In my C++ application, I am taking a series of JPEG images, manipulating their data using FreeImage, and then encoding the bitmaps as H264 using the ffmpeg/libx264 C API. The output is an MP4 which shows the series of 22 images at 12fps. My code is adapted from the "muxing" example that comes with ffmpeg C source code.
My problem: no matter how I tune the codec parameters, a certain number of frames at the end of the sequence which are passed to the encoder do not appear in the final output. I've set the AVCodecContext parameters like this:
//set context params
ctx->codec_id = AV_CODEC_ID_H264;
ctx->bit_rate = 4000 * 1000;
ctx->width = _width;
ctx->height = _height;
ost->st->time_base = AVRational{ 1, 12 };
ctx->time_base = ost->st->time_base;
ctx->gop_size = 1;
ctx->pix_fmt = AV_PIX_FMT_YUV420P;
I have found that the higher the gop_size the more frames are dropped from the end of the video. I can also see from the output that, with this gop size (where I'm essentially directing that all output frames be I frames) that only 9 frames are written.
I'm not sure why this is occurring. I experimented with encoding duplicate frames and making a much longer video. This resulted in no frames being dropped. I know with the ffmpeg command line tool there is a concatenation command that accomplishes what I am trying to do, but I'm not sure how to accomplish the same goal using the C API.
Here's the output I'm getting from the console:
[libx264 @ 026d81c0] using cpu capabilities: MMX2 SSE2Fast SSSE3 SSE4.2 AVX FMA3 BMI2 AVX2 [libx264 @ 026d81c0] profile High, level 3.1 [libx264 @ 026d81c0] 264 - core 152 r2851 ba24899 - H.264/MPEG-4 AVC codec - Cop yleft 2003-2017 - http://www.videolan.org/x264.html - options: cabac=1 ref=1 deb lock=1:0:0 analyse=0x3:0x113 me=hex subme=7 psy=1 psy_rd=1.00:0.00 mixed_ref=0 m e_range=16 chroma_me=1 trellis=1 8x8dct=1 cqm=0 deadzone=21,11 fast_pskip=1 chro ma_qp_offset=-2 threads=12 lookahead_threads=2 sliced_threads=0 nr=0 decimate=1 interlaced=0 bluray_compat=0 constrained_intra=0 bframes=0 weightp=0 keyint=1 ke yint_min=1 scenecut=40 intra_refresh=0 rc=abr mbtree=0 bitrate=4000 ratetol=1.0 qcomp=0.60 qpmin=0 qpmax=69 qpstep=4 ip_ratio=1.40 aq=1:1.00 Output #0, mp4, to '....\images\c411a991-46f6-400c-8bb0-77af3738559a.mp4': Stream #0:0: Video: h264, yuv420p, 700x700, q=2-31, 4000 kb/s, 12 tbn
[libx264 @ 026d81c0] frame I:9 Avg QP:17.83 size:111058 [libx264 @ 026d81c0] mb I I16..4: 1.9% 47.7% 50.5% [libx264 @ 026d81c0] final ratefactor: 19.14 [libx264 @ 026d81c0] 8x8 transform intra:47.7% [libx264 @ 026d81c0] coded y,uvDC,uvAC intra: 98.4% 96.9% 89.5% [libx264 @ 026d81c0] i16 v,h,dc,p: 64% 6% 2% 28% [libx264 @ 026d81c0] i8 v,h,dc,ddl,ddr,vr,hd,vl,hu: 32% 15% 9% 5% 5% 6% 8% 10% 10% [libx264 @ 026d81c0] i4 v,h,dc,ddl,ddr,vr,hd,vl,hu: 28% 18% 7% 6% 8% 8% 8% 9% 8% [libx264 @ 026d81c0] i8c dc,h,v,p: 43% 22% 25% 10% [libx264 @ 026d81c0] kb/s:10661.53
Code included below:
MP4Writer.h
#ifndef MPEG_WRITER
#define MPEG_WRITER
#include <iostream>
#include <string>
#include <vector>
#include <ImgData.h>
extern "C" {
#include <libavformat/avformat.h>
#include <libswscale/swscale.h>
#include <libswresample/swresample.h>
#include <libswscale/swscale.h>
}
typedef struct OutputStream
{
AVStream *st;
AVCodecContext *enc;
//pts of the next frame that will be generated
int64_t next_pts;
int samples_count;
AVFrame *frame;
AVFrame *tmp_frame;
float t, tincr, tincr2;
struct SwsContext *sws_ctx;
struct SwrContext *swr_ctx;
};
