MP4文件格式简要解析

准备

安装mediaInfo、QTAtomViewer.exe、MP4info.exe等软件

视频一些参数

封装格式：我也不太清楚

编码格式：我也不太清楚

分辨率：图像的长和宽。

帧率：一秒播放的图片数，如15fps、30fps，体现在画面的流畅度。

码率：压缩后每秒显示的图片数据量，通俗地说，就是流量。如298kbps，体现在视频文件的大小和画质。

压缩比：分辨率*帧率/码率，简单来说就是压缩前每秒的数据量比上压缩后每秒的数据量。

音频一些参数

声道：录音时录音的位置不同产生不同的声道，有单声道、立体声（左右声道）、四声道环绕、5.1声道等。放歌时切换到左声道，两个音箱都播放左声道，切换到右声道时，两个音箱都播放右声道，切换到立体音时，左音箱播放左声道，右音箱播放右声道。可用暴风影音体验一下。

音轨： ①电视上看到录音时歌手都是在录音棚里戴着耳机唱歌，其实这就使用了多音轨技术，先把伴奏录好（音轨1），歌手再根据伴奏唱歌（音轨2）。②在K房使用的也是多音轨（原唱和伴唱）③有些老歌经常是粤语版和国语版一起唱，这也是多音轨技术

采样率：录制声音时对声音的采样频率，一般为44.1KHz

MP4文件组成

MP4由一个个的box（以前叫atom）组成，如下图，每个box包含不同的信息（再如下图），在程序中用结构体表示各个box，形象地说MP4文件就是由各个结构体组织起来的。所以，首先要了解这些box的组成和意义。

主要概念：sample、chunk、track

sample，在这里实际上就是一帧或连续几帧

chunk，一个或几个sample组成的单元

track，视频或音频序列，sample的集合

MP4主要由三个box组成，ftyp、mdat和moov。ftyp指示一些头部信息，通过ftyp box可以判断文件类型。mdat box存放的是媒体数据，就是播放的音频实实在在的元数据，比如我们需要取到第一帧的图片数据就存放在这个box里面。这么多数据，怎么知道这些数据的准确位置呢？这就要靠moov box来解析了。一般情况下，一个moov box包含一个mvhd和若干个track box。①mvhd box中存放一些create/modify time（基准时间是1904-1-1 0:00 AM）等信息，duration/time scale就是视频的长度（单位为s）。②track box一般包含tkhd和mdia两个box，tkhd中存放有视频长度和宽度等信息，，，，，，，最重要的是mdia -> minf -> stbl box下面的6个box，分别是stsd、stts、stss、stsc、stsz、stco，比如要查找第N秒的图片数据，步骤如下：

1、确定第N秒在媒体时间坐标系统中的位置，即N*time_scale

2、根据stts（time to sample）确定sample序号

3、根据stsc（sample to chunk）确定对应该sample的chunk

4、根据stco（chunk offset）中提取该chunk的偏移量

5、根据stsz（sample size）找到该sample在chunk内的偏移量和sample的大小。

例如：以test.mp4为例说明，用QTAtomViewer.exe查看此文件可知time_scale为90000，即一秒分为90000个刻度，现在假设要查看第一秒的数据，步骤如下：

1、一秒在媒体时间坐标系统中的位置，即1*90000 = 90000；

2、由stts，sample duration = 6000， 90000/6000 = 15，即sample序号为15

3、由stsc，因为samples Per chunk = 1，所以chunk序号也为15

4、由stco，序号为15的chunk的offset = 62719

5、由stsz，sample 在 chunk内的偏移为0，且大小为3554。

也就是说，程序到62719的地址开始，读取3554byte的数据，就是第一秒的图片数据。如果要查看第一帧（第0秒）的数据，同样按照上面的步骤可定位到地址28，读取5280byte的数据

