cropping, cropping_x, cropping_y, cropping_width, cropping_height (all 2 bytes).*/
  359.         av_dlog(avctx, "Subtitle Placement ID=%d, x=%d, y=%d\n", reference->picture_id, reference->x, reference->y);
  360.  
  361.         if (reference->x > avctx->width || reference->y > avctx->height) {
  362.             av_log(avctx, AV_LOG_ERROR, "Subtitle out of video bounds. x = %d, y = %d, video width = %d, video height = %d.\n",
  363.                    reference->x, reference->y, avctx->width, avctx->height);
  364.             reference->x = 0;
  365.             reference->y = 0;
  366.         }
  367.     }
  368. }
  369.  
  370. /**
  371.  * Parse the display segment packet.
  372.  *
  373.  * The display segment controls the updating of the display.
  374.  *
  375.  * @param avctx contains the current codec context
  376.  * @param data pointer to the data pertaining the subtitle to display
  377.  * @param buf pointer to the packet to process
  378.  * @param buf_size size of packet to process
  379.  * @todo TODO: Fix start time, relies on correct PTS, currently too late
  380.  *
  381.  * @todo TODO: Fix end time, normally cleared by a second display
  382.  * @todo       segment, which is currently ignored as it clears
  383.  * @todo       the subtitle too early.
  384.  */
  385. static int display_end_segment(AVCodecContext *avctx, void *data,
  386.                                const uint8_t *buf, int buf_size)
  387. {
  388.     AVSubtitle    *sub = data;
  389.     PGSSubContext *ctx = avctx->priv_data;
  390.     int64_t pts;
  391.  
  392.     uint16_t rect;
  393.  
  394.     /*
  395.      *      The end display time is a timeout value and is only reached
  396.      *      if the next subtitle is later than timeout or subtitle has
  397.      *      not been cleared by a subsequent empty display command.
  398.      */
  399.  
  400.     pts = ctx->presentation.pts != AV_NOPTS_VALUE ? ctx->presentation.pts : sub->pts;
  401.     memset(sub, 0, sizeof(*sub));
  402.     sub->pts = pts;
  403.     ctx->presentation.pts = AV_NOPTS_VALUE;
  404.  
  405.     // Blank if last object_count was 0.
  406.     if (!ctx->presentation.object_count)
  407.         return 1;
  408.  
  409.     sub->start_display_time = 0;
  410.     sub->end_display_time   = 20000;
  411.     sub->format             = 0;
  412.  
  413.     sub->num_rects = ctx->presentation.object_count;
  414.     sub->rects     = av_mallocz(sizeof(*sub->rects) * sub->num_rects);
  415.  
  416.     for (rect = 0; rect < sub->num_rects; ++rect) {
  417.         uint16_t picture_id    = ctx->presentation.objects[rect].picture_id;
  418.         sub->rects[rect]       = av_mallocz(sizeof(*sub->rects[rect]));
  419.         sub->rects[rect]->x    = ctx->presentation.objects[rect].x;
  420.         sub->rects[rect]->y    = ctx->presentation.objects[rect].y;
  421.         sub->rects[rect]->w    = ctx->pictures[picture_id].w;
  422.         sub->rects[rect]->h    = ctx->pictures[picture_id].h;
  423.         sub->rects[rect]->type = SUBTITLE_BITMAP;
  424.  
  425.         /* Process bitmap */
  426.         sub->rects[rect]->pict.linesize[0] = ctx->pictures[picture_id].w;
  427.         if (ctx->pictures[picture_id].rle) {
  428.             if (ctx->pictures[picture_id].rle_remaining_len)
  429.                 av_log(avctx, AV_LOG_ERROR, "RLE data length %u is %u bytes shorter than expected\n",
  430.                        ctx->pictures[picture_id].rle_data_len, ctx->pictures[picture_id].rle_remaining_len);
  431.             if (decode_rle(avctx, sub, rect, ctx->pictures[picture_id].rle, ctx->pictures[picture_id].rle_data_len) < 0)
  432.                 return 0;
  433.         }
  434.  
  435.         /* Allocate memory for colors */
  436.         sub->rects[rect]->nb_colors    = 256;
  437.         sub->rects[rect]->pict.data[1] = av_mallocz(AVPALETTE_SIZE);
  438.  
  439.         /* Copy the forced flag */
  440.         sub->rects[rect]->flags = (ctx->presentation.objects[rect].composition & 0x40) != 0 ? AV_SUBTITLE_FLAG_FORCED : 0;
  441.  
