h.264視頻編碼加生成QT7可播放的mp4文件

實際上只是一個頭文件h264_util.h。h264_util.h封裝了x264庫。我做的工作就是使用模板封裝了一下，保證使用任何mp4 container庫都可以方便地輸出mp4文件。使用模板的目的是爲了避免任何回調的開銷。
h264_util.h依賴于x264庫。
main.cpp依賴于h264_util.h和libmp42（可選）和libmp4av（可選），後兩者都是mpeg4ip的一部分。使用的 mpeg4ip是爲了生成Quicktime7可播放的mp4文件（相信我，這點很難的，我試了很多庫了。原因不是這些庫不好，而是QT7太糟糕）。
附帶說明一下，所用到的庫和我寫的代碼都是可以同時在window下和linux編譯運行的（我已經都試過了）。要在windows下編譯x264很簡 單的，參考說明文檔就可以。麻煩的是mpeg4ip，其文檔聲稱不再會對windows下的兼容性負責了。要在windows下編譯並使用 mpeg4ip,有兩個辦法，或者自己提供gettimeofday的定義，或者使用其libmissing（和libmissing60是一回事，60 表示vs6下編譯通過，實際上我用vc7也編譯運行過了）。
如果同時使用x264和mpeg4ip，又會發現一些兼容性問題。問題的關鍵在于 windows操作下默認沒有提供int8_t之類的類型定義，于是mpeg4ip會自己提供一個mpeg4ip_version.h，補上這些定義。所 有的跨平台的開源項目似乎都會自說自話地爲windows平台（linux下有標准頭文件）提供這些定義，如果我同時使用兩個這樣的開源項目，就可能會産 生類型重複定義的編譯錯誤。x264是個特例，它的libx264根本就沒有提供這些定義，所以我在h264_util.h補上了這些定義。又由于這些類 型定義可能和mpeg4ip_version.h中的定義沖突，所以我在h264_util.h中預先加入了判斷避免沖突的宏。要這些宏起作 用，mpeg4ip的頭文件必須先被#include于我的h264_util.h之前。
要編譯x264,必須安裝nasm（一種彙編語言編譯器），mpeg4ip如果是只編譯上文中的兩個庫的話，不需要，否則也需要nasm。
下面就是源代碼，包括兩個源代碼文件，h264_util.h和main.cpp和一個makefile（僅用于linux），GNUmakefile。
h264_util.h封裝了x264，main.cpp提供了兩個測試例子，一個僅使用h264_util.h演示了如何進行視頻編碼，另一個測試例 子說明了如何生成mp4。test_h264_utilities是在內存中快速生成rgb888的raw image的工具，僅僅是爲了我測試方便用，你完全可以不使用它。
使用方法很簡單（僅限于于linux，windows下懶得寫了），安裝 libx264,libmp4ip,libmp4v2(./configure;make;make install總會吧？mpeg4ip安裝完整的軟件包也許會有問題，但是安裝庫是沒問題的（先在源代碼包的根目錄下./configure，然後還是在 對應的庫代碼的目錄裏make;make install），把下述三個文件拷貝到同一個目錄裏，然後make clean;make;./test_main,就會生成一個叫test.mp4的文件。
//h264_util.h
#ifndef __H264_UTIL_H__
#define __H264_UTIL_H__
#include <cstdio>
#include <cassert>
#include <climits>
#include <cstdlib>
#ifdef HAVE_MALLOC_H
#include <malloc.h>
#endif
#ifdef __cplusplus
extern "C" {
#endif
#if (defined HAVE_INTTYPES_H)
#include <inttypes.h>
#elif (defined HAVE_STDINT_H)
#include <stdint.h>
#else
#ifndef __MPEG4IP_WIN32_H__
typedef signed char int8_t;
typedef signed short int16_t;
typedef signed int int32_t;
typedef unsigned char uint8_t;
typedef unsigned short uint16_t;
typedef unsigned int uint32_t;
#if (defined WIN32)||(defined _WIN32)
typedef signed __int64 int64_t;
typedef unsigned __int64 uint64_t;
#else //other OS
typedef signed long long int64_t;
typedef unsigned long long uint64_t;
#endif
#endif
#endif
#include "x264.h"
#ifdef __cplusplus
}
#endif
struct ContainerTraits{
//insert+=return of write_nalu ; x264_nal_encode( _bit_buffer+inserter ......);
static int write_nalu(unsigned char *nalu, int size){ return 0;}
//encoded bit buffer, contains raw h.264 data
static void set_eop(unsigned char* bit_buffer, int size){ }
};
template <typename CTraits=ContainerTraits>
class h264_utilities
{
public:
h264_utilities(int width, int height):_h(0),_bit_buffer(0),_bit_buffer_capacity(1024*256)
{
assert(width%16==0);
assert(height%16==0);