现在知道了MP4文件的组成，也知道了怎么去获取第N帧的数据，那么无论程序多么复杂，它都是围绕这个主题来做文章了

最后附上一小段简单的视频解析的代码

#include <stdio.h>
#include <stdlib.h>
#include <time.h>#define BOX_FTYP 0x66747970
#define MDAT_BOX 0x6D646174
#define MOOV_BOX 0x6D6F6F76
#define MVHD_BOX 0x6D766864
#define TRAK_BOX 0x7472616B
#define IODS_BOX 0x696F6473
#define UDTA_BOX 0x75647461
#define FREE_BOX 0x66726565
#define TKHD_BOX 0x746B6864
#define MDIA_BOX 0x6D646961
#define MDHD_BOX 0x6D646864
#define HDLR_BOX 0x68646C72
#define MINF_BOX 0x6D696E66
#define DINF_BOX 0x64696E66
#define TREF_BOX 0x74726566
#define STBL_BOX 0x7374626C
#define STSD_BOX 0x73747364
#define MP4S_BOX 0x6D703473
#define ESDS_BOX 0x65736473
#define STTS_BOX 0x73747473
#define STSC_BOX 0x73747363
#define STSZ_BOX 0x7374737A
#define STCO_BOX 0x7374636F
#define STSS_BOX 0x73747373
#define CTTS_BOX 0x63747473
#define EDTS_BOX 0x65647473
#define VMHD_BOX 0x766D6864
#define HMHD_BOX 0x686D6864
#define SMHD_BOX 0x736D6864
#define NMHD_BOX 0xffffffff#define ODSM 0x6F64736D
#define SDSM    0x7364736D
#define VIDE    0x76696465
#define HINT    0x68696E74
#define SOUN    0x736F756Etypedef struct BOX_HEAD
{unsigned int size;unsigned int type;
}BOX_HEAD;typedef struct FTYP_BOX
{unsigned int ftyp_size;unsigned int ftyp_type;
}FTYP_BOX;typedef struct MOOV_HEADER_BOX
{unsigned int create_time;unsigned int modification_time;unsigned int time_scale;unsigned int duration;unsigned int preferred_rate;unsigned short preferred_volume;char reserved[10];char matrix_structure[36];char pre_defined[24];unsigned int next_track_id;
}MOOV_HEADER_BOX;typedef struct Track_head_box
{unsigned int create_time;unsigned int modification_time;unsigned int track_id;unsigned int reserved1;unsigned int duration;unsigned int reserved2[2];unsigned short layer;unsigned short alternate_group;unsigned short volume;unsigned short reserved3;char matrix[36];unsigned int width;unsigned int height;
}TRACK_HEAD_BOX;typedef struct Mdhd_head_box
{unsigned int create_time;unsigned int modification_time;unsigned int time_scale;unsigned int duration;unsigned short language;unsigned short quality;
}MDHD_HEAD_BOX;typedef struct Ret_Boxhead_Fileptr
{BOX_HEAD box_head;FILE *p;
}Ret_Boxhead_Fileptr;struct MP4_INFO
{unsigned int create_time;unsigned int modification_time;
}mp4_info;unsigned int big_to_small_endian_32bit(unsigned int const big)
{unsigned int data = big;unsigned int ret = 0;ret += (data>>0 & 0xff) << 24;ret += (data>>8 & 0xff) << 16;ret += (data>>16 & 0xff) << 8;ret += (data>>24 & 0xff) << 0;return ret;
}__int64 big_to_small_endian_64bit(__int64 const big)
{__int64 data = big;__int64 ret = 0;ret += (data>>0 & 0xff) << 56;ret += (data>>8 & 0xff) << 48;ret += (data>>16 & 0xff) << 40;ret += (data>>24 & 0xff) << 32;ret += (data>>32 & 0xff) << 24;ret += (data>>40 & 0xff) << 16;ret += (data>>48 & 0xff) << 8;ret += (data>>56 & 0xff) << 0;return ret;
}void print_box_type(unsigned int box_type, unsigned int num)
