adpcm.c

ffmpeg的完整源代码和作者自己写的文档。不但有在Linux的工程哦

/*
 * ADPCM codecs
 * Copyright (c) 2001-2003 The ffmpeg Project
 *
 * This file is part of FFmpeg.
 *
 * FFmpeg is free software; you can redistribute it and/or
 * 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.
 *
 * FFmpeg is distributed in the hope that it will be useful,
 * 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
 * License along with FFmpeg; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */
#include "avcodec.h"
#include "bitstream.h"
#include "bytestream.h"

/**
 * @file adpcm.c
 * ADPCM codecs.
 * First version by Francois Revol ([e-mail])
 * Fringe ADPCM codecs (e.g., DK3, DK4, Westwood)
 *   by Mike Melanson ([e-mail])
 * CD-ROM XA ADPCM codec by BERO
 * EA ADPCM decoder by Robin Kay ([e-mail])
 * THP ADPCM decoder by Marco Gerards ([e-mail])
 *
 * Features and limitations:
 *
 * Reference documents:
 * http://www.pcisys.net/~melanson/codecs/simpleaudio.html
 * http://www.geocities.com/SiliconValley/8682/aud3.txt
 * http://openquicktime.sourceforge.net/plugins.htm
 * XAnim sources (xa_codec.c) http://www.rasnaimaging.com/people/lapus/download.html
 * http://www.cs.ucla.edu/~leec/mediabench/applications.html
 * SoX source code http://home.sprynet.com/~cbagwell/sox.html
 *
 * CD-ROM XA:
 * http://ku-www.ss.titech.ac.jp/~yatsushi/xaadpcm.html
 * vagpack & depack http://homepages.compuserve.de/bITmASTER32/psx-index.html
 * readstr http://www.geocities.co.jp/Playtown/2004/
 */

#define BLKSIZE 1024

/* step_table[] and index_table[] are from the ADPCM reference source */
/* This is the index table: */
static const int index_table[16] = {
    -1, -1, -1, -1, 2, 4, 6, 8,
    -1, -1, -1, -1, 2, 4, 6, 8,
};

/**
 * This is the step table. Note that many programs use slight deviations from
 * this table, but such deviations are negligible:
 */
static const int step_table[89] = {
    7, 8, 9, 10, 11, 12, 13, 14, 16, 17,
    19, 21, 23, 25, 28, 31, 34, 37, 41, 45,
    50, 55, 60, 66, 73, 80, 88, 97, 107, 118,
    130, 143, 157, 173, 190, 209, 230, 253, 279, 307,
    337, 371, 408, 449, 494, 544, 598, 658, 724, 796,
    876, 963, 1060, 1166, 1282, 1411, 1552, 1707, 1878, 2066,
    2272, 2499, 2749, 3024, 3327, 3660, 4026, 4428, 4871, 5358,
    5894, 6484, 7132, 7845, 8630, 9493, 10442, 11487, 12635, 13899,
    15289, 16818, 18500, 20350, 22385, 24623, 27086, 29794, 32767
};

/* These are for MS-ADPCM */
/* AdaptationTable[], AdaptCoeff1[], and AdaptCoeff2[] are from libsndfile */
static const int AdaptationTable[] = {
        230, 230, 230, 230, 307, 409, 512, 614,
        768, 614, 512, 409, 307, 230, 230, 230
};

static const int AdaptCoeff1[] = {
        256, 512, 0, 192, 240, 460, 392
};

static const int AdaptCoeff2[] = {
        0, -256, 0, 64, 0, -208, -232
};

/* These are for CD-ROM XA ADPCM */
static const int xa_adpcm_table[5][2] = {
   {   0,   0 },
   {  60,   0 },
   { 115, -52 },
   {  98, -55 },
   { 122, -60 }
};

static const int ea_adpcm_table[] = {
    0, 240, 460, 392, 0, 0, -208, -220, 0, 1,
    3, 4, 7, 8, 10, 11, 0, -1, -3, -4
};

static const int ct_adpcm_table[8] = {
    0x00E6, 0x00E6, 0x00E6, 0x00E6,
    0x0133, 0x0199, 0x0200, 0x0266
};

// padded to zero where table size is less then 16
static const int swf_index_tables[4][16] = {
    /*2*/ { -1, 2 },
    /*3*/ { -1, -1, 2, 4 },
    /*4*/ { -1, -1, -1, -1, 2, 4, 6, 8 },
    /*5*/ { -1, -1, -1, -1, -1, -1, -1, -1, 1, 2, 4, 6, 8, 10, 13, 16 }
};

static const int yamaha_indexscale[] = {
    230, 230, 230, 230, 307, 409, 512, 614,
    230, 230, 230, 230, 307, 409, 512, 614
};

static const int yamaha_difflookup[] = {
    1, 3, 5, 7, 9, 11, 13, 15,
    -1, -3, -5, -7, -9, -11, -13, -15
};