class MP4Writer {
public:
MP4Writer();
void Init();
int16_t SetOutput( const std::string & path );
int16_t AddFrame( uint8_t * imgData );
int16_t Write( std::vector<ImgData> & imgData );
int16_t Finalize();
void SetHeight( const int height ) { _height = _width = height; } //assuming 1:1 aspect ratio
private:
int16_t AddStream( OutputStream * ost, AVFormatContext * formatCtx, AVCodec ** codec, enum AVCodecID codecId );
int16_t OpenVideo( AVFormatContext * formatCtx, AVCodec *codec, OutputStream * ost, AVDictionary * optArg );
static AVFrame * AllocPicture( enum AVPixelFormat pixFmt, int width, int height );
static AVFrame * GetVideoFrame( uint8_t * imgData, OutputStream * ost, const int width, const int height );
static int WriteFrame( AVFormatContext * formatCtx, const AVRational * timeBase, AVStream * stream, AVPacket * packet );
int _width;
int _height;
OutputStream _ost;
AVFormatContext * _formatCtx;
AVDictionary * _dict;
};
#endif //MPEG_WRITER
MP4Writer.cpp
#include <MP4Writer.h>
#include <algorithm>
MP4Writer::MP4Writer()
{
_width = 0;
_height = 0;
}
void MP4Writer::Init()
{
av_register_all();
}
/**
sets up output stream for the specified path.
note that the output format is deduced automatically from the file extension passed
@param path: output file path
@returns: -1 = output could not be deduced, -2 = invalid codec, -3 = error opening output file,
-4 = error writing header
*/
int16_t MP4Writer::SetOutput( const std::string & path )
{
int error;
AVCodec * codec;
AVOutputFormat * format;
_ost = OutputStream{}; //TODO reset state in a more focused way?
//allocate output media context
avformat_alloc_output_context2( &_formatCtx, NULL, NULL, path.c_str() );
if ( !_formatCtx ) {
std::cout << "could not deduce output format from file extension. aborting" << std::endl;
return -1;
}
//set format
format = _formatCtx->oformat;
if ( format->video_codec != AV_CODEC_ID_NONE ) {
AddStream( &_ost, _formatCtx, &codec, format->video_codec );
}
else {
std::cout << "there is no video codec set. aborting" << std::endl;
return -2;
}
OpenVideo( _formatCtx, codec, &_ost, _dict );
av_dump_format( _formatCtx, 0, path.c_str(), 1 );
//open output file
if ( !( format->flags & AVFMT_NOFILE )) {
error = avio_open( &_formatCtx->pb, path.c_str(), AVIO_FLAG_WRITE );
if ( error < 0 ) {
std::cout << "there was an error opening output file " << path << ". aborting" << std::endl;
return -3;
}
}
//write header
error = avformat_write_header( _formatCtx, &_dict );
if ( error < 0 ) {
std::cout << "an error occurred writing header. aborting" << std::endl;
return -4;
}
return 0;
}
/**
initialize the output stream
@param ost: the output stream
@param formatCtx: the context format
@param codec: the output codec
@param codec: the ffmpeg enumerated id of the codec
@returns: -1 = encoder not found, -2 = stream could not be allocated, -3 = encoding context could not be allocated
*/
int16_t MP4Writer::AddStream( OutputStream * ost, AVFormatContext * formatCtx, AVCodec ** codec, enum AVCodecID codecId )
{
AVCodecContext * ctx; //TODO not sure why this is here, could just set ost->enc directly
int i;
//detect the encoder
*codec = avcodec_find_encoder( codecId );
if ( (*codec) == NULL ) {
std::cout << "could not find encoder. aborting" << std::endl;
return -1;
}
//allocate stream
ost->st = avformat_new_stream( formatCtx, NULL );
if ( ost->st == NULL ) {
std::cout << "could not allocate stream. aborting" << std::endl;
return -2;
}
//allocate encoding context