x264_param_default(&_param);
_param.i_width = width;
_param.i_height= height;

_h=x264_encoder_open(&_param);
assert(_h);

//http://www.via.ecp.fr/via/ml/vlc-devel/2005-02/msg00371.html or google "x264_encoder_encode"

//if we do not allocate the new buffer ,
//the code "RGB_TO_I420( rgb_to_i420, 0, 1, 2, 3 );" will raise segment fault
x264_picture_alloc(&_pic,X264_CSP_RGB,_param.i_width ,_param.i_height);
// Do not force any parameters
_pic.i_type = X264_TYPE_AUTO;
_pic.i_qpplus1 = 0;

_bit_buffer_capacity=(_bit_buffer_capacity>_param.i_width*_param.i_height*4? _bit_buffer_capacity:_param.i_width*_param.i_height*4);
if (!_bit_buffer)
_bit_buffer = (unsigned char*)av_malloc(_bit_buffer_capacity);
}
~h264_utilities()
{
//assert(0);
if(_bit_buffer)
{
av_free(_bit_buffer);
_bit_buffer=0;
}

x264_encoder_close(_h);
}
//only accept row images (rgb888)
bool compress(unsigned char* frame)
{
memcpy(reinterpret_cast<void*>(_pic.img.plane[0]), reinterpret_cast<void*>(frame),_param.i_width*_param.i_height*3);
if(x264_encoder_encode(_h,&_nals,&_nnal,&_pic,&_pic_out)<0)
{
fprintf(stderr, "x264_encoder_encode failed\n" );
assert(0);
return false;
}
assert(_nnal);

int inserter=0;
for(int b = 0; b < _nnal; b++ )
{
int size;
if( ( size = x264_nal_encode( _bit_buffer+inserter, &_bit_buffer_capacity, 1, _nals+b ) ) < 0 )
{
fprintf( stderr, "need to increase buffer size (size=%d)\n", -size );
assert(0);
return false;
}

//nal unit start from
inserter+=CTraits::write_nalu(_bit_buffer+inserter,size);
}

if(_nnal)
CTraits::set_eop(_bit_buffer,inserter);

return true;
}
public:
x264_picture_t _pic;
x264_picture_t _pic_out;
x264_param_t _param;
x264_nal_t *_nals;
int _nnal;
x264_t *_h;

unsigned char * _bit_buffer;
int _bit_buffer_capacity;
public:
static void *av_malloc(unsigned int size)
{
static void *ptr;
#ifdef MEMALIGN_HACK
int diff;
#endif
/* lets disallow possible ambiguous cases */
if(size > INT_MAX)
return 0;

#ifdef MEMALIGN_HACK
ptr = malloc(size+16+1);
diff= ((-(int)ptr - 1)&15) + 1;
ptr += diff;
((char*)ptr)[-1]= diff;
#elif defined (HAVE_MEMALIGN)
ptr = memalign(16,size);
// Why 64? Indeed, we should align it:
// on 4 for 386
// on 16 for 486
// on 32 for 586, PPro - k6-III
// on 64 for K7 (maybe for P3 too).
// Because L1 and L2 caches are aligned on those values.
// But I don't want to code such logic here!