/* end of tables */

typedef struct ADPCMChannelStatus {
    int predictor;
    short int step_index;
    int step;
    /* for encoding */
    int prev_sample;

    /* MS version */
    short sample1;
    short sample2;
    int coeff1;
    int coeff2;
    int idelta;
} ADPCMChannelStatus;

typedef struct ADPCMContext {
    int channel; /* for stereo MOVs, decode left, then decode right, then tell it's decoded */
    ADPCMChannelStatus status[2];
    short sample_buffer[32]; /* hold left samples while waiting for right samples */
} ADPCMContext;

/* XXX: implement encoding */

#ifdef CONFIG_ENCODERS
static int adpcm_encode_init(AVCodecContext *avctx)
{
    if (avctx->channels > 2)
        return -1; /* only stereo or mono =) */
    switch(avctx->codec->id) {
    case CODEC_ID_ADPCM_IMA_QT:
        av_log(avctx, AV_LOG_ERROR, "ADPCM: codec adpcm_ima_qt unsupported for encoding !\n");
        avctx->frame_size = 64; /* XXX: can multiple of avctx->channels * 64 (left and right blocks are interleaved) */
        return -1;
        break;
    case CODEC_ID_ADPCM_IMA_WAV:
        avctx->frame_size = (BLKSIZE - 4 * avctx->channels) * 8 / (4 * avctx->channels) + 1; /* each 16 bits sample gives one nibble */
                                                             /* and we have 4 bytes per channel overhead */
        avctx->block_align = BLKSIZE;
        /* seems frame_size isn't taken into account... have to buffer the samples :-( */
        break;
    case CODEC_ID_ADPCM_MS:
        avctx->frame_size = (BLKSIZE - 7 * avctx->channels) * 2 / avctx->channels + 2; /* each 16 bits sample gives one nibble */
                                                             /* and we have 7 bytes per channel overhead */
        avctx->block_align = BLKSIZE;
        break;
    case CODEC_ID_ADPCM_YAMAHA:
        avctx->frame_size = BLKSIZE * avctx->channels;
        avctx->block_align = BLKSIZE;
        break;
    case CODEC_ID_ADPCM_SWF:
        if (avctx->sample_rate != 11025 &&
            avctx->sample_rate != 22050 &&
            avctx->sample_rate != 44100) {
            av_log(avctx, AV_LOG_ERROR, "Sample rate must be 11025, 22050 or 44100\n");
            return -1;
        }
        avctx->frame_size = 512 * (avctx->sample_rate / 11025);
        break;
    default:
        return -1;
        break;
    }

    avctx->coded_frame= avcodec_alloc_frame();
    avctx->coded_frame->key_frame= 1;

    return 0;
}

static int adpcm_encode_close(AVCodecContext *avctx)
{
    av_freep(&avctx->coded_frame);

    return 0;
}


static inline unsigned char adpcm_ima_compress_sample(ADPCMChannelStatus *c, short sample)
{
    int delta = sample - c->prev_sample;
    int nibble = FFMIN(7, abs(delta)*4/step_table[c->step_index]) + (delta<0)*8;
    c->prev_sample += ((step_table[c->step_index] * yamaha_difflookup[nibble]) / 8);
    c->prev_sample = av_clip_int16(c->prev_sample);
    c->step_index = av_clip(c->step_index + index_table[nibble], 0, 88);
    return nibble;
}

static inline unsigned char adpcm_ms_compress_sample(ADPCMChannelStatus *c, short sample)
{
    int predictor, nibble, bias;

    predictor = (((c->sample1) * (c->coeff1)) + ((c->sample2) * (c->coeff2))) / 256;

    nibble= sample - predictor;
    if(nibble>=0) bias= c->idelta/2;
    else          bias=-c->idelta/2;

    nibble= (nibble + bias) / c->idelta;
    nibble= av_clip(nibble, -8, 7)&0x0F;

    predictor += (signed)((nibble & 0x08)?(nibble - 0x10):(nibble)) * c->idelta;

    c->sample2 = c->sample1;
    c->sample1 = av_clip_int16(predictor);

    c->idelta = (AdaptationTable[(int)nibble] * c->idelta) >> 8;
    if (c->idelta < 16) c->idelta = 16;

    return nibble;
}

static inline unsigned char adpcm_yamaha_compress_sample(ADPCMChannelStatus *c, short sample)
{
    int nibble, delta;

    if(!c->step) {
        c->predictor = 0;
        c->step = 127;
    }

    delta = sample - c->predictor;

    nibble = FFMIN(7, abs(delta)*4/c->step) + (delta<0)*8;

    c->predictor += ((c->step * yamaha_difflookup[nibble]) / 8);
    c->predictor = av_clip_int16(c->predictor);
    c->step = (c->step * yamaha_indexscale[nibble]) >> 8;
    c->step = av_clip(c->step, 127, 24567);

