dvdsubenc.c 15.1 KB
Newer Older
1
/*
2
 * DVD subtitle encoding
3
 * Copyright (c) 2005 Wolfram Gloger
4
 *
5 6 7
 * This file is part of FFmpeg.
 *
 * FFmpeg is free software; you can redistribute it and/or
8 9
 * modify it under the terms of the GNU Lesser General Public
 * 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 14 15 16 17
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * 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
 */
#include "avcodec.h"
22
#include "bytestream.h"
23
#include "internal.h"
24
#include "libavutil/avassert.h"
25 26
#include "libavutil/bprint.h"
#include "libavutil/imgutils.h"
27
#include "libavutil/opt.h"
28

29
typedef struct {
30
    AVClass *class;
31
    uint32_t global_palette[16];
32
    int even_rows_fix;
33
} DVDSubtitleContext;
34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63

// ncnt is the nibble counter
#define PUTNIBBLE(val)\
do {\
    if (ncnt++ & 1)\
        *q++ = bitbuf | ((val) & 0x0f);\
    else\
        bitbuf = (val) << 4;\
} while(0)

static void dvd_encode_rle(uint8_t **pq,
                           const uint8_t *bitmap, int linesize,
                           int w, int h,
                           const int cmap[256])
{
    uint8_t *q;
    unsigned int bitbuf = 0;
    int ncnt;
    int x, y, len, color;

    q = *pq;

    for (y = 0; y < h; ++y) {
        ncnt = 0;
        for(x = 0; x < w; x += len) {
            color = bitmap[x];
            for (len=1; x+len < w; ++len)
                if (bitmap[x+len] != color)
                    break;
            color = cmap[color];
64
            av_assert0(color < 4);
65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96
            if (len < 0x04) {
                PUTNIBBLE((len << 2)|color);
            } else if (len < 0x10) {
                PUTNIBBLE(len >> 2);
                PUTNIBBLE((len << 2)|color);
            } else if (len < 0x40) {
                PUTNIBBLE(0);
                PUTNIBBLE(len >> 2);
                PUTNIBBLE((len << 2)|color);
            } else if (x+len == w) {
                PUTNIBBLE(0);
                PUTNIBBLE(0);
                PUTNIBBLE(0);
                PUTNIBBLE(color);
            } else {
                if (len > 0xff)
                    len = 0xff;
                PUTNIBBLE(0);
                PUTNIBBLE(len >> 6);
                PUTNIBBLE(len >> 2);
                PUTNIBBLE((len << 2)|color);
            }
        }
        /* end of line */
        if (ncnt & 1)
            PUTNIBBLE(0);
        bitmap += linesize;
    }

    *pq = q;
}

97 98 99
static int color_distance(uint32_t a, uint32_t b)
{
    int r = 0, d, i;
100
    int alpha_a = 8, alpha_b = 8;
101

102 103 104
    for (i = 24; i >= 0; i -= 8) {
        d = alpha_a * (int)((a >> i) & 0xFF) -
            alpha_b * (int)((b >> i) & 0xFF);
105
        r += d * d;
106 107
        alpha_a = a >> 28;
        alpha_b = b >> 28;
108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139
    }
    return r;
}

/**
 * Count colors used in a rectangle, quantizing alpha and grouping by
 * nearest global palette entry.
 */
static void count_colors(AVCodecContext *avctx, unsigned hits[33],
                         const AVSubtitleRect *r)
{
    DVDSubtitleContext *dvdc = avctx->priv_data;
    unsigned count[256] = { 0 };
    uint32_t *palette = (uint32_t *)r->pict.data[1];
    uint32_t color;
    int x, y, i, j, match, d, best_d, av_uninit(best_j);
    uint8_t *p = r->pict.data[0];

    for (y = 0; y < r->h; y++) {
        for (x = 0; x < r->w; x++)
            count[*(p++)]++;
        p += r->pict.linesize[0] - r->w;
    }
    for (i = 0; i < 256; i++) {
        if (!count[i]) /* avoid useless search */
            continue;
        color = palette[i];
        /* 0: transparent, 1-16: semi-transparent, 17-33 opaque */
        match = color < 0x33000000 ? 0 : color < 0xCC000000 ? 1 : 17;
        if (match) {
            best_d = INT_MAX;
            for (j = 0; j < 16; j++) {
140 141
                d = color_distance(0xFF000000 | color,
                                   0xFF000000 | dvdc->global_palette[j]);
142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212
                if (d < best_d) {
                    best_d = d;
                    best_j = j;
                }
            }
            match += best_j;
        }
        hits[match] += count[i];
    }
}

static void select_palette(AVCodecContext *avctx, int out_palette[4],
                           int out_alpha[4], unsigned hits[33])
{
    DVDSubtitleContext *dvdc = avctx->priv_data;
    int i, j, bright, mult;
    uint32_t color;
    int selected[4] = { 0 };
    uint32_t pseudopal[33] = { 0 };
    uint32_t refcolor[3] = { 0x00000000, 0xFFFFFFFF, 0xFF000000 };