/* Why 16?
because some cpus need alignment, for example SSE2 on P4, & most RISC cpus
it will just trigger an exception and the unaligned load will be done in the
exception handler or it will just segfault (SSE2 on P4)
Why not larger? because i didnt see a difference in benchmarks ...
*/
/* benchmarks with p3
memalign(64)+1 3071,3051,3032
memalign(64)+2 3051,3032,3041
memalign(64)+4 2911,2896,2915
memalign(64)+8 2545,2554,2550
memalign(64)+16 2543,2572,2563
memalign(64)+32 2546,2545,2571
memalign(64)+64 2570,2533,2558

btw, malloc seems to do 8 byte alignment by default here
*/
#else
ptr = malloc(size);
#endif
return ptr;
}
/**
* av_realloc semantics (same as glibc): if ptr is 0 and size > 0,
* identical to malloc(size). If size is zero, it is identical to
* free(ptr) and 0 is returned.
*/
static void *av_realloc(void *ptr, unsigned int size)
{
#ifdef MEMALIGN_HACK
int diff;
#endif
/* lets disallow possible ambiguous cases */
if(size > INT_MAX)
return 0;
#ifdef MEMALIGN_HACK
//FIXME this isnt aligned correctly though it probably isnt needed
if(!ptr) return av_malloc(size);
diff= ((char*)ptr)[-1];
return realloc(ptr - diff, size + diff) + diff;
#else
return realloc(ptr, size);
#endif
}
/* NOTE: ptr = 0 is explicetly allowed */
static void av_free(void *ptr)
{
/* XXX: this test should not be needed on most libcs */
if (ptr)
#ifdef MEMALIGN_HACK
free(ptr - ((char*)ptr)[-1]);
#else
free(ptr);
#endif
}

public:
void debug_dump(void)
{
//debug parameters
fprintf(stderr,"dump parameters now:\n");
fprintf(stderr,"cpu=%d\n",_param.cpu);
fprintf(stderr,"i_threads=%d\n",_param.i_threads);
fprintf(stderr,"i_width=%d\n",_param.i_width);
fprintf(stderr,"i_height=%d\n",_param.i_height);
fprintf(stderr,"i_csp=%d\n",_param.i_csp);
fprintf(stderr,"i_level_idc=%d\n",_param.i_level_idc);
fprintf(stderr,"vui.i_sar_height=%d\n",_param.vui.i_sar_height);
fprintf(stderr,"vui.i_sar_width=%d\n",_param.vui.i_sar_width);
fprintf(stderr,"i_fps_num=%d\n",_param.i_fps_num);
fprintf(stderr,"i_fps_den=%d\n",_param.i_fps_den);
fprintf(stderr,"i_frame_reference=%d\n",_param.i_frame_reference);
fprintf(stderr,"i_keyint_max=%d\n",_param.i_keyint_max);
fprintf(stderr,"i_keyint_min=%d\n",_param.i_keyint_min);
fprintf(stderr,"i_scenecut_threshold=%d\n",_param.i_scenecut_threshold);
fprintf(stderr,"i_bframe=%d\n",_param.i_bframe);
fprintf(stderr,"psz_cqm_file=%s\n",_param.psz_cqm_file);
}
};
//prepare row images rgb888
class test_h264_utilities{
public:
test_h264_utilities(int w,int h,int f):WIDTH(w),HEIGHT(h),FRAMES(f),buf(0)
{
buf=(unsigned char*)malloc(HEIGHT*WIDTH*3*FRAMES);
produce_images();
dump_images();
}
~test_h264_utilities()
{
free(buf);
}
unsigned char* frame(int j)
{
return buf+j*HEIGHT*WIDTH*3;
}

