truemotion1.c

从FFMPEG转换而来的H264解码程序,VC下编译..

/*
 * Duck TrueMotion 1.0 Decoder
 * Copyright (C) 2003 Alex Beregszaszi & Mike Melanson
 *
 * 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
 */

/**
 * @file truemotion1.c
 * Duck TrueMotion v1 Video Decoder by
 * Alex Beregszaszi and
 * Mike Melanson ([e-mail])
 *
 * The TrueMotion v1 decoder presently only decodes 16-bit TM1 data and
 * outputs RGB555 (or RGB565) data. 24-bit TM1 data is not supported yet.
 */

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>

#include "avcodec.h"
#include "dsputil.h"

#include "truemotion1data.h"

typedef struct TrueMotion1Context {
    AVCodecContext *avctx;
    AVFrame frame;
    AVFrame prev_frame;

    uint8_t *buf;
    int size;

    uint8_t *mb_change_bits;
    int mb_change_bits_row_size;
    uint8_t *index_stream;
    int index_stream_size;

    int flags;
    int x, y, w, h;

    uint32_t y_predictor_table[1024];
    uint32_t c_predictor_table[1024];
    uint32_t fat_y_predictor_table[1024];
    uint32_t fat_c_predictor_table[1024];

    int compression;
    int block_type;
    int block_width;
    int block_height;

    int16_t ydt[8];
    int16_t cdt[8];
    int16_t fat_ydt[8];
    int16_t fat_cdt[8];

    int last_deltaset, last_vectable;

    unsigned int *vert_pred;

} TrueMotion1Context;

#define FLAG_SPRITE         32
#define FLAG_KEYFRAME       16
#define FLAG_INTERFRAME      8
#define FLAG_INTERPOLATED    4

struct frame_header {
    uint8_t header_size;
    uint8_t compression;
    uint8_t deltaset;
    uint8_t vectable;
    uint16_t ysize;
    uint16_t xsize;
    uint16_t checksum;
    uint8_t version;
    uint8_t header_type;
    uint8_t flags;
    uint8_t control;
    uint16_t xoffset;
    uint16_t yoffset;
    uint16_t width;
    uint16_t height;
};

#define ALGO_NOP        0
#define ALGO_RGB16V     1
#define ALGO_RGB16H     2
#define ALGO_RGB24H     3

/* these are the various block sizes that can occupy a 4x4 block */
#define BLOCK_2x2  0
#define BLOCK_2x4  1
#define BLOCK_4x2  2
#define BLOCK_4x4  3

typedef struct comp_types {
    int algorithm;
    int block_width; // vres
    int block_height; // hres
    int block_type;
} comp_types;

/* { valid for metatype }, algorithm, num of deltas, vert res, horiz res */
static comp_types compression_types[17] = {
    { ALGO_NOP,    0, 0, 0 },

    { ALGO_RGB16V, 4, 4, BLOCK_4x4 },
    { ALGO_RGB16H, 4, 4, BLOCK_4x4 },
    { ALGO_RGB16V, 4, 2, BLOCK_4x2 },
    { ALGO_RGB16H, 4, 2, BLOCK_4x2 },

    { ALGO_RGB16V, 2, 4, BLOCK_2x4 },
    { ALGO_RGB16H, 2, 4, BLOCK_2x4 },
    { ALGO_RGB16V, 2, 2, BLOCK_2x2 },
    { ALGO_RGB16H, 2, 2, BLOCK_2x2 },

    { ALGO_NOP,    4, 4, BLOCK_4x4 },
    { ALGO_RGB24H, 4, 4, BLOCK_4x4 },
    { ALGO_NOP,    4, 2, BLOCK_4x2 },
    { ALGO_RGB24H, 4, 2, BLOCK_4x2 },

