[ffmpeg系列 03] 文件、流地址（视频）解码为YUV

一  代码

ffmpeg版本5.1.2，dll是：ffmpeg-5.1.2-full_build-shared。x64的。

文件、流地址对使用者来说是一样。
流地址(RTMP、HTTP-FLV、RTSP等)：信令完成后，才进行音视频传输。信令包括音视频格式、参数等协商。

接流的在实际中的应用：1 展示，播放。2 给算法用，一般是需要RGB格式的。

#ifndef _DECODE_H264_H_
#define _DECODE_H264_H_
#include <string>extern  "C"
{
#include "libavformat/avformat.h"
#include "libavcodec/avcodec.h"
#include "libswscale/swscale.h"
#include "libavutil/avutil.h"
#include "libavutil/mathematics.h"
#include "libavutil/time.h"
#include "libavutil/pixdesc.h"
#include "libavutil/display.h"
};#pragma  comment(lib, "avformat.lib")
#pragma  comment(lib, "avutil.lib")
#pragma  comment(lib, "avcodec.lib")
#pragma  comment(lib, "swscale.lib")class CDecodeH264
{
public:CDecodeH264();~CDecodeH264();public:public:int DecodeH264();int  Start();int  Close();int  DecodeH264File_Init();int  ReleaseDecode();void H264Decode_Thread_Fun();std::string  dup_wchar_to_utf8(const wchar_t* wstr);double get_rotation(AVStream *st);public:AVFormatContext*       m_pInputFormatCtx = nullptr;AVCodecContext*        m_pVideoDecodeCodecCtx = nullptr;const AVCodec*           m_pCodec = nullptr;	SwsContext*                m_pSwsContext = nullptr;AVFrame*                    m_pFrameScale = nullptr;AVFrame*                    m_pFrameYUV = nullptr;AVPacket*                   m_pAVPacket = nullptr;enum AVMediaType		m_CodecType;int                                    m_output_pix_fmt;int                                    m_nVideoStream = -1;int                                    m_nFrameHeight = 0;int                                    m_nFrameWidth = 0;int                                    m_nFPS;int                                    m_nVideoSeconds;FILE*         m_pfOutYUV = nullptr;FILE*         m_pfOutYUV2 = nullptr;
};
#endif