private:
void produce_images()
{
//make a big picture
//the frames (* is red pixel, - is black pixel:
// frame1 frame2 frame3 frame4
// **-- --** ---- ----
// **-- --** ---- ----
// ---- ---- --** **--
// ---- ---- --** **--
for(int j=0;j<FRAMES;j++)
{
for(int y=0;y<HEIGHT;y++)
{
for(int x=0;x<WIDTH;x++)
{
switch(j%4)
{
case 0:
if(x<=WIDTH/2&&y<=HEIGHT/2)
{
//red pixel
frame(j)[y*WIDTH*3+x*3]=255;
}
break;
case 1:
if(x>WIDTH/2&&y<=HEIGHT/2)
{
//red pixel
frame(j)[y*WIDTH*3+x*3]=255;
}
break;
case 2:
if(x>WIDTH/2&&y>HEIGHT/2)
{
//red pixel
frame(j)[y*WIDTH*3+x*3]=255;
}
break;
case 3:
if(x<=WIDTH/2&&y>HEIGHT/2)
{
//red pixel
frame(j)[y*WIDTH*3+x*3]=255;
}
break;
default:
assert(0);
}//end of switch
}//end of for
} //end of for
} //end of for
}
void dump_images()
{
for(int j=0;j<FRAMES;j++)
{
//debug, before writing the real file, let's dump it on screen
fprintf(stdout,"image %d:\n",j);
for(int y=0;y<HEIGHT;y++)
{
for(int x=0;x<WIDTH;x++)
{
unsigned char* row=frame(j)+y*WIDTH*3;
if(row[x*3]==255)
{
fprintf(stdout,"%d",1);
}
else
{
fprintf(stdout,"%d",0);
}
}
fprintf(stdout,"\n");
}//end of for
}//end of for
}
public:
int WIDTH;
int HEIGHT;
int FRAMES;
unsigned char* buf;
};
#endif //__H264_UTIL_H__
//main.cpp
#include <assert.h>
#include <stdio.h>
#include <string.h>
#if 1
#include "mp4.h"
#include "mp4av_h264.h" //not included in mp4.h
//--------------missing declaration -------begin
//not in mp4av_h264.h or mp4av.h :(
typedef struct
{
struct
{
int size_min;
int next;
int cnt;
int idx[17];
int poc[17];
} dpb;
int cnt;
int cnt_max;
int *frame;
} h264_dpb_t;
//--------------missing declaration -------end
struct MP4Container
{
//interface
static int write_nalu(unsigned char* nalu, int size)
{
int32_t poc = 0;
unsigned char type=nalu[4] & 0x1f;

samplesWritten++;
switch(type)
{
//sps
case 0x7:
//ISO 14496-10 (Video), Advanced Video Coding (AVC), also known as H.264
memset(&h264_dec, 0, sizeof(h264_dec));
if (h264_read_seq_info(nalu, size, &h264_dec) == -1)
{
assert(0);
fprintf(stderr, "Could not decode Sequence header\n");
return 0;
}
trackId=MP4AddH264VideoTrack(mp4File,
Mp4TimeScale, //time scale,
mp4FrameDuration(),
h264_dec.pic_width,
h264_dec.pic_height,
nalu[5], //profile (baseline profile, main profile, etc. see
nalu[6], //compatible profile
nalu[7], //levels
3);
// doesn't get added to sample buffer
// remove header
MP4AddH264SequenceParameterSet(mp4File, trackId,
nalu + 4,
size - 4);
return 0;
break;
//pps
case 0x8:
// doesn't get added to sample buffer
MP4AddH264PictureParameterSet(mp4File, trackId,
nalu + 4,
size - 4);
return 0;
break;

//filler data
case 0xc:
// doesn't get copied
break;
//access unit
//sei
case 0x6:
case 0x9:
// note - may not want to copy this - not needed
default:
if(h264_nal_unit_type_is_slice(type))
{
//slice type
slice_is_idr = h264_dec.nal_unit_type == 0x5; //type idr slice
poc = h264_dec.pic_order_cnt;
nal_is_sync = h264_slice_is_idr(&h264_dec);
}
nalu[0] = ((size-4) >> 24) & 0xff;
nalu[1] = ((size-4) >> 16) & 0xff;
nalu[2] = ((size-4) >> 8) & 0xff;
nalu[3] = ((size-4) >> 0) & 0xff;

return size;
}//end of switch

return 0;
}
static void set_eop(unsigned char* bit_buffer, int size)
{
samplesWritten++;
double thiscalc;
thiscalc = samplesWritten;
thiscalc *= Mp4TimeScale;
thiscalc /= VideoFrameRate;

thisTime = (MP4Duration)thiscalc;
MP4Duration dur;
dur = thisTime - lastTime;
if(!MP4WriteSample(mp4File,trackId,bit_buffer,size,dur,0,nal_is_sync))
{
fprintf(stderr,"can't write video frame\n");
MP4DeleteTrack(mp4File, trackId);
return;
}
lastTime = thisTime;
DpbAdd(&h264_dpb, h264_dec.pic_order_cnt, slice_is_idr);
return;

