之前的文章里已經(jīng)介紹了音頻碼流AAC和視頻碼流H.264微猖，接下來就是要把他們打包并封裝格式了涕刚。

FLV用途

FLV(Flash Video)是Adobe公司設(shè)計開發(fā)的一種流行的流媒體格式几晤，由于其視頻文件體積輕巧宫补、封裝簡單等特點盛撑，使其很適合在互聯(lián)網(wǎng)上進(jìn)行應(yīng)用碎节。 除了播放視頻，在直播時也可以使用抵卫。采用FLV格式封裝的文件后綴為.flv

FLV封裝格式

FLV封裝格式和其他格式一樣狮荔，一個文件頭+一個文件體。結(jié)構(gòu)如下圖：

FLV format

文件頭

FLV文件頭總共有9Byte介粘，文件頭部的內(nèi)容結(jié)構(gòu)如下：

header

typedef unsigned char byte;
typedef struct {
    byte Signature[3];
    byte Version;
    byte Flags;
    uint DataOffset;
} FLV_HEADER;

其中Signature[3]是存儲了"F"(0x46)轴合，"L"(0x4c)，“V”(0x56)簽名的3個字節(jié)碗短；然后是version版本受葛，表示FLV的版本，占1個字節(jié)偎谁；Flags是音頻流標(biāo)識总滩，前5位為保留位，第7位同樣保留巡雨，全為0闰渔，第6位表示是否存在音頻tag，第8位表示是否存在視頻tag铐望。剩下的4個字節(jié)為整個文件頭所占的大小，數(shù)據(jù)的起始位置就是從文件開頭偏移這么多的大小。

音頻tag解析

Flags音頻流標(biāo)志位的第6位是判斷是否存在音頻tag锻全，音頻tag的結(jié)構(gòu)如下：

音頻tag

音頻tag開始的第一個字節(jié)包含了音頻數(shù)據(jù)的參數(shù)信息泪漂，從第二個字節(jié)開始為音頻流赘风。如上圖所示，在第1個字節(jié)中劣砍，前4個bit為音頻編碼類型：

0 Linear PCM装畅，platform endian
1 ADPCM
2 MP3
3 Linear PCM蚂夕，little endian
4 Nellymoser 16-kHz mono
5 Nellymoser 8-kHz mono
6 Nellymoser
7 G.711 A-law logarithmic PCM
8 G.711 mu-law logarithmic PCM
9 reserved
10 AAC
14 MP3 8-Khz
15 Device-specific sound

第5牍汹、6個bit表示音頻采樣率：

0 5.5kHz
1 11KHz
2 22 kHz
3 44 kHz

FLV并不支持48KHz的采樣率

第7個bit表示音頻采樣精度：

0 8bits
1 16bits

第8個表示音頻類型：

0 sndMono
1 sndStereo

視頻tag解析

Flags音頻流標(biāo)志位的第8位是判斷是否存在視頻tag露该，視頻tag的結(jié)構(gòu)如下：

視頻tag

2個字節(jié)第一個字節(jié)包含視頻數(shù)據(jù)的參數(shù)信息，第2個字節(jié)為視頻流數(shù)據(jù)底靠。第一個字節(jié)的前4個bit的數(shù)值表示幀類型：

1 keyframe （for AVC害晦，a seekable frame）
2 inter frame （for AVC，a nonseekable frame）
3 disposable inter frame （H.263 only）
4 generated keyframe （reserved for server use）
5 video info/command frame

第1位的后4個bit的數(shù)值表示視頻編碼類型：

1 JPEG （currently unused）
2 Sorenson H.263
3 Screen video
4 On2 VP6
5 On2 VP6 with alpha channel
6 Screen video version 2
7 AVC

文件體

FLV文件體結(jié)構(gòu)如下：

body

FLV文件體是由一連串的back-pointers和tags構(gòu)成：

body-pointers+tags

Tag

這里的tag同樣是由header和data兩部分組成暑中，tag header里存放的數(shù)據(jù)如下：

tag header

Tag的類型可以是視頻壹瘟、音頻和Script(腳本類型)，視頻tag和音頻tag在上面已經(jīng)介紹過了鳄逾，下面介紹一下Script Tag這種類型稻轨。

typedef unsigned char byte;
typedef unsigned int uint;
typedef struct {
    byte TagType;
    byte DataSize[3];
    byte Timestamp[3];
    uint Reserved;
} TAG_HEADER;

