vp3.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

/*
 * Copyright (C) 2003-2004 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
 */

/**
 * @file vp3.c
 * On2 VP3 Video Decoder
 *
 * VP3 Video Decoder by Mike Melanson (mike at multimedia.cx)
 * For more information about the VP3 coding process, visit:
 *   http://multimedia.cx/
 *
 * Theora decoder by Alex Beregszaszi
 */

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

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

#include "vp3data.h"
#include "xiph.h"

#define FRAGMENT_PIXELS 8

/*
 * Debugging Variables
 *
 * Define one or more of the following compile-time variables to 1 to obtain
 * elaborate information about certain aspects of the decoding process.
 *
 * KEYFRAMES_ONLY: set this to 1 to only see keyframes (VP3 slideshow mode)
 * DEBUG_VP3: high-level decoding flow
 * DEBUG_INIT: initialization parameters
 * DEBUG_DEQUANTIZERS: display how the dequanization tables are built
 * DEBUG_BLOCK_CODING: unpacking the superblock/macroblock/fragment coding
 * DEBUG_MODES: unpacking the coding modes for individual fragments
 * DEBUG_VECTORS: display the motion vectors
 * DEBUG_TOKEN: display exhaustive information about each DCT token
 * DEBUG_VLC: display the VLCs as they are extracted from the stream
 * DEBUG_DC_PRED: display the process of reversing DC prediction
 * DEBUG_IDCT: show every detail of the IDCT process
 */

#define KEYFRAMES_ONLY 0

#define DEBUG_VP3 0
#define DEBUG_INIT 0
#define DEBUG_DEQUANTIZERS 0
#define DEBUG_BLOCK_CODING 0
#define DEBUG_MODES 0
#define DEBUG_VECTORS 0
#define DEBUG_TOKEN 0
#define DEBUG_VLC 0
#define DEBUG_DC_PRED 0
#define DEBUG_IDCT 0

#if DEBUG_VP3
#define debug_vp3(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
#else
static inline void debug_vp3(const char *format, ...) { }
#endif

#if DEBUG_INIT
#define debug_init(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
#else
static inline void debug_init(const char *format, ...) { }
#endif

#if DEBUG_DEQUANTIZERS
#define debug_dequantizers(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
#else
static inline void debug_dequantizers(const char *format, ...) { }
#endif

#if DEBUG_BLOCK_CODING
#define debug_block_coding(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
#else
static inline void debug_block_coding(const char *format, ...) { }
#endif

#if DEBUG_MODES
#define debug_modes(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
#else
static inline void debug_modes(const char *format, ...) { }
#endif

#if DEBUG_VECTORS
#define debug_vectors(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
#else
static inline void debug_vectors(const char *format, ...) { }
#endif

#if DEBUG_TOKEN
#define debug_token(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
#else
static inline void debug_token(const char *format, ...) { }
#endif

#if DEBUG_VLC
#define debug_vlc(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
#else
static inline void debug_vlc(const char *format, ...) { }
#endif

#if DEBUG_DC_PRED
#define debug_dc_pred(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
#else
static inline void debug_dc_pred(const char *format, ...) { }
#endif

#if DEBUG_IDCT
#define debug_idct(args...) av_log(NULL, AV_LOG_DEBUG, ## args)
#else
static inline void debug_idct(const char *format, ...) { }
#endif

typedef struct Coeff {
    struct Coeff *next;
    DCTELEM coeff;
    uint8_t index;
} Coeff;

//FIXME split things out into their own arrays
typedef struct Vp3Fragment {
    Coeff *next_coeff;
    /* address of first pixel taking into account which plane the fragment
     * lives on as well as the plane stride */
    int first_pixel;
    /* this is the macroblock that the fragment belongs to */
    uint16_t macroblock;
    uint8_t coding_method;
    uint8_t coeff_count;
    int8_t motion_x;
    int8_t motion_y;
} Vp3Fragment;

#define SB_NOT_CODED        0
#define SB_PARTIALLY_CODED  1
#define SB_FULLY_CODED      2

#define MODE_INTER_NO_MV      0
#define MODE_INTRA            1
#define MODE_INTER_PLUS_MV    2
#define MODE_INTER_LAST_MV    3
#define MODE_INTER_PRIOR_LAST 4
#define MODE_USING_GOLDEN     5
#define MODE_GOLDEN_MV        6
#define MODE_INTER_FOURMV     7
#define CODING_MODE_COUNT     8

/* special internal mode */
#define MODE_COPY             8

/* There are 6 preset schemes, plus a free-form scheme */
static int ModeAlphabet[7][CODING_MODE_COUNT] =
{
    /* this is the custom scheme */
    { 0, 0, 0, 0, 0, 0, 0, 0 },

    /* scheme 1: Last motion vector dominates */
    {    MODE_INTER_LAST_MV,    MODE_INTER_PRIOR_LAST,
         MODE_INTER_PLUS_MV,    MODE_INTER_NO_MV,
         MODE_INTRA,            MODE_USING_GOLDEN,
         MODE_GOLDEN_MV,        MODE_INTER_FOURMV },

