utvideoenc.c 19.3 KB
Newer Older
1 2 3 4
/*
 * Ut Video encoder
 * Copyright (c) 2012 Jan Ekström
 *
5
 * This file is part of FFmpeg.
6
 *
7
 * FFmpeg is free software; you can redistribute it and/or
8 9 10 11
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
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 20 21 22 23 24 25 26
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

/**
 * @file
 * Ut Video encoder
 */

27
#include "libavutil/imgutils.h"
28 29 30
#include "libavutil/intreadwrite.h"
#include "avcodec.h"
#include "internal.h"
31
#include "bswapdsp.h"
32 33
#include "bytestream.h"
#include "put_bits.h"
34
#include "huffyuvencdsp.h"
35 36
#include "mathops.h"
#include "utvideo.h"
37
#include "huffman.h"
38 39 40 41 42 43 44 45 46 47 48

/* Compare huffentry symbols */
static int huff_cmp_sym(const void *a, const void *b)
{
    const HuffEntry *aa = a, *bb = b;
    return aa->sym - bb->sym;
}

static av_cold int utvideo_encode_close(AVCodecContext *avctx)
{
    UtvideoContext *c = avctx->priv_data;
49
    int i;
50 51 52

    av_freep(&avctx->coded_frame);
    av_freep(&c->slice_bits);
53 54
    for (i = 0; i < 4; i++)
        av_freep(&c->slice_buffer[i]);
55 56 57 58 59 60 61

    return 0;
}

static av_cold int utvideo_encode_init(AVCodecContext *avctx)
{
    UtvideoContext *c = avctx->priv_data;
62
    int i, subsampled_height;
63 64 65 66
    uint32_t original_format;

    c->avctx           = avctx;
    c->frame_info_size = 4;
67
    c->slice_stride    = FFALIGN(avctx->width, 32);
68 69

    switch (avctx->pix_fmt) {
70
    case AV_PIX_FMT_RGB24:
71 72 73 74
        c->planes        = 3;
        avctx->codec_tag = MKTAG('U', 'L', 'R', 'G');
        original_format  = UTVIDEO_RGB;
        break;
75
    case AV_PIX_FMT_RGBA:
76 77 78 79
        c->planes        = 4;
        avctx->codec_tag = MKTAG('U', 'L', 'R', 'A');
        original_format  = UTVIDEO_RGBA;
        break;
80
    case AV_PIX_FMT_YUV420P:
81 82 83 84 85 86
        if (avctx->width & 1 || avctx->height & 1) {
            av_log(avctx, AV_LOG_ERROR,
                   "4:2:0 video requires even width and height.\n");
            return AVERROR_INVALIDDATA;
        }
        c->planes        = 3;
87 88 89 90
        if (avctx->colorspace == AVCOL_SPC_BT709)
            avctx->codec_tag = MKTAG('U', 'L', 'H', '0');
        else
            avctx->codec_tag = MKTAG('U', 'L', 'Y', '0');
91 92
        original_format  = UTVIDEO_420;
        break;
93
    case AV_PIX_FMT_YUV422P:
94 95 96 97 98 99
        if (avctx->width & 1) {
            av_log(avctx, AV_LOG_ERROR,
                   "4:2:2 video requires even width.\n");
            return AVERROR_INVALIDDATA;
        }
        c->planes        = 3;
100 101 102 103
        if (avctx->colorspace == AVCOL_SPC_BT709)
            avctx->codec_tag = MKTAG('U', 'L', 'H', '2');
        else
            avctx->codec_tag = MKTAG('U', 'L', 'Y', '2');
104 105 106 107 108 109 110 111
        original_format  = UTVIDEO_422;
        break;
    default:
        av_log(avctx, AV_LOG_ERROR, "Unknown pixel format: %d\n",
               avctx->pix_fmt);
        return AVERROR_INVALIDDATA;
    }

112
    ff_bswapdsp_init(&c->bdsp);
113
    ff_huffyuvencdsp_init(&c->hdsp);
114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136

    /* Check the prediction method, and error out if unsupported */
    if (avctx->prediction_method < 0 || avctx->prediction_method > 4) {
        av_log(avctx, AV_LOG_WARNING,
               "Prediction method %d is not supported in Ut Video.\n",
               avctx->prediction_method);
        return AVERROR_OPTION_NOT_FOUND;
    }

    if (avctx->prediction_method == FF_PRED_PLANE) {
        av_log(avctx, AV_LOG_ERROR,
               "Plane prediction is not supported in Ut Video.\n");
        return AVERROR_OPTION_NOT_FOUND;
    }