    /* Bonus for transparent: if the rectangle fits tightly the text, the
       background color can be quite rare, but it would be ugly without it */
    hits[0] *= 16;
    /* Bonus for bright colors */
    for (i = 0; i < 16; i++) {
        if (!(hits[1 + i] + hits[17 + i]))
            continue; /* skip unused colors to gain time */
        color = dvdc->global_palette[i];
        bright = 0;
        for (j = 0; j < 3; j++, color >>= 8)
            bright += (color & 0xFF) < 0x40 || (color & 0xFF) >= 0xC0;
        mult = 2 + FFMIN(bright, 2);
        hits[ 1 + i] *= mult;
        hits[17 + i] *= mult;
    }

    /* Select four most frequent colors */
    for (i = 0; i < 4; i++) {
        for (j = 0; j < 33; j++)
            if (hits[j] > hits[selected[i]])
                selected[i] = j;
        hits[selected[i]] = 0;
    }

    /* Order the colors like in most DVDs:
       0: background, 1: foreground, 2: outline */
    for (i = 0; i < 16; i++) {
        pseudopal[ 1 + i] = 0x80000000 | dvdc->global_palette[i];
        pseudopal[17 + i] = 0xFF000000 | dvdc->global_palette[i];
    }
    for (i = 0; i < 3; i++) {
        int best_d = color_distance(refcolor[i], pseudopal[selected[i]]);
        for (j = i + 1; j < 4; j++) {
            int d = color_distance(refcolor[i], pseudopal[selected[j]]);
            if (d < best_d) {
                FFSWAP(int, selected[i], selected[j]);
                best_d = d;
            }
        }
    }

    /* Output */
    for (i = 0; i < 4; i++) {
        out_palette[i] = selected[i] ? (selected[i] - 1) & 0xF : 0;
        out_alpha  [i] = !selected[i] ? 0 : selected[i] < 17 ? 0x80 : 0xFF;
    }
}

static void build_color_map(AVCodecContext *avctx, int cmap[],
                            const uint32_t palette[],
213
                            const int out_palette[], unsigned int const out_alpha[])
214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251
{
    DVDSubtitleContext *dvdc = avctx->priv_data;
    int i, j, d, best_d;
    uint32_t pseudopal[4];

    for (i = 0; i < 4; i++)
        pseudopal[i] = (out_alpha[i] << 24) |
                       dvdc->global_palette[out_palette[i]];
    for (i = 0; i < 256; i++) {
        best_d = INT_MAX;
        for (j = 0; j < 4; j++) {
            d = color_distance(pseudopal[j], palette[i]);
            if (d < best_d) {
                cmap[i] = j;
                best_d = d;
            }
        }
    }
}

static void copy_rectangle(AVSubtitleRect *dst, AVSubtitleRect *src, int cmap[])
{
    int x, y;
    uint8_t *p, *q;

    p = src->pict.data[0];
    q = dst->pict.data[0] + (src->x - dst->x) +
                            (src->y - dst->y) * dst->pict.linesize[0];
    for (y = 0; y < src->h; y++) {
        for (x = 0; x < src->w; x++)
            *(q++) = cmap[*(p++)];
        p += src->pict.linesize[0] - src->w;
        q += dst->pict.linesize[0] - src->w;
    }
}

static int encode_dvd_subtitles(AVCodecContext *avctx,
                                uint8_t *outbuf, int outbuf_size,
252 253
                                const AVSubtitle *h)
{
254
    DVDSubtitleContext *dvdc = avctx->priv_data;
255
    uint8_t *q, *qq;
256 257 258 259 260 261 262 263 264
    int offset1, offset2;
    int i, rects = h->num_rects, ret;
    unsigned global_palette_hits[33] = { 0 };
    int cmap[256];
    int out_palette[4];
    int out_alpha[4];
    AVSubtitleRect vrect;
    uint8_t *vrect_data = NULL;
    int x2, y2;
265
    int forced = 0;
266