    /* scheme 2 */
    {    MODE_INTER_LAST_MV,    MODE_INTER_PRIOR_LAST,
         MODE_INTER_NO_MV,      MODE_INTER_PLUS_MV,
         MODE_INTRA,            MODE_USING_GOLDEN,
         MODE_GOLDEN_MV,        MODE_INTER_FOURMV },

    /* scheme 3 */
    {    MODE_INTER_LAST_MV,    MODE_INTER_PLUS_MV,
         MODE_INTER_PRIOR_LAST, MODE_INTER_NO_MV,
         MODE_INTRA,            MODE_USING_GOLDEN,
         MODE_GOLDEN_MV,        MODE_INTER_FOURMV },

    /* scheme 4 */
    {    MODE_INTER_LAST_MV,    MODE_INTER_PLUS_MV,
         MODE_INTER_NO_MV,      MODE_INTER_PRIOR_LAST,
         MODE_INTRA,            MODE_USING_GOLDEN,
         MODE_GOLDEN_MV,        MODE_INTER_FOURMV },

    /* scheme 5: No motion vector dominates */
    {    MODE_INTER_NO_MV,      MODE_INTER_LAST_MV,
         MODE_INTER_PRIOR_LAST, MODE_INTER_PLUS_MV,
         MODE_INTRA,            MODE_USING_GOLDEN,
         MODE_GOLDEN_MV,        MODE_INTER_FOURMV },

    /* scheme 6 */
    {    MODE_INTER_NO_MV,      MODE_USING_GOLDEN,
         MODE_INTER_LAST_MV,    MODE_INTER_PRIOR_LAST,
         MODE_INTER_PLUS_MV,    MODE_INTRA,
         MODE_GOLDEN_MV,        MODE_INTER_FOURMV },

};

#define MIN_DEQUANT_VAL 2

typedef struct Vp3DecodeContext {
    AVCodecContext *avctx;
    int theora, theora_tables;
    int version;
    int width, height;
    AVFrame golden_frame;
    AVFrame last_frame;
    AVFrame current_frame;
    int keyframe;
    DSPContext dsp;
    int flipped_image;

    int qis[3];
    int nqis;
    int quality_index;
    int last_quality_index;

    int superblock_count;
    int superblock_width;
    int superblock_height;
    int y_superblock_width;
    int y_superblock_height;
    int c_superblock_width;
    int c_superblock_height;
    int u_superblock_start;
    int v_superblock_start;
    unsigned char *superblock_coding;

    int macroblock_count;
    int macroblock_width;
    int macroblock_height;

    int fragment_count;
    int fragment_width;
    int fragment_height;

    Vp3Fragment *all_fragments;
    Coeff *coeffs;
    Coeff *next_coeff;
    int fragment_start[3];

    ScanTable scantable;

    /* tables */
    uint16_t coded_dc_scale_factor[64];
    uint32_t coded_ac_scale_factor[64];
    uint8_t base_matrix[384][64];
    uint8_t qr_count[2][3];
    uint8_t qr_size [2][3][64];
    uint16_t qr_base[2][3][64];

    /* this is a list of indices into the all_fragments array indicating
     * which of the fragments are coded */
    int *coded_fragment_list;
    int coded_fragment_list_index;
    int pixel_addresses_inited;

    VLC dc_vlc[16];
    VLC ac_vlc_1[16];
    VLC ac_vlc_2[16];
    VLC ac_vlc_3[16];
    VLC ac_vlc_4[16];

    VLC superblock_run_length_vlc;
    VLC fragment_run_length_vlc;
    VLC mode_code_vlc;
    VLC motion_vector_vlc;

    /* these arrays need to be on 16-byte boundaries since SSE2 operations
     * index into them */
    DECLARE_ALIGNED_16(int16_t, qmat[2][4][64]);        //<qmat[is_inter][plane]

    /* This table contains superblock_count * 16 entries. Each set of 16
     * numbers corresponds to the fragment indices 0..15 of the superblock.
     * An entry will be -1 to indicate that no entry corresponds to that
     * index. */
    int *superblock_fragments;

    /* This table contains superblock_count * 4 entries. Each set of 4
     * numbers corresponds to the macroblock indices 0..3 of the superblock.
     * An entry will be -1 to indicate that no entry corresponds to that
     * index. */
    int *superblock_macroblocks;

    /* This table contains macroblock_count * 6 entries. Each set of 6
     * numbers corresponds to the fragment indices 0..5 which comprise
     * the macroblock (4 Y fragments and 2 C fragments). */
    int *macroblock_fragments;
    /* This is an array that indicates how a particular macroblock
     * is coded. */
    unsigned char *macroblock_coding;

    int first_coded_y_fragment;
    int first_coded_c_fragment;
    int last_coded_y_fragment;
    int last_coded_c_fragment;

    uint8_t edge_emu_buffer[9*2048]; //FIXME dynamic alloc
    int8_t qscale_table[2048]; //FIXME dynamic alloc (width+15)/16

    /* Huffman decode */
    int hti;
    unsigned int hbits;
    int entries;
    int huff_code_size;
    uint16_t huffman_table[80][32][2];

    uint32_t filter_limit_values[64];
    int bounding_values_array[256];
} Vp3DecodeContext;

/************************************************************************
 * VP3 specific functions
 ************************************************************************/

/*