    { ALGO_NOP,    2, 4, BLOCK_2x4 },
    { ALGO_RGB24H, 2, 4, BLOCK_2x4 },
    { ALGO_NOP,    2, 2, BLOCK_2x2 },
    { ALGO_RGB24H, 2, 2, BLOCK_2x2 }
};

static void select_delta_tables(TrueMotion1Context *s, int delta_table_index)
{
    int i;

    if (delta_table_index > 3)
        return;

    memcpy(s->ydt, ydts[delta_table_index], 8 * sizeof(int16_t));
    memcpy(s->cdt, cdts[delta_table_index], 8 * sizeof(int16_t));
    memcpy(s->fat_ydt, fat_ydts[delta_table_index], 8 * sizeof(int16_t));
    memcpy(s->fat_cdt, fat_cdts[delta_table_index], 8 * sizeof(int16_t));

    /* Y skinny deltas need to be halved for some reason; maybe the
     * skinny Y deltas should be modified */
    for (i = 0; i < 8; i++)
    {
        /* drop the lsb before dividing by 2-- net effect: round down
         * when dividing a negative number (e.g., -3/2 = -2, not -1) */
        s->ydt[i] &= 0xFFFE;
        s->ydt[i] /= 2;
    }
}

#ifdef WORDS_BIGENDIAN
static int make_ydt15_entry(int p2, int p1, int16_t *ydt)
#else
static int make_ydt15_entry(int p1, int p2, int16_t *ydt)
#endif
{
    int lo, hi;

    lo = ydt[p1];
    lo += (lo << 5) + (lo << 10);
    hi = ydt[p2];
    hi += (hi << 5) + (hi << 10);
    return ((lo + (hi << 16)) << 1);
}

#ifdef WORDS_BIGENDIAN
static int make_cdt15_entry(int p2, int p1, int16_t *cdt)
#else
static int make_cdt15_entry(int p1, int p2, int16_t *cdt)
#endif
{
    int r, b, lo;

    b = cdt[p2];
    r = cdt[p1] << 10;
    lo = b + r;
    return ((lo + (lo << 16)) << 1);
}

#ifdef WORDS_BIGENDIAN
static int make_ydt16_entry(int p2, int p1, int16_t *ydt)
#else
static int make_ydt16_entry(int p1, int p2, int16_t *ydt)
#endif
{
    int lo, hi;

    lo = ydt[p1];
    lo += (lo << 6) + (lo << 11);
    hi = ydt[p2];
    hi += (hi << 6) + (hi << 11);
    return ((lo + (hi << 16)) << 1);
}

#ifdef WORDS_BIGENDIAN
static int make_cdt16_entry(int p2, int p1, int16_t *cdt)
#else
static int make_cdt16_entry(int p1, int p2, int16_t *cdt)
#endif
{
    int r, b, lo;

    b = cdt[p2];
    r = cdt[p1] << 11;
    lo = b + r;
    return ((lo + (lo << 16)) << 1);
}

#ifdef WORDS_BIGENDIAN
static int make_ydt24_entry(int p2, int p1, int16_t *ydt)
#else
static int make_ydt24_entry(int p1, int p2, int16_t *ydt)
#endif
{
    int lo, hi;

    lo = ydt[p1];
    hi = ydt[p2];
    return ((lo + (hi << 8) + (hi << 16)) << 1);
}

#ifdef WORDS_BIGENDIAN
static int make_cdt24_entry(int p2, int p1, int16_t *cdt)
#else
static int make_cdt24_entry(int p1, int p2, int16_t *cdt)
#endif
{
    int r, b;

    b = cdt[p2];
    r = cdt[p1]<<16;
    return ((b+r) << 1);
}

static void gen_vector_table15(TrueMotion1Context *s, const uint8_t *sel_vector_table)
{
    int len, i, j;
    unsigned char delta_pair;

    for (i = 0; i < 1024; i += 4)
    {
        len = *sel_vector_table++ / 2;
        for (j = 0; j < len; j++)
        {
            delta_pair = *sel_vector_table++;
            s->y_predictor_table[i+j] = 0xfffffffe &
                make_ydt15_entry(delta_pair >> 4, delta_pair & 0xf, s->ydt);
            s->c_predictor_table[i+j] = 0xfffffffe &
                make_cdt15_entry(delta_pair >> 4, delta_pair & 0xf, s->cdt);
        }
        s->y_predictor_table[i+(j-1)] |= 1;
        s->c_predictor_table[i+(j-1)] |= 1;
    }
}