267
    if (rects == 0 || !h->rects)
268
        return AVERROR(EINVAL);
269 270 271 272 273
    for (i = 0; i < rects; i++)
        if (h->rects[i]->type != SUBTITLE_BITMAP) {
            av_log(avctx, AV_LOG_ERROR, "Bitmap subtitle required\n");
            return AVERROR(EINVAL);
        }
274 275 276 277 278 279
    /* Mark this subtitle forced if any of the rectangles is forced. */
    for (i = 0; i < rects; i++)
        if ((h->rects[i]->flags & AV_SUBTITLE_FLAG_FORCED) != 0) {
            forced = 1;
            break;
        }
280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305
    vrect = *h->rects[0];

    if (rects > 1) {
        /* DVD subtitles can have only one rectangle: build a virtual
           rectangle containing all actual rectangles.
           The data of the rectangles will be copied later, when the palette
           is decided, because the rectangles may have different palettes. */
        int xmin = h->rects[0]->x, xmax = xmin + h->rects[0]->w;
        int ymin = h->rects[0]->y, ymax = ymin + h->rects[0]->h;
        for (i = 1; i < rects; i++) {
            xmin = FFMIN(xmin, h->rects[i]->x);
            ymin = FFMIN(ymin, h->rects[i]->y);
            xmax = FFMAX(xmax, h->rects[i]->x + h->rects[i]->w);
            ymax = FFMAX(ymax, h->rects[i]->y + h->rects[i]->h);
        }
        vrect.x = xmin;
        vrect.y = ymin;
        vrect.w = xmax - xmin;
        vrect.h = ymax - ymin;
        if ((ret = av_image_check_size(vrect.w, vrect.h, 0, avctx)) < 0)
            return ret;

        /* Count pixels outside the virtual rectangle as transparent */
        global_palette_hits[0] = vrect.w * vrect.h;
        for (i = 0; i < rects; i++)
            global_palette_hits[0] -= h->rects[i]->w * h->rects[i]->h;
306
    }
307 308 309 310 311 312 313 314 315 316 317 318 319 320

    for (i = 0; i < rects; i++)
        count_colors(avctx, global_palette_hits, h->rects[i]);
    select_palette(avctx, out_palette, out_alpha, global_palette_hits);

    if (rects > 1) {
        if (!(vrect_data = av_calloc(vrect.w, vrect.h)))
            return AVERROR(ENOMEM);
        vrect.pict.data    [0] = vrect_data;
        vrect.pict.linesize[0] = vrect.w;
        for (i = 0; i < rects; i++) {
            build_color_map(avctx, cmap, (uint32_t *)h->rects[i]->pict.data[1],
                            out_palette, out_alpha);
            copy_rectangle(&vrect, h->rects[i], cmap);
321
        }
322 323 324 325 326
        for (i = 0; i < 4; i++)
            cmap[i] = i;
    } else {
        build_color_map(avctx, cmap, (uint32_t *)h->rects[0]->pict.data[1],
                        out_palette, out_alpha);
327 328
    }

329 330 331 332 333 334
    av_log(avctx, AV_LOG_DEBUG, "Selected palette:");
    for (i = 0; i < 4; i++)
        av_log(avctx, AV_LOG_DEBUG, " 0x%06x@@%02x (0x%x,0x%x)",
               dvdc->global_palette[out_palette[i]], out_alpha[i],
               out_palette[i], out_alpha[i] >> 4);
    av_log(avctx, AV_LOG_DEBUG, "\n");
335 336 337

    // encode data block
    q = outbuf + 4;
338 339 340 341 342 343 344 345 346 347 348 349
    offset1 = q - outbuf;
    // worst case memory requirement: 1 nibble per pixel..
    if ((q - outbuf) + vrect.w * vrect.h / 2 + 17 + 21 > outbuf_size) {
        av_log(NULL, AV_LOG_ERROR, "dvd_subtitle too big\n");
        ret = AVERROR_BUFFER_TOO_SMALL;
        goto fail;
    }
    dvd_encode_rle(&q, vrect.pict.data[0], vrect.w * 2,
                   vrect.w, (vrect.h + 1) >> 1, cmap);
    offset2 = q - outbuf;
    dvd_encode_rle(&q, vrect.pict.data[0] + vrect.w, vrect.w * 2,
                   vrect.w, vrect.h >> 1, cmap);
350

351 352 353 354 355 356 357
    if (dvdc->even_rows_fix && (vrect.h & 1)) {
        // Work-around for some players that want the height to be even.
        vrect.h++;
        *q++ = 0x00; // 0x00 0x00 == empty row, i.e. fully transparent
        *q++ = 0x00;
    }

358 359
    // set data packet size
    qq = outbuf + 2;
360
    bytestream_put_be16(&qq, q - outbuf);
361 362

    // send start display command
363
    bytestream_put_be16(&q, (h->start_display_time*90) >> 10);
364
    bytestream_put_be16(&q, (q - outbuf) /*- 2 */ + 8 + 12 + 2);
365
    *q++ = 0x03; // palette - 4 nibbles
366 367
    *q++ = (out_palette[3] << 4) | out_palette[2];
    *q++ = (out_palette[1] << 4) | out_palette[0];
368
    *q++ = 0x04; // alpha - 4 nibbles
369 370
    *q++ = (out_alpha[3] & 0xF0) | (out_alpha[2] >> 4);
    *q++ = (out_alpha[1] & 0xF0) | (out_alpha[0] >> 4);
371 372