Script Tag

該類型Tag又被稱為MetaData Tag,存放一些關(guān)于FLV視頻和音頻的元信息，比如：duration雕凹、width殴俱、height等。通常該類型Tag會作為FLV文件的第一個tag枚抵，并且只有一個线欲，跟在File Header后。該類型Tag Data的結(jié)構(gòu)如下所示:

script tag

針對第一個AMF包：第1個字節(jié)表示AMF類型俄精，一般為0x02询筏，表示字符串；第2竖慧、3個字節(jié)用來標(biāo)識字符長度嫌套，一般總是0x000A。后面字節(jié)為具體的字符串圾旨。

針對第二個AMF包：第1個字節(jié)表示AMF包類型踱讨，一般總是0x08，表示數(shù)組砍的。第2-5個字節(jié)表示數(shù)組元素個數(shù)痹筛，后面幾位各數(shù)組元素的封裝，數(shù)組元素為(名稱-值)組成的對：

duration 時長
width 視頻寬度
height 視頻高度
videodatarate 視頻碼率
framerate 視頻幀率
videocodecid 視頻編碼方式
audiosamplerate 音頻采樣率
audiosamplesize 音頻采樣精度
stereo 是否為立體聲
audiocodecid 音頻編碼方式
filesize 文件大小

FLV解析

來自雷神

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
 
//Important!
#pragma pack(1)
 
 
#define TAG_TYPE_SCRIPT 18
#define TAG_TYPE_AUDIO  8
#define TAG_TYPE_VIDEO  9
 
typedef unsigned char byte;
typedef unsigned int uint;
 
typedef struct {
    byte Signature[3];
    byte Version;
    byte Flags;
    uint DataOffset;
} FLV_HEADER;
 
typedef struct {
    byte TagType;
    byte DataSize[3];
    byte Timestamp[3];
    uint Reserved;
} TAG_HEADER;
 
 
//reverse_bytes - turn a BigEndian byte array into a LittleEndian integer
uint reverse_bytes(byte *p, char c) {
    int r = 0;
    int i;
    for (i=0; i<c; i++) 
        r |= ( *(p+i) << (((c-1)*8)-8*i));
    return r;
}
 
/**
 * Analysis FLV file
 * @param url    Location of input FLV file.
 */
 
int simplest_flv_parser(char *url){
 
    //whether output audio/video stream
    int output_a=1;
    int output_v=1;
    //-------------
    FILE *ifh=NULL,*vfh=NULL, *afh = NULL;
 
    //FILE *myout=fopen("output_log.txt","wb+");
    FILE *myout=stdout;
 
    FLV_HEADER flv;
    TAG_HEADER tagheader;
    uint previoustagsize, previoustagsize_z=0;
    uint ts=0, ts_new=0;
 
    ifh = fopen(url, "rb+");
    if ( ifh== NULL) {
        printf("Failed to open files!");
        return -1;
    }
 
    //FLV file header
    fread((char *)&flv,1,sizeof(FLV_HEADER),ifh);
 
    fprintf(myout,"============== FLV Header ==============\n");
    fprintf(myout,"Signature:  0x %c %c %c\n",flv.Signature[0],flv.Signature[1],flv.Signature[2]);
    fprintf(myout,"Version:    0x %X\n",flv.Version);
    fprintf(myout,"Flags  :    0x %X\n",flv.Flags);
    fprintf(myout,"HeaderSize: 0x %X\n",reverse_bytes((byte *)&flv.DataOffset, sizeof(flv.DataOffset)));
    fprintf(myout,"========================================\n");
 
    //move the file pointer to the end of the header
    fseek(ifh, reverse_bytes((byte *)&flv.DataOffset, sizeof(flv.DataOffset)), SEEK_SET);
 
    //process each tag
    do {
 
        previoustagsize = _getw(ifh);
 
        fread((void *)&tagheader,sizeof(TAG_HEADER),1,ifh);
 