    /* Convert from libavcodec prediction type to Ut Video's */
    c->frame_pred = ff_ut_pred_order[avctx->prediction_method];

    if (c->frame_pred == PRED_GRADIENT) {
        av_log(avctx, AV_LOG_ERROR, "Gradient prediction is not supported.\n");
        return AVERROR_OPTION_NOT_FOUND;
    }

137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156
    /*
     * Check the asked slice count for obviously invalid
     * values (> 256 or negative).
     */
    if (avctx->slices > 256 || avctx->slices < 0) {
        av_log(avctx, AV_LOG_ERROR,
               "Slice count %d is not supported in Ut Video (theoretical range is 0-256).\n",
               avctx->slices);
        return AVERROR(EINVAL);
    }

    /* Check that the slice count is not larger than the subsampled height */
    subsampled_height = avctx->height >> av_pix_fmt_desc_get(avctx->pix_fmt)->log2_chroma_h;
    if (avctx->slices > subsampled_height) {
        av_log(avctx, AV_LOG_ERROR,
               "Slice count %d is larger than the subsampling-applied height %d.\n",
               avctx->slices, subsampled_height);
        return AVERROR(EINVAL);
    }

157
    avctx->coded_frame = av_frame_alloc();
158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176

    if (!avctx->coded_frame) {
        av_log(avctx, AV_LOG_ERROR, "Could not allocate frame.\n");
        utvideo_encode_close(avctx);
        return AVERROR(ENOMEM);
    }

    /* extradata size is 4 * 32bit */
    avctx->extradata_size = 16;

    avctx->extradata = av_mallocz(avctx->extradata_size +
                                  FF_INPUT_BUFFER_PADDING_SIZE);

    if (!avctx->extradata) {
        av_log(avctx, AV_LOG_ERROR, "Could not allocate extradata.\n");
        utvideo_encode_close(avctx);
        return AVERROR(ENOMEM);
    }

177
    for (i = 0; i < c->planes; i++) {
178
        c->slice_buffer[i] = av_malloc(c->slice_stride * (avctx->height + 2) +
179 180 181 182 183 184
                                       FF_INPUT_BUFFER_PADDING_SIZE);
        if (!c->slice_buffer[i]) {
            av_log(avctx, AV_LOG_ERROR, "Cannot allocate temporary buffer 1.\n");
            utvideo_encode_close(avctx);
            return AVERROR(ENOMEM);
        }
185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205
    }

    /*
     * Set the version of the encoder.
     * Last byte is "implementation ID", which is
     * obtained from the creator of the format.
     * Libavcodec has been assigned with the ID 0xF0.
     */
    AV_WB32(avctx->extradata, MKTAG(1, 0, 0, 0xF0));

    /*
     * Set the "original format"
     * Not used for anything during decoding.
     */
    AV_WL32(avctx->extradata + 4, original_format);

    /* Write 4 as the 'frame info size' */
    AV_WL32(avctx->extradata + 8, c->frame_info_size);

    /*
     * Set how many slices are going to be used.
206 207
     * By default uses multiple slices depending on the subsampled height.
     * This enables multithreading in the official decoder.
208
     */
209 210 211
    if (!avctx->slices) {
        c->slices = subsampled_height / 120;

212
        if (!c->slices)
213
            c->slices = 1;
214
        else if (c->slices > 256)
215 216 217 218
            c->slices = 256;
    } else {
        c->slices = avctx->slices;
    }
219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238

    /* Set compression mode */
    c->compression = COMP_HUFF;

    /*
     * Set the encoding flags:
     * - Slice count minus 1
     * - Interlaced encoding mode flag, set to zero for now.
     * - Compression mode (none/huff)
     * And write the flags.
     */
    c->flags  = (c->slices - 1) << 24;
    c->flags |= 0 << 11; // bit field to signal interlaced encoding mode
    c->flags |= c->compression;