  442.         if (!ctx->forced_subs_only || ctx->presentation.objects[rect].composition & 0x40)
  443.         memcpy(sub->rects[rect]->pict.data[1], ctx->clut, sub->rects[rect]->nb_colors * sizeof(uint32_t));
  444.     }
  445.  
  446.     return 1;
  447. }
  448.  
  449. static int decode(AVCodecContext *avctx, void *data, int *data_size,
  450.                   AVPacket *avpkt)
  451. {
  452.     const uint8_t *buf = avpkt->data;
  453.     int buf_size       = avpkt->size;
  454.     AVSubtitle *sub    = data;
  455.  
  456.     const uint8_t *buf_end;
  457.     uint8_t       segment_type;
  458.     int           segment_length;
  459.     int i;
  460.  
  461.     av_dlog(avctx, "PGS sub packet:\n");
  462.  
  463.     for (i = 0; i < buf_size; i++) {
  464.         av_dlog(avctx, "%02x ", buf[i]);
  465.         if (i % 16 == 15)
  466.             av_dlog(avctx, "\n");
  467.     }
  468.  
  469.     if (i & 15)
  470.         av_dlog(avctx, "\n");
  471.  
  472.     *data_size = 0;
  473.  
  474.     /* Ensure that we have received at a least a segment code and segment length */
  475.     if (buf_size < 3)
  476.         return -1;
  477.  
  478.     buf_end = buf + buf_size;
  479.  
  480.     /* Step through buffer to identify segments */
  481.     while (buf < buf_end) {
  482.         segment_type   = bytestream_get_byte(&buf);
  483.         segment_length = bytestream_get_be16(&buf);
  484.  
  485.         av_dlog(avctx, "Segment Length %d, Segment Type %x\n", segment_length, segment_type);
  486.  
  487.         if (segment_type != DISPLAY_SEGMENT && segment_length > buf_end - buf)
  488.             break;
  489.  
  490.         switch (segment_type) {
  491.         case PALETTE_SEGMENT:
  492.             parse_palette_segment(avctx, buf, segment_length);
  493.             break;
  494.         case PICTURE_SEGMENT:
  495.             parse_picture_segment(avctx, buf, segment_length);
  496.             break;
  497.         case PRESENTATION_SEGMENT:
  498.             parse_presentation_segment(avctx, buf, segment_length, sub->pts);
  499.             break;
  500.         case WINDOW_SEGMENT:
  501.             /*
  502.              * Window Segment Structure (No new information provided):
  503.              *     2 bytes: Unknown,
  504.              *     2 bytes: X position of subtitle,
  505.              *     2 bytes: Y position of subtitle,
  506.              *     2 bytes: Width of subtitle,
  507.              *     2 bytes: Height of subtitle.
  508.              */
  509.             break;
  510.         case DISPLAY_SEGMENT:
  511.             *data_size = display_end_segment(avctx, data, buf, segment_length);
  512.             break;
  513.         default:
  514.             av_log(avctx, AV_LOG_ERROR, "Unknown subtitle segment type 0x%x, length %d\n",
  515.                    segment_type, segment_length);
  516.             break;
  517.         }
  518.  
  519.         buf += segment_length;
  520.     }
  521.  
  522.     return buf_size;
  523. }
  524.  
  525. #define OFFSET(x) offsetof(PGSSubContext, x)
  526. #define SD AV_OPT_FLAG_SUBTITLE_PARAM | AV_OPT_FLAG_DECODING_PARAM
  527. static const AVOption options[] = {
  528.     {"forced_subs_only", "Only show forced subtitles", OFFSET(forced_subs_only), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, SD},
  529.     { NULL },
  530. };
  531.  
  532. static const AVClass pgsdec_class = {
  533.     .class_name = "PGS subtitle decoder",
  534.     .item_name  = av_default_item_name,
  535.     .option     = options,
  536.     .version    = LIBAVUTIL_VERSION_INT,
  537. };
  538.  
  539. AVCodec ff_pgssub_decoder = {
  540.     .name           = "pgssub",
  541.     .long_name      = NULL_IF_CONFIG_SMALL("HDMV Presentation Graphic Stream subtitles"),
  542.     .type           = AVMEDIA_TYPE_SUBTITLE,
  543.     .id             = AV_CODEC_ID_HDMV_PGS_SUBTITLE,
  544.     .priv_data_size = sizeof(PGSSubContext),
  545.     .init           = init_decoder,
  546.     .close          = close_decoder,
  547.     .decode         = decode,
  548.     .priv_class     = &pgsdec_class,
  549. };
  550.