ost->st->id = formatCtx->nb_streams - 1;
ctx = avcodec_alloc_context3( *codec );
if ( ctx == NULL ) {
std::cout << "could not allocate encoding context. aborting" << std::endl;
return -3;
}
ost->enc = ctx;
//set context params
ctx->codec_id = AV_CODEC_ID_H264;
ctx->bit_rate = 4000 * 1000;
ctx->width = _width;
ctx->height = _height;
ost->st->time_base = AVRational{ 1, 12 };
ctx->time_base = ost->st->time_base;
ctx->gop_size = 1;
ctx->pix_fmt = AV_PIX_FMT_YUV420P;
//if neccesary, set stream headers and formats separately
if ( formatCtx->oformat->flags & AVFMT_GLOBALHEADER ) {
std::cout << "setting stream and headers to be separate" << std::endl;
ctx->flags |= AV_CODEC_FLAG_GLOBAL_HEADER;
}
return 0;
}
/**
open the video for writing
@param formatCtx: the format context
@param codec: output codec
@param ost: output stream
@param optArg: dictionary
@return: -1 = error opening codec, -2 = allocate new frame, -3 = copy stream params
*/
int16_t MP4Writer::OpenVideo( AVFormatContext * formatCtx, AVCodec *codec, OutputStream * ost, AVDictionary * optArg )
{
int error;
AVCodecContext * ctx = ost->enc;
AVDictionary * dict = NULL;
av_dict_copy( &dict, optArg, 0 );
//open codec
error = avcodec_open2( ctx, codec, &dict );
av_dict_free( &dict );
if ( error < 0 ) {
std::cout << "there was an error opening the codec. aborting" << std::endl;
return -1;
}
//allocate new frame
ost->frame = AllocPicture( ctx->pix_fmt, ctx->width, ctx->height );
if ( ost->frame == NULL ) {
std::cout << "there was an error allocating a new frame. aborting" << std::endl;
return -2;
}
//copy steam params
error = avcodec_parameters_from_context( ost->st->codecpar, ctx );
if ( error < 0 ) {
std::cout << "could not copy stream parameters. aborting" << std::endl;
return -3;
}
return 0;
}
/**
allocate a new frame
@param pixFmt: ffmpeg enumerated pixel format
@param width: output width
@param height: output height
@returns: an inititalized frame
*/
AVFrame * MP4Writer::AllocPicture( enum AVPixelFormat pixFmt, int width, int height )
{
AVFrame * picture;
int error;
//allocate the frame
picture = av_frame_alloc();
if ( picture == NULL ) {
std::cout << "there was an error allocating the picture" << std::endl;
return NULL;
}
picture->format = pixFmt;
picture->width = width;
picture->height = height;
//allocate the frame's data buffer
error = av_frame_get_buffer( picture, 32 );
if ( error < 0 ) {
std::cout << "could not allocate frame data" << std::endl;
return NULL;
}
picture->pts = 0;
return picture;
}
/**
convert raw RGB buffer to YUV frame
@return: frame that contains image data
*/
AVFrame * MP4Writer::GetVideoFrame( uint8_t * imgData, OutputStream * ost, const int width, const int height )
{
int error;
AVCodecContext * ctx = ost->enc;
//prepare the frame
error = av_frame_make_writable( ost->frame );
if ( error < 0 ) {
std::cout << "could not make frame writeable" << std::endl;
return NULL;
}
//TODO set this context one time per run, or even better, one time at init
//convert RGB to YUV
struct SwsContext* fooContext = sws_getContext( width, height, AV_PIX_FMT_BGR24,
width, height, AV_PIX_FMT_YUV420P, SWS_BICUBIC, NULL, NULL, NULL );
int inLinesize[1] = { 3 * width }; // RGB stride
uint8_t * inData[1] = { imgData };
int sliceHeight = sws_scale( fooContext, inData, inLinesize, 0, height, ost->frame->data, ost->frame->linesize );
sws_freeContext( fooContext );
ost->frame->pts = ost->next_pts++;
//TODO does the frame need to be returned here as it is available at the class level?