static void gen_vector_table16(TrueMotion1Context *s, const uint8_t *sel_vector_table)
{
    int len, i, j;
    unsigned char delta_pair;

    for (i = 0; i < 1024; i += 4)
    {
        len = *sel_vector_table++ / 2;
        for (j = 0; j < len; j++)
        {
            delta_pair = *sel_vector_table++;
            s->y_predictor_table[i+j] = 0xfffffffe &
                make_ydt16_entry(delta_pair >> 4, delta_pair & 0xf, s->ydt);
            s->c_predictor_table[i+j] = 0xfffffffe &
                make_cdt16_entry(delta_pair >> 4, delta_pair & 0xf, s->cdt);
        }
        s->y_predictor_table[i+(j-1)] |= 1;
        s->c_predictor_table[i+(j-1)] |= 1;
    }
}

static void gen_vector_table24(TrueMotion1Context *s, const uint8_t *sel_vector_table)
{
    int len, i, j;
    unsigned char delta_pair;

    for (i = 0; i < 1024; i += 4)
    {
        len = *sel_vector_table++ / 2;
        for (j = 0; j < len; j++)
        {
            delta_pair = *sel_vector_table++;
            s->y_predictor_table[i+j] = 0xfffffffe &
                make_ydt24_entry(delta_pair >> 4, delta_pair & 0xf, s->ydt);
            s->c_predictor_table[i+j] = 0xfffffffe &
                make_cdt24_entry(delta_pair >> 4, delta_pair & 0xf, s->cdt);
            s->fat_y_predictor_table[i+j] = 0xfffffffe &
                make_ydt24_entry(delta_pair >> 4, delta_pair & 0xf, s->fat_ydt);
            s->fat_c_predictor_table[i+j] = 0xfffffffe &
                make_cdt24_entry(delta_pair >> 4, delta_pair & 0xf, s->fat_cdt);
        }
        s->y_predictor_table[i+(j-1)] |= 1;
        s->c_predictor_table[i+(j-1)] |= 1;
        s->fat_y_predictor_table[i+(j-1)] |= 1;
        s->fat_c_predictor_table[i+(j-1)] |= 1;
    }
}

/* Returns the number of bytes consumed from the bytestream. Returns -1 if
 * there was an error while decoding the header */
static int truemotion1_decode_header(TrueMotion1Context *s)
{
    int i;
    struct frame_header header;
    uint8_t header_buffer[128];  /* logical maximum size of the header */
    const uint8_t *sel_vector_table;

    /* There is 1 change bit per 4 pixels, so each change byte represents
     * 32 pixels; divide width by 4 to obtain the number of change bits and
     * then round up to the nearest byte. */
    s->mb_change_bits_row_size = ((s->avctx->width >> 2) + 7) >> 3;

    header.header_size = ((s->buf[0] >> 5) | (s->buf[0] << 3)) & 0x7f;
    if (s->buf[0] < 0x10)
    {
        av_log(s->avctx, AV_LOG_ERROR, "invalid header size (%d)\n", s->buf[0]);
        return -1;
    }

    /* unscramble the header bytes with a XOR operation */
    memset(header_buffer, 0, 128);
    for (i = 1; i < header.header_size; i++)
        header_buffer[i - 1] = s->buf[i] ^ s->buf[i + 1];

    header.compression = header_buffer[0];
    header.deltaset = header_buffer[1];
    header.vectable = header_buffer[2];
    header.ysize = AV_RL16(&header_buffer[3]);
    header.xsize = AV_RL16(&header_buffer[5]);
    header.checksum = AV_RL16(&header_buffer[7]);
    header.version = header_buffer[9];
    header.header_type = header_buffer[10];
    header.flags = header_buffer[11];
    header.control = header_buffer[12];