        //int temp_datasize1=reverse_bytes((byte *)&tagheader.DataSize, sizeof(tagheader.DataSize));
        int tagheader_datasize=tagheader.DataSize[0]*65536+tagheader.DataSize[1]*256+tagheader.DataSize[2];
        int tagheader_timestamp=tagheader.Timestamp[0]*65536+tagheader.Timestamp[1]*256+tagheader.Timestamp[2];
 
        char tagtype_str[10];
        switch(tagheader.TagType){
        case TAG_TYPE_AUDIO:sprintf(tagtype_str,"AUDIO");break;
        case TAG_TYPE_VIDEO:sprintf(tagtype_str,"VIDEO");break;
        case TAG_TYPE_SCRIPT:sprintf(tagtype_str,"SCRIPT");break;
        default:sprintf(tagtype_str,"UNKNOWN");break;
        }
        fprintf(myout,"[%6s] %6d %6d |",tagtype_str,tagheader_datasize,tagheader_timestamp);
 
        //if we are not past the end of file, process the tag
        if (feof(ifh)) {
            break;
        }
 
        //process tag by type
        switch (tagheader.TagType) {
 
        case TAG_TYPE_AUDIO:{ 
            char audiotag_str[100]={0};
            strcat(audiotag_str,"| ");
            char tagdata_first_byte;
            tagdata_first_byte=fgetc(ifh);
            int x=tagdata_first_byte&0xF0;
            x=x>>4;
            switch (x)
            {
            case 0:strcat(audiotag_str,"Linear PCM, platform endian");break;
            case 1:strcat(audiotag_str,"ADPCM");break;
            case 2:strcat(audiotag_str,"MP3");break;
            case 3:strcat(audiotag_str,"Linear PCM, little endian");break;
            case 4:strcat(audiotag_str,"Nellymoser 16-kHz mono");break;
            case 5:strcat(audiotag_str,"Nellymoser 8-kHz mono");break;
            case 6:strcat(audiotag_str,"Nellymoser");break;
            case 7:strcat(audiotag_str,"G.711 A-law logarithmic PCM");break;
            case 8:strcat(audiotag_str,"G.711 mu-law logarithmic PCM");break;
            case 9:strcat(audiotag_str,"reserved");break;
            case 10:strcat(audiotag_str,"AAC");break;
            case 11:strcat(audiotag_str,"Speex");break;
            case 14:strcat(audiotag_str,"MP3 8-Khz");break;
            case 15:strcat(audiotag_str,"Device-specific sound");break;
            default:strcat(audiotag_str,"UNKNOWN");break;
            }
            strcat(audiotag_str,"| ");
            x=tagdata_first_byte&0x0C;
            x=x>>2;
            switch (x)
            {
            case 0:strcat(audiotag_str,"5.5-kHz");break;
            case 1:strcat(audiotag_str,"1-kHz");break;
            case 2:strcat(audiotag_str,"22-kHz");break;
            case 3:strcat(audiotag_str,"44-kHz");break;
            default:strcat(audiotag_str,"UNKNOWN");break;
            }
            strcat(audiotag_str,"| ");
            x=tagdata_first_byte&0x02;
            x=x>>1;
            switch (x)
            {
            case 0:strcat(audiotag_str,"8Bit");break;
            case 1:strcat(audiotag_str,"16Bit");break;
            default:strcat(audiotag_str,"UNKNOWN");break;
            }
            strcat(audiotag_str,"| ");
            x=tagdata_first_byte&0x01;
            switch (x)
            {
            case 0:strcat(audiotag_str,"Mono");break;
            case 1:strcat(audiotag_str,"Stereo");break;
            default:strcat(audiotag_str,"UNKNOWN");break;
            }
            fprintf(myout,"%s",audiotag_str);
 
            //if the output file hasn't been opened, open it.
            if(output_a!=0&&afh == NULL){
                afh = fopen("output.mp3", "wb");
            }
 
            //TagData - First Byte Data
            int data_size=reverse_bytes((byte *)&tagheader.DataSize, sizeof(tagheader.DataSize))-1;
            if(output_a!=0){
                //TagData+1
                for (int i=0; i<data_size; i++)
                    fputc(fgetc(ifh),afh);
 