DpbFlush(&h264_dpb);
if (h264_dpb.dpb.size_min > 0)
{
unsigned int ix;
for (ix = 0; ix < samplesWritten; ix++)
{;
const int offset = DpbFrameOffset(&h264_dpb, ix);
//fprintf( stderr, "dts=%d pts=%d offset=%d\n", ix, ix+offset, offset );
MP4SetSampleRenderingOffset(mp4File, trackId, 1 + ix, offset*mp4FrameDuration());
}
}
DpbClean(&h264_dpb);
}
//not interface, can be deleled safely
static MP4FileHandle mp4File;
//consts
static uint32_t Mp4TimeScale;
static double VideoFrameRate;
static uint32_t mp4FrameDuration()
{
return (u_int32_t)(((double)Mp4TimeScale) / VideoFrameRate);
}
static MP4SampleId samplesWritten;
static uint8_t *nal_buffer;
static uint32_t nal_buffer_size;
static uint32_t nal_buffer_size_max;
static MP4Timestamp lastTime;
static MP4Timestamp thisTime;
static MP4TrackId trackId;
static bool nal_is_sync;
static h264_dpb_t h264_dpb;
static h264_decode_t h264_dec;
static bool slice_is_idr;
//some tools
static void DpbInit( h264_dpb_t *p )
{
p->dpb.cnt = 0;
p->dpb.next = 0;
p->dpb.size_min = 0;
p->cnt = 0;
p->cnt_max = 0;
p->frame = NULL;
}
static void DpbClean( h264_dpb_t *p )
{
free( p->frame );
}
static void DpbUpdate( h264_dpb_t *p, int is_forced )
{
int i;
int pos;
if (!is_forced && p->dpb.cnt < 16)
return;
/* find the lowest poc */
pos = 0;
for (i = 1; i < p->dpb.cnt; i++)
{
if (p->dpb.poc[i] < p->dpb.poc[pos])
pos = i;
}
//fprintf( stderr, "lowest=%d\n", pos );
/* save the idx */
if (p->dpb.idx[pos] >= p->cnt_max)
{
int inc = 1000 + (p->dpb.idx[pos]-p->cnt_max);
p->cnt_max += inc;
p->frame = (int*)realloc( p->frame, sizeof(int)*p->cnt_max );
for (i=0;i<inc;i++)
p->frame[p->cnt_max-inc+i] = -1; /* To detect errors latter */
}
p->frame[p->dpb.idx[pos]] = p->cnt++;
/* Update the dpb minimal size */
if (pos > p->dpb.size_min)
p->dpb.size_min = pos;
/* update dpb */
for (i = pos; i < p->dpb.cnt-1; i++)
{
p->dpb.idx[i] = p->dpb.idx[i+1];
p->dpb.poc[i] = p->dpb.poc[i+1];
}
p->dpb.cnt--;
}
static void DpbFlush( h264_dpb_t *p )
{
while (p->dpb.cnt > 0)
DpbUpdate( p, true );
}
static void DpbAdd( h264_dpb_t *p, int poc, int is_idr )
{
if (is_idr)
DpbFlush( p );
p->dpb.idx[p->dpb.cnt] = p->dpb.next;
p->dpb.poc[p->dpb.cnt] = poc;
p->dpb.cnt++;
p->dpb.next++;