    AV_WL32(avctx->extradata + 12, c->flags);

    return 0;
}

239 240
static void mangle_rgb_planes(uint8_t *dst[4], int dst_stride, uint8_t *src,
                              int step, int stride, int width, int height)
241 242
{
    int i, j;
243
    int k = 2 * dst_stride;
244
    unsigned int g;
245 246

    for (j = 0; j < height; j++) {
247 248 249 250
        if (step == 3) {
            for (i = 0; i < width * step; i += step) {
                g         = src[i + 1];
                dst[0][k] = g;
251
                g        += 0x80;
252 253 254 255 256 257 258 259
                dst[1][k] = src[i + 2] - g;
                dst[2][k] = src[i + 0] - g;
                k++;
            }
        } else {
            for (i = 0; i < width * step; i += step) {
                g         = src[i + 1];
                dst[0][k] = g;
260
                g        += 0x80;
261 262 263 264 265
                dst[1][k] = src[i + 2] - g;
                dst[2][k] = src[i + 0] - g;
                dst[3][k] = src[i + 3];
                k++;
            }
266
        }
267
        k += dst_stride - width;
268 269 270 271 272
        src += stride;
    }
}

/* Write data to a plane with left prediction */
273
static void left_predict(uint8_t *src, uint8_t *dst, int stride,
274 275 276 277 278 279 280
                         int width, int height)
{
    int i, j;
    uint8_t prev;

    prev = 0x80; /* Set the initial value */
    for (j = 0; j < height; j++) {
281
        for (i = 0; i < width; i++) {
282 283 284 285 286 287 288 289
            *dst++ = src[i] - prev;
            prev   = src[i];
        }
        src += stride;
    }
}

/* Write data to a plane with median prediction */
290
static void median_predict(UtvideoContext *c, uint8_t *src, uint8_t *dst, int stride,
291 292 293
                           int width, int height)
{
    int i, j;
294
    int A, B;
295 296 297 298
    uint8_t prev;

    /* First line uses left neighbour prediction */
    prev = 0x80; /* Set the initial value */
299
    for (i = 0; i < width; i++) {
300 301 302 303 304 305 306 307 308 309 310 311 312
        *dst++ = src[i] - prev;
        prev   = src[i];
    }

    if (height == 1)
        return;

    src += stride;

    /*
     * Second line uses top prediction for the first sample,
     * and median for the rest.
     */
313
    A = B = 0;
314 315

    /* Rest of the coded part uses median prediction */
316
    for (j = 1; j < height; j++) {
317
        c->hdsp.sub_hfyu_median_pred(dst, src - stride, src, width, &A, &B);
318
        dst += width;
319 320 321 322 323 324
        src += stride;
    }
}

/* Count the usage of values in a plane */
static void count_usage(uint8_t *src, int width,
325
                        int height, uint64_t *counts)
326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391
{
    int i, j;

    for (j = 0; j < height; j++) {
        for (i = 0; i < width; i++) {
            counts[src[i]]++;
        }
        src += width;
    }
}

/* Calculate the actual huffman codes from the code lengths */
static void calculate_codes(HuffEntry *he)
{
    int last, i;
    uint32_t code;

    qsort(he, 256, sizeof(*he), ff_ut_huff_cmp_len);

    last = 255;
    while (he[last].len == 255 && last)
        last--;

    code = 1;
    for (i = last; i >= 0; i--) {
        he[i].code  = code >> (32 - he[i].len);
        code       += 0x80000000u >> (he[i].len - 1);
    }

    qsort(he, 256, sizeof(*he), huff_cmp_sym);
}

/* Write huffman bit codes to a memory block */
static int write_huff_codes(uint8_t *src, uint8_t *dst, int dst_size,
                            int width, int height, HuffEntry *he)
{
    PutBitContext pb;
    int i, j;
    int count;

    init_put_bits(&pb, dst, dst_size);

    /* Write the codes */
    for (j = 0; j < height; j++) {
        for (i = 0; i < width; i++)
            put_bits(&pb, he[src[i]].len, he[src[i]].code);

        src += width;
    }

    /* Pad output to a 32bit boundary */
    count = put_bits_count(&pb) & 0x1F;

    if (count)
        put_bits(&pb, 32 - count, 0);

    /* Get the amount of bits written */
    count = put_bits_count(&pb);

    /* Flush the rest with zeroes */
    flush_put_bits(&pb);

    return count;
}

static int encode_plane(AVCodecContext *avctx, uint8_t *src,
392
                        uint8_t *dst, int stride, int plane_no,
393 394 395 396
                        int width, int height, PutByteContext *pb)
{
    UtvideoContext *c        = avctx->priv_data;
    uint8_t  lengths[256];
397
    uint64_t counts[256]     = { 0 };
398 399 400 401

    HuffEntry he[256];

    uint32_t offset = 0, slice_len = 0;
402
    const int cmask = ~(!plane_no && avctx->pix_fmt == AV_PIX_FMT_YUV420P);
403 404
    int      i, sstart, send = 0;
    int      symbol;
405
    int      ret;
406 407 408 409 410 411