            }else{
                for (int i=0; i<data_size; i++)
                    fgetc(ifh);
            }
            break;
        }
        case TAG_TYPE_VIDEO:{
            char videotag_str[100]={0};
            strcat(videotag_str,"| ");
            char tagdata_first_byte;
            tagdata_first_byte=fgetc(ifh);
            int x=tagdata_first_byte&0xF0;
            x=x>>4;
            switch (x)
            {
            case 1:strcat(videotag_str,"key frame  ");break;
            case 2:strcat(videotag_str,"inter frame");break;
            case 3:strcat(videotag_str,"disposable inter frame");break;
            case 4:strcat(videotag_str,"generated keyframe");break;
            case 5:strcat(videotag_str,"video info/command frame");break;
            default:strcat(videotag_str,"UNKNOWN");break;
            }
            strcat(videotag_str,"| ");
            x=tagdata_first_byte&0x0F;
            switch (x)
            {
            case 1:strcat(videotag_str,"JPEG (currently unused)");break;
            case 2:strcat(videotag_str,"Sorenson H.263");break;
            case 3:strcat(videotag_str,"Screen video");break;
            case 4:strcat(videotag_str,"On2 VP6");break;
            case 5:strcat(videotag_str,"On2 VP6 with alpha channel");break;
            case 6:strcat(videotag_str,"Screen video version 2");break;
            case 7:strcat(videotag_str,"AVC");break;
            default:strcat(videotag_str,"UNKNOWN");break;
            }
            fprintf(myout,"%s",videotag_str);
 
            fseek(ifh, -1, SEEK_CUR);
            //if the output file hasn't been opened, open it.
            if (vfh == NULL&&output_v!=0) {
                //write the flv header (reuse the original file's hdr) and first previoustagsize
                    vfh = fopen("output.flv", "wb");
                    fwrite((char *)&flv,1, sizeof(flv),vfh);
                    fwrite((char *)&previoustagsize_z,1,sizeof(previoustagsize_z),vfh);
            }
#if 0
            //Change Timestamp
            //Get Timestamp
            ts = reverse_bytes((byte *)&tagheader.Timestamp, sizeof(tagheader.Timestamp));
            ts=ts*2;
            //Writeback Timestamp
            ts_new = reverse_bytes((byte *)&ts, sizeof(ts));
            memcpy(&tagheader.Timestamp, ((char *)&ts_new) + 1, sizeof(tagheader.Timestamp));
#endif
 
 
            //TagData + Previous Tag Size
            int data_size=reverse_bytes((byte *)&tagheader.DataSize, sizeof(tagheader.DataSize))+4;
            if(output_v!=0){
                //TagHeader
                fwrite((char *)&tagheader,1, sizeof(tagheader),vfh);
                //TagData
                for (int i=0; i<data_size; i++)
                    fputc(fgetc(ifh),vfh);
            }else{
                for (int i=0; i<data_size; i++)
                    fgetc(ifh);
            }
            //rewind 4 bytes, because we need to read the previoustagsize again for the loop's sake
            fseek(ifh, -4, SEEK_CUR);
 
            break;
            }
        default:
 
            //skip the data of this tag
            fseek(ifh, reverse_bytes((byte *)&tagheader.DataSize, sizeof(tagheader.DataSize)), SEEK_CUR);
        }
 
        fprintf(myout,"\n");
 
    } while (!feof(ifh));
 
 
    _fcloseall();
 
    return 0;
}

此程序可以分離FLV中的視頻碼流和音頻碼流廓鞠。

首先定義了兩個header帚稠，F(xiàn)LV_HEADER和TAG_HEADER，然后reverse_bytes()是將大端存儲的byte array翻轉(zhuǎn)為小端存儲的整型：

uint reverse_bytes(byte *p, char c) {
    int r = 0;
    int i;
    for (i=0; i<c; i++)  //直接和char型的c比較是因為char是“短的int”
        r |= ( *(p+i) << (((c-1)*8)-8*i));
    return r;
}