{for(unsigned int i = 0; i < num; i++){printf("\t");}switch(big_to_small_endian_32bit(box_type)){case BOX_FTYP:{printf("%s\n", "ftyp box");}break;case MOOV_BOX:{printf("%s\n", "moov box");}break;case MDAT_BOX:{printf("%s\n", "mdat box");}break;case MVHD_BOX:{printf("%s\n", "mvhd box");}break;case IODS_BOX:{printf("%s\n", "iods box");}break;case UDTA_BOX:{printf("%s\n", "udta box");}break;case TRAK_BOX:{printf("%s\n", "trak box");}break;case TKHD_BOX:{printf("%s\n", "tkhd box");}break;case MDIA_BOX:{printf("%s\n", "mdia box");}break;case MDHD_BOX:{printf("%s\n", "mdhd box");}break;case HDLR_BOX:{printf("%s\n", "hdlr box");}break;case MINF_BOX:{printf("%s\n", "minf box");}break;case DINF_BOX:{printf("%s\n", "dinf box");}break;case TREF_BOX:{printf("%s\n", "tref box");}break;case STBL_BOX:{printf("%s\n", "stbl box");}break;case STSD_BOX:{printf("%s\n", "stsd box");}break;case MP4S_BOX:{printf("%s\n", "mp4s box");}break;case ESDS_BOX:{printf("%s\n", "esds box");}break;case STTS_BOX:{printf("%s\n", "stts box");}break;case STSC_BOX:{printf("%s\n", "stsc box");}break;case STSZ_BOX:{printf("%s\n", "stsz box");}break;case STCO_BOX:{printf("%s\n", "stco box");}break;case STSS_BOX:{printf("%s\n", "stss box");}break;case CTTS_BOX:{printf("%s\n", "ctts box");}break;case EDTS_BOX:{printf("%s\n", "edts box");}break;case VMHD_BOX:{printf("%s\n", "vmhd box");        }break;case SMHD_BOX:{printf("%s\n", "smhd box");}break;case HMHD_BOX:{printf("%s\n", "hmhd box");}break;case NMHD_BOX:{printf("%s\n", "nmhd box");}break;}
}enum timeType
{CREATE_TIME,MODIFY_TIME
};void create_and_modify_time(unsigned int time, enum timeType type)
{int month_days[12] = {31, 28, 31, 30, 31, 30, 31, 31, 30, 31, 30, 31};int leap_month_days[12] = {31, 29, 31, 30, 31, 30, 31, 31, 30, 31, 30, 31};unsigned long local_time = (long)big_to_small_endian_32bit(time);unsigned int local_year = 1904+local_time/3600/24/365;//此处当做是每年365天，下面再微调int leap_years_count;//闰年数量//平年，不精确计算if(local_year % 4){leap_years_count = (local_year - 1904)/4 + 1;}//闰年else{leap_years_count = (local_year - 1904)/4;}unsigned int day_count;//由于上面把每年当作365天，此处计算到年初时过去的实际天数day_count = leap_years_count*366 + (local_year-1904-leap_years_count)*365;int leave_day_count = local_time/3600/24 - day_count;//算出剩下的天数int i=0;int sum = 31;int local_month = 0;int local_day = 0;/*下面程序算出一年中的哪一个月和哪一天*///平年if(local_year % 4){while(leave_day_count > sum){local_day = leave_day_count - sum + 1;//因为从零算起，所以天数要加1sum += month_days[++i];}}//闰年else{while(leave_day_count > sum){local_day = leave_day_count - sum + 1;sum += leap_month_days[++i];}}local_month = i+1;if(local_month == 1)//如果是一月份{local_day = leave_day_count + 1;}/*上面程序算出一年中的哪一个月和哪一天*/int out_of_day = leave_day_count + day_count;int leave_second = local_time - out_of_day * 3600 *24;//计算剩下的秒数if(type == CREATE_TIME){printf("编码时间：%d-%d-%d %d:%d:%d\n", local_year, local_month, local_day, leave_second/3600, leave_second%3600/60, leave_second%60);}else if(type == MODIFY_TIME){printf("修订时间：%d-%d-%d %d:%d:%d\n", local_year, local_month, local_day, leave_second/3600, leave_second%3600/60, leave_second%60);}}void mainmenu()
{printf("------------------------function menu-------------------------\n");printf("1.编码时间与修订时间\n");printf("2.影片长度\n");printf("3.影片声音采样频率、视频帧率、码率（比特率）、分辨率\n");printf("4.视频关键帧列表\n");printf("q.输入q退出\n");printf("------------------------function menu-------------------------\n");
}char get_user_input()
{char select;
/*这里scanf必须有\n或者空格，看下面的解释scanf("%c",&c) 与 scanf(" %c",&c)，后者只是在%前多了个空格，似乎没有什么区别，
但使用起来区别是很大的。scanf()作单字符输入时规定只接收一个字符，但它却把回车符
也作为字符对待的。这就造成程序中只有一个输入字符的scanf()语句时，问题还不大，
但如果后面还跟着第二个scanf()字符输入语句，这个scanf()就把前面输入的驾车符当作
输入字符了。这就在输入逻辑上造成了混乱，达不到人们预期的愿望。有了这个空格，
因为scanf()是跳过空格读字符的，就回避了这个问题。实践证明，这个空格放在%c后面也不能达到目的。
应当说，这也是比较巧妙的应用！   */scanf(" %c", &select);switch(select){case '1':{create_and_modify_time(mp4_info.create_time, CREATE_TIME);create_and_modify_time(mp4_info.modification_time, MODIFY_TIME);}break;case '2':{}break;case '3':{}break;case '4':{}break;case '5':{}break;}return select;