    return nibble;
}

typedef struct TrellisPath {
    int nibble;
    int prev;
} TrellisPath;

typedef struct TrellisNode {
    uint32_t ssd;
    int path;
    int sample1;
    int sample2;
    int step;
} TrellisNode;

static void adpcm_compress_trellis(AVCodecContext *avctx, const short *samples,
                                   uint8_t *dst, ADPCMChannelStatus *c, int n)
{
#define FREEZE_INTERVAL 128
    //FIXME 6% faster if frontier is a compile-time constant
    const int frontier = 1 << avctx->trellis;
    const int stride = avctx->channels;
    const int version = avctx->codec->id;
    const int max_paths = frontier*FREEZE_INTERVAL;
    TrellisPath paths[max_paths], *p;
    TrellisNode node_buf[2][frontier];
    TrellisNode *nodep_buf[2][frontier];
    TrellisNode **nodes = nodep_buf[0]; // nodes[] is always sorted by .ssd
    TrellisNode **nodes_next = nodep_buf[1];
    int pathn = 0, froze = -1, i, j, k;

    assert(!(max_paths&(max_paths-1)));

    memset(nodep_buf, 0, sizeof(nodep_buf));
    nodes[0] = &node_buf[1][0];
    nodes[0]->ssd = 0;
    nodes[0]->path = 0;
    nodes[0]->step = c->step_index;
    nodes[0]->sample1 = c->sample1;
    nodes[0]->sample2 = c->sample2;
    if((version == CODEC_ID_ADPCM_IMA_WAV) || (version == CODEC_ID_ADPCM_SWF))
        nodes[0]->sample1 = c->prev_sample;
    if(version == CODEC_ID_ADPCM_MS)
        nodes[0]->step = c->idelta;
    if(version == CODEC_ID_ADPCM_YAMAHA) {
        if(c->step == 0) {
            nodes[0]->step = 127;
            nodes[0]->sample1 = 0;
        } else {
            nodes[0]->step = c->step;
            nodes[0]->sample1 = c->predictor;
        }
    }

    for(i=0; i<n; i++) {
        TrellisNode *t = node_buf[i&1];
        TrellisNode **u;
        int sample = samples[i*stride];
        memset(nodes_next, 0, frontier*sizeof(TrellisNode*));
        for(j=0; j<frontier && nodes[j]; j++) {
            // higher j have higher ssd already, so they're unlikely to use a suboptimal next sample too
            const int range = (j < frontier/2) ? 1 : 0;
            const int step = nodes[j]->step;
            int nidx;
            if(version == CODEC_ID_ADPCM_MS) {
                const int predictor = ((nodes[j]->sample1 * c->coeff1) + (nodes[j]->sample2 * c->coeff2)) / 256;
                const int div = (sample - predictor) / step;
                const int nmin = av_clip(div-range, -8, 6);
                const int nmax = av_clip(div+range, -7, 7);
                for(nidx=nmin; nidx<=nmax; nidx++) {
                    const int nibble = nidx & 0xf;
                    int dec_sample = predictor + nidx * step;
#define STORE_NODE(NAME, STEP_INDEX)\
                    int d;\
                    uint32_t ssd;\
                    dec_sample = av_clip_int16(dec_sample);\
                    d = sample - dec_sample;\
                    ssd = nodes[j]->ssd + d*d;\
                    if(nodes_next[frontier-1] && ssd >= nodes_next[frontier-1]->ssd)\
                        continue;\
                    /* Collapse any two states with the same previous sample value. \
                     * One could also distinguish states by step and by 2nd to last
                     * sample, but the effects of that are negligible. */\
                    for(k=0; k<frontier && nodes_next[k]; k++) {\
                        if(dec_sample == nodes_next[k]->sample1) {\
                            assert(ssd >= nodes_next[k]->ssd);\
                            goto next_##NAME;\
                        }\
                    }\
                    for(k=0; k<frontier; k++) {\
                        if(!nodes_next[k] || ssd < nodes_next[k]->ssd) {\
                            TrellisNode *u = nodes_next[frontier-1];\
                            if(!u) {\
                                assert(pathn < max_paths);\
                                u = t++;\
                                u->path = pathn++;\
                            }\
                            u->ssd = ssd;\
                            u->step = STEP_INDEX;\
                            u->sample2 = nodes[j]->sample1;\
                            u->sample1 = dec_sample;\
                            paths[u->path].nibble = nibble;\
                            paths[u->path].prev = nodes[j]->path;\
                            memmove(&nodes_next[k+1], &nodes_next[k], (frontier-k-1)*sizeof(TrellisNode*));\
                            nodes_next[k] = u;\
                            break;\
                        }\
                    }\
                    next_##NAME:;