他的用法是這樣的：reverse_bytes((byte *)&flv.DataOffset, sizeof(flv.DataOffset))床佳，然后用一個循環(huán)滋早，將內(nèi)存中的數(shù)據(jù)先后通過|=與整型進(jìn)行比較，從而變成了小端存儲的整型砌们。

接著在simplest_flv_parser函數(shù)中先讀文件頭杆麸，再讀文件體搁进。然后一個do-while大循環(huán)，直到讀取到文件結(jié)尾昔头。利用 _getw()函數(shù) （是查找碼流中的整數(shù)的）讀取Previous tag size饼问，然后fread((void *)&tagheader,sizeof(TAG_HEADER),1,ifh);讀取tag header，再獲取tag data的字節(jié)數(shù):int tagheader_datasize=tagheader.DataSize[0]*pow(2, 16)+tagheader.DataSize[1]*pow(2, 8)+tagheader.DataSize[2];揭斧，獲取時戳：int tagheader_timestamp=tagheader.Timestamp[0]*pow(2, 16)+tagheader.Timestamp[1]*pow(2, 8)+tagheader.Timestamp[2];

接著就在之前讀取的tag header中tag typr所在的字節(jié)中莱革，進(jìn)行switch判斷：

如果是音頻：

case TAG_TYPE_AUDIO:{  //音頻
    char audiotag_str[100]={0};
    strcat(audiotag_str,"| ");
    char tagdata_first_byte;

    //讀取一個字符，音頻tag data區(qū)域的第一個字節(jié)未蝌，音頻的信息
    tagdata_first_byte=fgetc(ifh);
    // &操作獲取前四位驮吱，代表音頻格式
    int x=tagdata_first_byte&0xF0;
    //右移4位
    x=x>>4;
    //判斷音頻格式
    switch (x)
    {
        case 0:strcat(audiotag_str,"Linear PCM, platform endian");break;
        case 1:strcat(audiotag_str,"ADPCM");break;
        case 2:strcat(audiotag_str,"MP3");break;
        case 3:strcat(audiotag_str,"Linear PCM, little endian");break;
        case 4:strcat(audiotag_str,"Nellymoser 16-kHz mono");break;
        case 5:strcat(audiotag_str,"Nellymoser 8-kHz mono");break;
        case 6:strcat(audiotag_str,"Nellymoser");break;
        case 7:strcat(audiotag_str,"G.711 A-law logarithmic PCM");break;
        case 8:strcat(audiotag_str,"G.711 mu-law logarithmic PCM");break;
        case 9:strcat(audiotag_str,"reserved");break;
        case 10:strcat(audiotag_str,"AAC");break;
        case 11:strcat(audiotag_str,"Speex");break;
        case 14:strcat(audiotag_str,"MP3 8-Khz");break;
        case 15:strcat(audiotag_str,"Device-specific sound");break;
        default:strcat(audiotag_str,"UNKNOWN");break;
    }
    strcat(audiotag_str,"| ");

    //獲取5~6位茧妒，采樣率
    x=tagdata_first_byte&0x0C;
    //右移2位
    x=x>>2;
    //判斷采樣率
    switch (x)
    {
        case 0:strcat(audiotag_str,"5.5-kHz");break;
        case 1:strcat(audiotag_str,"1-kHz");break;
        case 2:strcat(audiotag_str,"22-kHz");break;
        case 3:strcat(audiotag_str,"44-kHz");break;
        default:strcat(audiotag_str,"UNKNOWN");break;
    }
    strcat(audiotag_str,"| ");

    //獲取第7位萧吠，采樣精度
    x=tagdata_first_byte&0x02;
    x=x>>1;
    switch (x)
    {
        case 0:strcat(audiotag_str,"8Bit");break;
        case 1:strcat(audiotag_str,"16Bit");break;
        default:strcat(audiotag_str,"UNKNOWN");break;
    }
    strcat(audiotag_str,"| ");

    //獲取第8位，音頻聲道數(shù)
    x=tagdata_first_byte&0x01;
    switch (x)
    {
        case 0:strcat(audiotag_str,"Mono");break;
        case 1:strcat(audiotag_str,"Stereo");break;
        default:strcat(audiotag_str,"UNKNOWN");break;
    }
    fprintf(myout,"%s",audiotag_str);