}Ret_Boxhead_Fileptr moov_box_parse(FILE *p)
{BOX_HEAD box_head;MOOV_HEADER_BOX mvhd_box;char version;char *flag = (char *)malloc(3);bool stbl_flag = false;fread(&box_head, sizeof(BOX_HEAD), 1, p);print_box_type(box_head.type, 1);//mvhd boxfread(&version, sizeof(char), 1, p);fread(flag, 3, 1, p);fread(&mvhd_box, sizeof(MOOV_HEADER_BOX), 1, p);mp4_info.create_time = mvhd_box.create_time;mp4_info.modification_time = mvhd_box.modification_time;fread(&box_head, sizeof(BOX_HEAD), 1, p);while(big_to_small_endian_32bit(box_head.type) != TRAK_BOX){print_box_type(box_head.type, 1);fseek(p, big_to_small_endian_32bit(box_head.size)-8, SEEK_CUR);fread(&box_head, sizeof(BOX_HEAD), 1, p);}print_box_type(box_head.type, 1);//trak boxwhile(big_to_small_endian_32bit(box_head.type) != MDAT_BOX &&big_to_small_endian_32bit(box_head.type) != FREE_BOX){if(0 == fread(&box_head, sizeof(BOX_HEAD), 1, p)){break;}while(big_to_small_endian_32bit(box_head.type) != TRAK_BOX &&big_to_small_endian_32bit(box_head.type) != MDAT_BOX &&big_to_small_endian_32bit(box_head.type) != FREE_BOX ){switch(big_to_small_endian_32bit(box_head.type)){case TKHD_BOX:{print_box_type(box_head.type, 2);//tkhd boxchar tkhd_version;char *tkhd_flag = (char *)malloc(3);TRACK_HEAD_BOX tkhd_box;fread(&tkhd_version, 1, 1, p);fread(tkhd_flag, 3, 1, p);fread(&tkhd_box, sizeof(TRACK_HEAD_BOX), 1, p);free(tkhd_flag);}break;case MDIA_BOX:{char mdhd_version;char *mdhd_flag = (char *)malloc(3);unsigned int minf_head_box_length;unsigned int handler_type;MDHD_HEAD_BOX mdhd_box;print_box_type(box_head.type, 2);//mdia boxfread(&box_head, sizeof(BOX_HEAD), 1, p);print_box_type(box_head.type, 3);//mdhd boxfread(&mdhd_version, 1, 1, p);fread(mdhd_flag, 3, 1, p);fread(&mdhd_box, sizeof(MDHD_HEAD_BOX), 1, p);fread(&box_head, sizeof(BOX_HEAD), 1, p);print_box_type(box_head.type, 3);//hdlr boxfseek(p, 8, SEEK_CUR);fread(&handler_type, 4, 1, p);fseek(p, big_to_small_endian_32bit(box_head.size)-20, SEEK_CUR);fread(&box_head, sizeof(BOX_HEAD), 1, p);print_box_type(box_head.type, 3);//minf box/*       fread(&minf_head_box_length, 4, 1, p);//minf head boxfseek(p, big_to_small_endian_32bit(minf_head_box_length)-4, SEEK_CUR);fread(&box_head, sizeof(BOX_HEAD), 1, p);while(big_to_small_endian_32bit(box_head.type) != STBL_BOX){print_box_type(box_head.type, 4);fseek(p, big_to_small_endian_32bit(box_head.size)-8, SEEK_CUR);fread(&box_head, sizeof(BOX_HEAD), 1, p);}print_box_type(box_head.type, 4);//stbl box*/switch(big_to_small_endian_32bit(handler_type)){case SOUN:case VIDE:case HINT:{unsigned int now_pos = ftell(p);unsigned int now_box_size = big_to_small_endian_32bit(box_head.size);while((ftell(p)-now_pos+8) != now_box_size){fread(&box_head, sizeof(BOX_HEAD), 1, p);fseek(p, big_to_small_endian_32bit(box_head.size)-8, SEEK_CUR);switch(big_to_small_endian_32bit(box_head.type)){case VMHD_BOX:case DINF_BOX:case HMHD_BOX:case SMHD_BOX:{print_box_type(box_head.type, 4);}break;case STBL_BOX:{print_box_type(box_head.type, 4);stbl_flag = true;fseek(p, 8-big_to_small_endian_32bit(box_head.size), SEEK_CUR);unsigned int stbl_pos = ftell(p);unsigned int stbl_box_size = big_to_small_endian_32bit(box_head.size);while((ftell(p)-stbl_pos+8) != stbl_box_size){fread(&box_head, sizeof(BOX_HEAD), 1, p);fseek(p, big_to_small_endian_32bit(box_head.size)-8, SEEK_CUR);switch(big_to_small_endian_32bit(box_head.type)){case