    /* Version 2 */
    if (header.version >= 2)
    {
        if (header.header_type > 3)
        {
            av_log(s->avctx, AV_LOG_ERROR, "invalid header type (%d)\n", header.header_type);
            return -1;
        } else if ((header.header_type == 2) || (header.header_type == 3)) {
            s->flags = header.flags;
            if (!(s->flags & FLAG_INTERFRAME))
                s->flags |= FLAG_KEYFRAME;
        } else
            s->flags = FLAG_KEYFRAME;
    } else /* Version 1 */
        s->flags = FLAG_KEYFRAME;

    if (s->flags & FLAG_SPRITE) {
        av_log(s->avctx, AV_LOG_INFO, "SPRITE frame found, please report the sample to the developers\n");
        /* FIXME header.width, height, xoffset and yoffset aren't initialized */
#if 0
        s->w = header.width;
        s->h = header.height;
        s->x = header.xoffset;
        s->y = header.yoffset;
#else
        return -1;
#endif
    } else {
        s->w = header.xsize;
        s->h = header.ysize;
        if (header.header_type < 2) {
            if ((s->w < 213) && (s->h >= 176))
            {
                s->flags |= FLAG_INTERPOLATED;
                av_log(s->avctx, AV_LOG_INFO, "INTERPOLATION selected, please report the sample to the developers\n");
            }
        }
    }

    if (header.compression > 17) {
        av_log(s->avctx, AV_LOG_ERROR, "invalid compression type (%d)\n", header.compression);
        return -1;
    }

    if ((header.deltaset != s->last_deltaset) ||
        (header.vectable != s->last_vectable))
        select_delta_tables(s, header.deltaset);

    if ((header.compression & 1) && header.header_type)
        sel_vector_table = pc_tbl2;
    else {
        if (header.vectable < 4)
            sel_vector_table = tables[header.vectable - 1];
        else {
            av_log(s->avctx, AV_LOG_ERROR, "invalid vector table id (%d)\n", header.vectable);
            return -1;
        }
    }

    // FIXME: where to place this ?!?!
    if (compression_types[header.compression].algorithm == ALGO_RGB24H)
        s->avctx->pix_fmt = PIX_FMT_RGB32;
    else
        s->avctx->pix_fmt = PIX_FMT_RGB555; // RGB565 is supported as well

    if ((header.deltaset != s->last_deltaset) || (header.vectable != s->last_vectable))
    {
        if (compression_types[header.compression].algorithm == ALGO_RGB24H)
            gen_vector_table24(s, sel_vector_table);
        else
        if (s->avctx->pix_fmt == PIX_FMT_RGB555)
            gen_vector_table15(s, sel_vector_table);
        else
            gen_vector_table16(s, sel_vector_table);
    }

    /* set up pointers to the other key data chunks */
    s->mb_change_bits = s->buf + header.header_size;
    if (s->flags & FLAG_KEYFRAME) {
        /* no change bits specified for a keyframe; only index bytes */
        s->index_stream = s->mb_change_bits;
    } else {
        /* one change bit per 4x4 block */
        s->index_stream = s->mb_change_bits +
            (s->mb_change_bits_row_size * (s->avctx->height >> 2));
    }
    s->index_stream_size = s->size - (s->index_stream - s->buf);

    s->last_deltaset = header.deltaset;
    s->last_vectable = header.vectable;
    s->compression = header.compression;
    s->block_width = compression_types[header.compression].block_width;
    s->block_height = compression_types[header.compression].block_height;
    s->block_type = compression_types[header.compression].block_type;

    if (s->avctx->debug & FF_DEBUG_PICT_INFO)
        av_log(s->avctx, AV_LOG_INFO, "tables: %d / %d c:%d %dx%d t:%d %s%s%s%s\n",
            s->last_deltaset, s->last_vectable, s->compression, s->block_width,
            s->block_height, s->block_type,
            s->flags & FLAG_KEYFRAME ? " KEY" : "",
            s->flags & FLAG_INTERFRAME ? " INTER" : "",
            s->flags & FLAG_SPRITE ? " SPRITE" : "",
            s->flags & FLAG_INTERPOLATED ? " INTERPOL" : "");

    return header.header_size;
}