    //if the output file hasn't been opened, open it.
    if(output_a!=0&&afh == NULL){
        afh = fopen("output.mp3", "wb");
    }

    //TagData - First Byte Data
    //獲取tag Data字節(jié)數(shù),需要減去Tag Data區(qū)域的第一個字節(jié)
    int data_size=reverse_bytes((byte *)&tagheader.DataSize, sizeof(tagheader.DataSize))-1;

    //循環(huán)獲取字節(jié)寫入文件
    if(output_a!=0){
        //TagData+1
        for (int i=0; i<data_size; i++)
            fputc(fgetc(ifh),afh);

    }else{
        for (int i=0; i<data_size; i++)
            fgetc(ifh);
    }
    break;
}

首先先讀取文件頭桐筏，再將音頻流導(dǎo)出成mp3格式纸型，看了上面的音頻tag應(yīng)該很好理解這段代碼。

如果是視頻：

case TAG_TYPE_VIDEO:{   //視頻
    char videotag_str[100]={0};
    strcat(videotag_str,"| ");
    //讀取TagData區(qū)域第一個字節(jié)梅忌，取出前4位狰腌。包含視頻幀類型
    char tagdata_first_byte;
    tagdata_first_byte=fgetc(ifh);
    int x=tagdata_first_byte&0xF0;
    x=x>>4;
    switch (x)
    {
        case 1:strcat(videotag_str,"key frame  ");break;
        case 2:strcat(videotag_str,"inter frame");break;
        case 3:strcat(videotag_str,"disposable inter frame");break;
        case 4:strcat(videotag_str,"generated keyframe");break;
        case 5:strcat(videotag_str,"video info/command frame");break;
        default:strcat(videotag_str,"UNKNOWN");break;
    }
    strcat(videotag_str,"| ");

    //讀取TagData區(qū)域第一個字節(jié)，取出后4位牧氮。包含視頻編碼類型
    x=tagdata_first_byte&0x0F;
    switch (x)
    {
        case 1:strcat(videotag_str,"JPEG (currently unused)");break;
        case 2:strcat(videotag_str,"Sorenson H.263");break;
        case 3:strcat(videotag_str,"Screen video");break;
        case 4:strcat(videotag_str,"On2 VP6");break;
        case 5:strcat(videotag_str,"On2 VP6 with alpha channel");break;
        case 6:strcat(videotag_str,"Screen video version 2");break;
        case 7:strcat(videotag_str,"AVC");break;
        default:strcat(videotag_str,"UNKNOWN");break;
    }
    fprintf(myout,"%s",videotag_str);

    fseek(ifh, -1, SEEK_CUR);
    //if the output file hasn't been opened, open it.
    if (vfh == NULL&&output_v!=0) {
        //write the flv header (reuse the original file's hdr) and first previoustagsize
        vfh = fopen("output.flv", "wb");
        fwrite((char *)&flv,1, sizeof(flv),vfh);
        fwrite((char *)&previoustagsize_z,1,sizeof(previoustagsize_z),vfh);
    }
    #if 0
    //Change Timestamp
    //Get Timestamp
    ts = reverse_bytes((byte *)&tagheader.Timestamp, sizeof(tagheader.Timestamp));
    ts=ts*2;
    //Writeback Timestamp
    ts_new = reverse_bytes((byte *)&ts, sizeof(ts));
    memcpy(&tagheader.Timestamp, ((char *)&ts_new) + 1, sizeof(tagheader.Timestamp));
    #endif


    //TagData + Previous Tag Size
    int data_size=reverse_bytes((byte *)&tagheader.DataSize, sizeof(tagheader.DataSize))+4;
    if(output_v!=0){
        //TagHeader
        fwrite((char *)&tagheader,1, sizeof(tagheader),vfh);
        //TagData
        for (int i=0; i<data_size; i++)
            fputc(fgetc(ifh),vfh);
    }else{
        for (int i=0; i<data_size; i++)
            fgetc(ifh);
    }
    //rewind 4 bytes, because we need to read the previoustagsize again for the loop's sake
    fseek(ifh, -4, SEEK_CUR);

    break;
}

結(jié)果如下：

flv output