    /* Do prediction / make planes */
    switch (c->frame_pred) {
    case PRED_NONE:
        for (i = 0; i < c->slices; i++) {
            sstart = send;
412
            send   = height * (i + 1) / c->slices & cmask;
413 414 415
            av_image_copy_plane(dst + sstart * width, width,
                                src + sstart * stride, stride,
                                width, send - sstart);
416 417 418 419 420
        }
        break;
    case PRED_LEFT:
        for (i = 0; i < c->slices; i++) {
            sstart = send;
421
            send   = height * (i + 1) / c->slices & cmask;
422
            left_predict(src + sstart * stride, dst + sstart * width,
423
                         stride, width, send - sstart);
424 425 426 427 428
        }
        break;
    case PRED_MEDIAN:
        for (i = 0; i < c->slices; i++) {
            sstart = send;
429
            send   = height * (i + 1) / c->slices & cmask;
430
            median_predict(c, src + sstart * stride, dst + sstart * width,
431
                           stride, width, send - sstart);
432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447
        }
        break;
    default:
        av_log(avctx, AV_LOG_ERROR, "Unknown prediction mode: %d\n",
               c->frame_pred);
        return AVERROR_OPTION_NOT_FOUND;
    }

    /* Count the usage of values */
    count_usage(dst, width, height, counts);

    /* Check for a special case where only one symbol was used */
    for (symbol = 0; symbol < 256; symbol++) {
        /* If non-zero count is found, see if it matches width * height */
        if (counts[symbol]) {
            /* Special case if only one symbol was used */
448
            if (counts[symbol] == width * (int64_t)height) {
449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471
                /*
                 * Write a zero for the single symbol
                 * used in the plane, else 0xFF.
                 */
                for (i = 0; i < 256; i++) {
                    if (i == symbol)
                        bytestream2_put_byte(pb, 0);
                    else
                        bytestream2_put_byte(pb, 0xFF);
                }

                /* Write zeroes for lengths */
                for (i = 0; i < c->slices; i++)
                    bytestream2_put_le32(pb, 0);

                /* And that's all for that plane folks */
                return 0;
            }
            break;
        }
    }

    /* Calculate huffman lengths */
472
    if ((ret = ff_huff_gen_len_table(lengths, counts, 256, 1)) < 0)
473
        return ret;
474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492

    /*
     * Write the plane's header into the output packet:
     * - huffman code lengths (256 bytes)
     * - slice end offsets (gotten from the slice lengths)
     */
    for (i = 0; i < 256; i++) {
        bytestream2_put_byte(pb, lengths[i]);

        he[i].len = lengths[i];
        he[i].sym = i;
    }

    /* Calculate the huffman codes themselves */
    calculate_codes(he);

    send = 0;
    for (i = 0; i < c->slices; i++) {
        sstart  = send;
493
        send    = height * (i + 1) / c->slices & cmask;
494 495 496 497 498 499

        /*
         * Write the huffman codes to a buffer,
         * get the offset in bits and convert to bytes.
         */
        offset += write_huff_codes(dst + sstart * width, c->slice_bits,
500
                                   width * height + 4, width,
501 502 503 504 505
                                   send - sstart, he) >> 3;

        slice_len = offset - slice_len;

        /* Byteswap the written huffman codes */
506 507 508
        c->bdsp.bswap_buf((uint32_t *) c->slice_bits,
                          (uint32_t *) c->slice_bits,
                          slice_len >> 2);
509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546

        /* Write the offset to the stream */
        bytestream2_put_le32(pb, offset);

        /* Seek to the data part of the packet */
        bytestream2_seek_p(pb, 4 * (c->slices - i - 1) +
                           offset - slice_len, SEEK_CUR);

        /* Write the slices' data into the output packet */
        bytestream2_put_buffer(pb, c->slice_bits, slice_len);

        /* Seek back to the slice offsets */
        bytestream2_seek_p(pb, -4 * (c->slices - i - 1) - offset,
                           SEEK_CUR);

        slice_len = offset;
    }

    /* And at the end seek to the end of written slice(s) */
    bytestream2_seek_p(pb, offset, SEEK_CUR);

    return 0;
}

static int utvideo_encode_frame(AVCodecContext *avctx, AVPacket *pkt,
                                const AVFrame *pic, int *got_packet)
{
    UtvideoContext *c = avctx->priv_data;
    PutByteContext pb;

    uint32_t frame_info;

    uint8_t *dst;

    int width = avctx->width, height = avctx->height;
    int i, ret = 0;