DpbUpdate( p, false );
}
static int DpbFrameOffset( h264_dpb_t *p, int idx )
{
if (idx >= p->cnt)
return 0;
if (p->frame[idx] < 0)
return p->dpb.size_min; /* We have an error (probably broken/truncated bitstream) */
return p->dpb.size_min + p->frame[idx] - idx;
}
};
MP4FileHandle MP4Container::mp4File=0;
uint32_t MP4Container::Mp4TimeScale=90000;
//http://www.via.ecp.fr/via/ml/x264-devel/2005-07/msg00056.html
//I think we need to get the latest x264 library to test it
double MP4Container::VideoFrameRate=25*2;
MP4SampleId MP4Container::samplesWritten=0;
uint8_t* MP4Container::nal_buffer=0;
uint32_t MP4Container::nal_buffer_size=0;
uint32_t MP4Container::nal_buffer_size_max=0;
MP4Timestamp MP4Container::lastTime;
MP4Timestamp MP4Container::thisTime;
MP4TrackId MP4Container::trackId;
bool MP4Container::nal_is_sync=false;
h264_dpb_t MP4Container::h264_dpb;
h264_decode_t MP4Container::h264_dec;
bool MP4Container::slice_is_idr;
#endif
//h264_util.h must be located after the mpeg4ip's headers, because those headers typedef its own int_8
#include "h264_util.h"
int main(void)
{
#if 1
{
//width, heigth, frames
int width=192;
int height=128;
int frames=25;
test_h264_utilities images(width,height,frames);
MP4FileHandle mp4File=MP4Create("7.mp4",1,0);
if (mp4File)
{
MP4SetTimeScale(mp4File, MP4Container::Mp4TimeScale);
}

MP4Container::mp4File=mp4File;

h264_utilities<MP4Container> h264(width,height);

for(int i=0;i<images.FRAMES;i++)
{
h264.compress(images.frame(i));
}

MP4Close(mp4File);
}
#endif //testcase1
#if 1
{
//width, height,frames
test_h264_utilities images(16,16,4);
h264_utilities<> h264(16,16);
for(int i=0;i<images.FRAMES;i++)
{
h264.compress(images.frame(i));
}
}
#endif //testcase2, test x264 (encapsulateed in a h264_util.h)
return 0;
}
//GNUmakefile
#to get release version , leave $(IDL_DEBUG) empty
IDL_DEBUG=hi
DEVELOP_ROOT := $(shell echo `cd ../../;pwd`)
#---------------------for libquicktime------------------begin
LIBX264=#$(DEVELOP_ROOT)/x264
INC_DIRS :=#$(LIBX264)/include
LIBS =mp4v2 mp4av x264
LIB_DIR =
STATIC_LIBS =#$(LIBX264)/lib/libx264.la
SHARED_LIBS =
SHARED_LIB_DIR =
#---------------------for libquicktime------------------end
OPT =-D_DEBUG -g -O0 -Wall
CC =gcc
CXX =g++
CPPFLAGS +=$(OPT) $(addprefix -I,$(INC_DIRS))
ifdef SHARED_LIB_DIR
LDFLAGS +=-Wl,-rpath,$(SHARED_LIB_DIR) -L$(SHARED_LIB_DIR)
endif
LDFLAGS +=$(addprefix -L,$(LIB_DIR))
LDFLAGS +=$(addprefix -l,$(SHARED_LIBS)) $(addprefix -l,$(LIBS))
ifdef IDL_DEBUG
endif
SOURCES :=$(wildcard *.c) $(wildcard *.cc) $(wildcard *.cxx) $(wildcard *.cpp)
OBJS = $(addsuffix .o, $(basename $(SOURCES)))
.PHONY : all clean mkdir targets
all : targets
# create the deployment directory if it does not yet exist
DEPLOYDIR = .
mkdir :
@mkdir -p $(DEPLOYDIR)
TARGET = $(DEPLOYDIR)/test_main
$(TARGET) : $(OBJS)
libtool $(CXX) $(LDFLAGS) $^ $(STATIC_LIBS) -o $@
targets : mkdir $(TARGET)
clean :
-$(RM) $(OBJS) $(OBJS:.o=.d) core* *.bak $(TARGET) *.mov *.mp4 *.d *.mpg *.mpeg
# common block
# add auto dependency generation to default rule
COMPILE.c += -MMD
COMPILE.cc += -MMD
DFILES := $(OBJS:.o=.d)
ifneq (${MAKECMDGOALS}, clean)
-include $(DFILES)
endif
# end of common block