STSD_BOX:{print_box_type(box_head.type, 5);}break;case STTS_BOX:{print_box_type(box_head.type, 5);}break;case STSC_BOX:{print_box_type(box_head.type, 5);}break;case STSZ_BOX:{print_box_type(box_head.type, 5);}break;case STCO_BOX:{print_box_type(box_head.type, 5);}break;case STSS_BOX:{print_box_type(box_head.type, 5);}break;}}       }break;}}}break;default:{fread(&minf_head_box_length, 4, 1, p);//nmhdprint_box_type(NMHD_BOX, 4);fseek(p, big_to_small_endian_32bit(minf_head_box_length)-4, SEEK_CUR);fread(&box_head, sizeof(BOX_HEAD), 1, p);while(big_to_small_endian_32bit(box_head.type) != STBL_BOX){print_box_type(box_head.type, 4);fseek(p, big_to_small_endian_32bit(box_head.size)-8, SEEK_CUR);fread(&box_head, sizeof(BOX_HEAD), 1, p);}print_box_type(box_head.type, 4);//stbl box}break;}if(stbl_flag == true){stbl_flag = false;free(mdhd_flag);break;}unsigned int cur_pos = ftell(p);unsigned int cur_box_size = big_to_small_endian_32bit(box_head.size);while((ftell(p)-cur_pos+8) != cur_box_size){fread(&box_head, sizeof(BOX_HEAD), 1, p);fseek(p, big_to_small_endian_32bit(box_head.size)-8, SEEK_CUR);switch(big_to_small_endian_32bit(box_head.type)){case STSD_BOX:{print_box_type(box_head.type, 5);}break;case STTS_BOX:{print_box_type(box_head.type, 5);}break;case STSC_BOX:{print_box_type(box_head.type, 5);}break;case STSZ_BOX:{print_box_type(box_head.type, 5);}break;case STCO_BOX:{print_box_type(box_head.type, 5);}break;case STSS_BOX:{print_box_type(box_head.type, 5);}break;case CTTS_BOX:{print_box_type(box_head.type, 5);}break;}}stbl_flag = false;free(mdhd_flag);}break;default:print_box_type(box_head.type, 2);fseek(p, big_to_small_endian_32bit(box_head.size)-8, SEEK_CUR);break;}/*读取到文件末尾跳出*/if(0 == fread(&box_head, sizeof(BOX_HEAD), 1, p)){break;}if(big_to_small_endian_32bit(box_head.type) == TRAK_BOX){print_box_type(box_head.type, 1);}}}free(flag);Ret_Boxhead_Fileptr ret;ret.box_head.size = box_head.size;ret.box_head.type = box_head.type;ret.p = p;return ret;
}FILE* mdat_box_parse(FILE *p, BOX_HEAD box_head)
{__int64 box_large_size;if(big_to_small_endian_32bit(box_head.size) != 1){fseek(p, big_to_small_endian_32bit(box_head.size)-8, SEEK_CUR);}else{fread(&box_large_size, sizeof(__int64), 1, p);fseek(p, big_to_small_endian_64bit(box_large_size)-sizeof(__int64)-8, SEEK_CUR);}return p;
}int main( void )
{FILE *p;FTYP_BOX ftyp_box;BOX_HEAD box_head;char filename[256];printf("please input mp4 filename:");scanf("%s", filename);if((p = fopen(filename,"r")) == NULL){printf("open fail\n");return -1;}printf("----------------------MP4 file structure----------------------\n");fread(&ftyp_box, sizeof(FTYP_BOX), 1, p);print_box_type(ftyp_box.ftyp_type, 0);//ftyp boxfseek(p, big_to_small_endian_32bit(ftyp_box.ftyp_size), SEEK_SET);fread(&box_head, sizeof(BOX_HEAD), 1, p);print_box_type(box_head.type, 0);//moov box OR mdat boxswitch(big_to_small_endian_32bit(box_head.type)){case MDAT_BOX:{p = mdat_box_parse(p, box_head);fread(&box_head, sizeof(BOX_HEAD), 1, p);print_box_type(box_head.type, 0);//moov boxmoov_box_parse(p);}break;case MOOV_BOX:{  Ret_Boxhead_Fileptr ret;ret = moov_box_parse(p);print_box_type(ret.box_head.type, 0);mdat_box_parse(ret.p, ret.box_head);}break;default:break;}fclose(p);printf("----------------------MP4 file structure----------------------\n\n\n");mainmenu();while(get_user_input() != 'q'){mainmenu();}return 0;
}

from: https://blog.csdn.net/chenshengfa/article/details/50512616