    /* Allocate a new packet if needed, and set it to the pointer dst */
547 548
    ret = ff_alloc_packet2(avctx, pkt, (256 + 4 * c->slices + width * height) *
                           c->planes + 4);
549

550
    if (ret < 0)
551 552 553 554 555 556
        return ret;

    dst = pkt->data;

    bytestream2_init_writer(&pb, dst, pkt->size);

557
    av_fast_padded_malloc(&c->slice_bits, &c->slice_bits_size, width * height + 4);
558 559 560 561 562 563 564

    if (!c->slice_bits) {
        av_log(avctx, AV_LOG_ERROR, "Cannot allocate temporary buffer 2.\n");
        return AVERROR(ENOMEM);
    }

    /* In case of RGB, mangle the planes to Ut Video's format */
565
    if (avctx->pix_fmt == AV_PIX_FMT_RGBA || avctx->pix_fmt == AV_PIX_FMT_RGB24)
566 567
        mangle_rgb_planes(c->slice_buffer, c->slice_stride, pic->data[0],
                          c->planes, pic->linesize[0], width, height);
568 569 570

    /* Deal with the planes */
    switch (avctx->pix_fmt) {
571 572
    case AV_PIX_FMT_RGB24:
    case AV_PIX_FMT_RGBA:
573
        for (i = 0; i < c->planes; i++) {
574
            ret = encode_plane(avctx, c->slice_buffer[i] + 2 * c->slice_stride,
575
                               c->slice_buffer[i], c->slice_stride, i,
576 577 578 579 580 581 582 583
                               width, height, &pb);

            if (ret) {
                av_log(avctx, AV_LOG_ERROR, "Error encoding plane %d.\n", i);
                return ret;
            }
        }
        break;
584
    case AV_PIX_FMT_YUV422P:
585
        for (i = 0; i < c->planes; i++) {
586
            ret = encode_plane(avctx, pic->data[i], c->slice_buffer[0],
587
                               pic->linesize[i], i, width >> !!i, height, &pb);
588 589 590 591 592 593 594

            if (ret) {
                av_log(avctx, AV_LOG_ERROR, "Error encoding plane %d.\n", i);
                return ret;
            }
        }
        break;
595
    case AV_PIX_FMT_YUV420P:
596
        for (i = 0; i < c->planes; i++) {
597
            ret = encode_plane(avctx, pic->data[i], c->slice_buffer[0],
598
                               pic->linesize[i], i, width >> !!i, height >> !!i,
599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638
                               &pb);

            if (ret) {
                av_log(avctx, AV_LOG_ERROR, "Error encoding plane %d.\n", i);
                return ret;
            }
        }
        break;
    default:
        av_log(avctx, AV_LOG_ERROR, "Unknown pixel format: %d\n",
               avctx->pix_fmt);
        return AVERROR_INVALIDDATA;
    }

    /*
     * Write frame information (LE 32bit unsigned)
     * into the output packet.
     * Contains the prediction method.
     */
    frame_info = c->frame_pred << 8;
    bytestream2_put_le32(&pb, frame_info);

    /*
     * At least currently Ut Video is IDR only.
     * Set flags accordingly.
     */
    avctx->coded_frame->key_frame = 1;
    avctx->coded_frame->pict_type = AV_PICTURE_TYPE_I;

    pkt->size   = bytestream2_tell_p(&pb);
    pkt->flags |= AV_PKT_FLAG_KEY;

    /* Packet should be done */
    *got_packet = 1;

    return 0;
}

AVCodec ff_utvideo_encoder = {
    .name           = "utvideo",
639
    .long_name      = NULL_IF_CONFIG_SMALL("Ut Video"),
640
    .type           = AVMEDIA_TYPE_VIDEO,
641
    .id             = AV_CODEC_ID_UTVIDEO,
642 643 644 645
    .priv_data_size = sizeof(UtvideoContext),
    .init           = utvideo_encode_init,
    .encode2        = utvideo_encode_frame,
    .close          = utvideo_encode_close,
646
    .capabilities   = CODEC_CAP_FRAME_THREADS | CODEC_CAP_INTRA_ONLY,
647 648 649
    .pix_fmts       = (const enum AVPixelFormat[]) {
                          AV_PIX_FMT_RGB24, AV_PIX_FMT_RGBA, AV_PIX_FMT_YUV422P,
                          AV_PIX_FMT_YUV420P, AV_PIX_FMT_NONE
650 651
                      },
};