return ost->frame;
}
/**
write frame to file
@param formatCtx: the output format context
@param timeBase: the framerate
@param stream: output stream
@param packet: data packet
@returns: see return values for av_interleaved_write_frame
*/
int MP4Writer::WriteFrame( AVFormatContext * formatCtx, const AVRational * timeBase, AVStream * stream, AVPacket * packet )
{
av_packet_rescale_ts( packet, *timeBase, stream->time_base );
packet->stream_index = stream->index;
//write compressed file to media file
return av_interleaved_write_frame( formatCtx, packet );
}
int16_t MP4Writer::Write( std::vector<ImgData> & imgData )
{
int16_t errorCount = 0;
int16_t retVal = 0;
bool countingUp = true;
size_t i = 0;
while ( true ) {
//don't show first frame again when counting back down
if ( !countingUp && i == 0 ) {
break;
}
uint8_t * pixels = imgData[i].GetBits( imgData[i].mp4Input );
AddFrame( pixels );
//handle inc/dec without repeating last frame
if ( countingUp ) {
if ( i == imgData.size() -1 ) {
countingUp = false;
i--;
}
else {
i++;
}
}
else {
i--;
}
}
Finalize();
return 0; //TODO return error code
}
/**
add another frame to output video
@param imgData: the raw image data
@returns -1 = error encoding video frame, -2 = error writing frame
*/
int16_t MP4Writer::AddFrame( uint8_t * imgData )
{
int error;
AVCodecContext * ctx;
AVFrame * frame;
int gotPacket = 0;
AVPacket pkt = { 0 };
ctx = _ost.enc;
av_init_packet( &pkt );
frame = GetVideoFrame( imgData, &_ost, _width, _height );
//encode the image
error = avcodec_encode_video2( ctx, &pkt, frame, &gotPacket );
if ( error < 0 ) {
std::cout << "there was an error encoding the video frame" << std::endl;
return -1;
}
//write the frame. NOTE: this doesn't kick in until the encoder has received a certain number of frames
if ( gotPacket ) {
error = WriteFrame( _formatCtx, &ctx->time_base, _ost.st, &pkt );
if ( error < 0 ) {
std::cout << "the video frame could not be written" << std::endl;
return -2;
}
}
return 0;
}
/**
finalize output video and cleanup
*/
int16_t MP4Writer::Finalize()
{
av_write_trailer( _formatCtx );
avcodec_free_context( &_ost.enc );
av_frame_free( &_ost.frame);
av_frame_free( &_ost.tmp_frame );
avio_closep( &_formatCtx->pb );
avformat_free_context( _formatCtx );
sws_freeContext( _ost.sws_ctx );
swr_free( &_ost.swr_ctx);
return 0;
}
usage
#include <FreeImage.h>
#include <MP4Writer.h>
#include <vector>
struct ImgData
{
unsigned int width;
unsigned int height;
std::string path;
FIBITMAP * mp4Input;
uint8_t * GetBits( FIBITMAP * bmp ) { return FreeImage_GetBits( bmp ); }
};
int main()
{
std::vector<ImgData> imgDataVec;
//load images and push to imgDataVec
MP4Writer mp4Writer;
mp4Writer.SetHeight( 1200 ); //assumes 1:1 aspect ratio
mp4Writer.Init();
mp4Writer.SetOutput( "test.mp4" );
mp4Writer.Write( imgDataVec );
}