    // 12 bytes per rect
373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389
    x2 = vrect.x + vrect.w - 1;
    y2 = vrect.y + vrect.h - 1;

    *q++ = 0x05;
    // x1 x2 -> 6 nibbles
    *q++ = vrect.x >> 4;
    *q++ = (vrect.x << 4) | ((x2 >> 8) & 0xf);
    *q++ = x2;
    // y1 y2 -> 6 nibbles
    *q++ = vrect.y >> 4;
    *q++ = (vrect.y << 4) | ((y2 >> 8) & 0xf);
    *q++ = y2;

    *q++ = 0x06;
    // offset1, offset2
    bytestream_put_be16(&q, offset1);
    bytestream_put_be16(&q, offset2);
390

391
    *q++ = forced ? 0x00 : 0x01; // start command
392 393 394
    *q++ = 0xff; // terminating command

    // send stop display command last
395 396
    bytestream_put_be16(&q, (h->end_display_time*90) >> 10);
    bytestream_put_be16(&q, (q - outbuf) - 2 /*+ 4*/);
397 398 399 400
    *q++ = 0x02; // set end
    *q++ = 0xff; // terminating command

    qq = outbuf;
401
    bytestream_put_be16(&qq, q - outbuf);
402

403
    av_log(NULL, AV_LOG_DEBUG, "subtitle_packet size=%"PTRDIFF_SPECIFIER"\n", q - outbuf);
404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426
    ret = q - outbuf;

fail:
    av_free(vrect_data);
    return ret;
}

static int dvdsub_init(AVCodecContext *avctx)
{
    DVDSubtitleContext *dvdc = avctx->priv_data;
    static const uint32_t default_palette[16] = {
        0x000000, 0x0000FF, 0x00FF00, 0xFF0000,
        0xFFFF00, 0xFF00FF, 0x00FFFF, 0xFFFFFF,
        0x808000, 0x8080FF, 0x800080, 0x80FF80,
        0x008080, 0xFF8080, 0x555555, 0xAAAAAA,
    };
    AVBPrint extradata;
    int i, ret;

    av_assert0(sizeof(dvdc->global_palette) == sizeof(default_palette));
    memcpy(dvdc->global_palette, default_palette, sizeof(dvdc->global_palette));

    av_bprint_init(&extradata, 0, 1);
427 428
    if (avctx->width && avctx->height)
        av_bprintf(&extradata, "size: %dx%d\n", avctx->width, avctx->height);
429 430 431 432 433
    av_bprintf(&extradata, "palette:");
    for (i = 0; i < 16; i++)
        av_bprintf(&extradata, " %06"PRIx32"%c",
                   dvdc->global_palette[i] & 0xFFFFFF, i < 15 ? ',' : '\n');

434
    ret = avpriv_bprint_to_extradata(avctx, &extradata);
435
    if (ret < 0)
436 437 438
        return ret;

    return 0;
439 440 441
}

static int dvdsub_encode(AVCodecContext *avctx,
442 443
                         unsigned char *buf, int buf_size,
                         const AVSubtitle *sub)
444 445 446 447
{
    //DVDSubtitleContext *s = avctx->priv_data;
    int ret;

448
    ret = encode_dvd_subtitles(avctx, buf, buf_size, sub);
449 450 451
    return ret;
}

452 453 454 455 456 457 458 459 460 461 462 463 464 465
#define OFFSET(x) offsetof(DVDSubtitleContext, x)
#define SE AV_OPT_FLAG_SUBTITLE_PARAM | AV_OPT_FLAG_ENCODING_PARAM
static const AVOption options[] = {
    {"even_rows_fix", "Make number of rows even (workaround for some players)", OFFSET(even_rows_fix), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, SE},
    { NULL },
};

static const AVClass dvdsubenc_class = {
    .class_name = "VOBSUB subtitle encoder",
    .item_name  = av_default_item_name,
    .option     = options,
    .version    = LIBAVUTIL_VERSION_INT,
};

466
AVCodec ff_dvdsub_encoder = {
467
    .name           = "dvdsub",
468
    .long_name      = NULL_IF_CONFIG_SMALL("DVD subtitles"),
469
    .type           = AVMEDIA_TYPE_SUBTITLE,
470
    .id             = AV_CODEC_ID_DVD_SUBTITLE,
471
    .init           = dvdsub_init,
472
    .encode_sub     = dvdsub_encode,
473
    .priv_class     = &dvdsubenc_class,
474
    .priv_data_size = sizeof(DVDSubtitleContext),
475
};