#include "DecodeH264.h"
#include <thread>
#include <functional>
#include <codecvt>
#include <locale>char av_error2[AV_ERROR_MAX_STRING_SIZE] = { 0 };
#define av_err2str2(errnum) av_make_error_string(av_error2, AV_ERROR_MAX_STRING_SIZE, errnum)CDecodeH264::CDecodeH264()
{	
}CDecodeH264::~CDecodeH264()
{ReleaseDecode();
}std::string  CDecodeH264::dup_wchar_to_utf8(const wchar_t* wstr)
{std::wstring_convert<std::codecvt_utf8<wchar_t>> converter;return converter.to_bytes(wstr);
}//Side data : 
//displaymatrix: rotation of - 90.00 degrees
double CDecodeH264::get_rotation(AVStream *st)
{uint8_t* displaymatrix = av_stream_get_side_data(st,AV_PKT_DATA_DISPLAYMATRIX, NULL);double theta = 0;if (displaymatrix)theta = -av_display_rotation_get((int32_t*)displaymatrix);theta -= 360 * floor(theta / 360 + 0.9 / 360);if (fabs(theta - 90 * round(theta / 90)) > 2)av_log(NULL, AV_LOG_WARNING, "Odd rotation angle.\n""If you want to help, upload a sample ""of this file to https://streams.videolan.org/upload/ ""and contact the ffmpeg-devel mailing list. (ffmpeg-devel@ffmpeg.org)");return theta;
}int CDecodeH264::DecodeH264File_Init()
{avformat_network_init(); //流地址需要m_pInputFormatCtx = avformat_alloc_context();std::string strFilename = dup_wchar_to_utf8(L"测试.h264");//std::string strFilename = dup_wchar_to_utf8(L"rtmp://127.0.0.1/live/now");	int ret = avformat_open_input(&m_pInputFormatCtx, strFilename.c_str(), nullptr, nullptr);if (ret != 0) {char* err_str = av_err2str2(ret);printf("fail to open filename: %s, return value: %d,  %s\n", strFilename.c_str(), ret, err_str);return -1;}ret = avformat_find_stream_info(m_pInputFormatCtx, nullptr);if (ret < 0) {char* err_str = av_err2str2(ret);printf("fail to get stream information: %d, %s\n", ret, err_str);		return -1;}for (int i = 0; i < m_pInputFormatCtx->nb_streams; ++i) {const AVStream* stream = m_pInputFormatCtx->streams[i];if (stream->codecpar->codec_type == AVMEDIA_TYPE_VIDEO) {m_nVideoStream = i;printf("type of the encoded data: %d, dimensions of the video frame in pixels: width: %d, height: %d, pixel format: %d\n",stream->codecpar->codec_id, stream->codecpar->width, stream->codecpar->height, stream->codecpar->format);}}if (m_nVideoStream == -1) {printf("no video stream\n");return -1;}printf("m_nVideoStream=%d\n", m_nVideoStream);//获取旋转角度double theta = get_rotation(m_pInputFormatCtx->streams[m_nVideoStream]);m_pVideoDecodeCodecCtx = avcodec_alloc_context3(nullptr);avcodec_parameters_to_context(m_pVideoDecodeCodecCtx,\m_pInputFormatCtx->streams[m_nVideoStream]->codecpar);m_pCodec = avcodec_find_decoder(m_pVideoDecodeCodecCtx->codec_id);if (m_pCodec == nullptr){return -1;}m_nFrameHeight = m_pVideoDecodeCodecCtx->height;m_nFrameWidth = m_pVideoDecodeCodecCtx->width;printf("w=%d h=%d\n", m_pVideoDecodeCodecCtx->width, m_pVideoDecodeCodecCtx->height);if (avcodec_open2(m_pVideoDecodeCodecCtx, m_pCodec, nullptr) < 0){return -1;}//读文件知道视频宽高m_output_pix_fmt = AV_PIX_FMT_YUV420P; //AV_PIX_FMT_NV12;m_pSwsContext = sws_getContext(m_pVideoDecodeCodecCtx->width, m_pVideoDecodeCodecCtx->height,m_pVideoDecodeCodecCtx->pix_fmt, m_pVideoDecodeCodecCtx->width, m_pVideoDecodeCodecCtx->height,(AVPixelFormat)m_output_pix_fmt, SWS_FAST_BILINEAR, nullptr, nullptr, nullptr);//解码后的视频数据m_pFrameScale = av_frame_alloc();m_pFrameScale->format = m_output_pix_fmt;m_pFrameYUV = av_frame_alloc();m_pFrameYUV->format = m_output_pix_fmt; //mAVFrame.format is not setm_pFrameYUV->width = m_pVideoDecodeCodecCtx->width;m_pFrameYUV->height = m_pVideoDecodeCodecCtx->height;printf("m_pFrameYUV pix_fmt=%d\n", m_pVideoDecodeCodecCtx->pix_fmt);av_frame_get_buffer(m_pFrameYUV, 64);char cYUVName[256];sprintf_s(cYUVName, "%d_%d_%s.yuv", m_nFrameWidth, m_nFrameHeight, av_get_pix_fmt_name(m_pVideoDecodeCodecCtx->pix_fmt));fopen_s(&m_pfOutYUV, cYUVName, "wb");char cYUVName2[256];sprintf_s(cYUVName2, "%d_%d_%s_2.yuv", m_nFrameWidth, m_nFrameHeight, av_get_pix_fmt_name(m_pVideoDecodeCodecCtx->pix_fmt));fopen_s(&m_pfOutYUV2, cYUVName2, "wb");printf("leave init\n");return   0;
}void CDecodeH264::H264Decode_Thread_Fun()
{int   nFrameFinished = 0;int i = 0;int  ret;m_pAVPacket = av_packet_alloc();while (true) {ret = av_read_frame(m_pInputFormatCtx, m_pAVPacket);if (ret == AVERROR(EAGAIN) || ret == AVERROR_EOF) {av_packet_unref(m_pAVPacket);printf("read_frame break");break;}if (m_pAVPacket->stream_index == m_nVideoStream){int  send_packet_ret = avcodec_send_packet(m_pVideoDecodeCodecCtx, m_pAVPacket);printf("encode video send_packet_ret %d\n", send_packet_ret);int receive_frame_ret = avcodec_receive_frame(m_pVideoDecodeCodecCtx, m_pFrameScale);char* err_str = av_err2str2(receive_frame_ret);printf("frame w=%d, h=%d, linesize[0]=%d, linesize[1]=%d\n", m_pFrameScale->width, m_pFrameScale->height, m_pFrameScale->linesize[0], m_pFrameScale->linesize[1]);if (receive_frame_ret == 0){++i;int iReturn = sws_scale(m_pSwsContext, m_pFrameScale->data,m_pFrameScale->linesize, 0, m_nFrameHeight,m_pFrameYUV->data, m_pFrameYUV->linesize);printf("frame w=%d, h=%d, linesize[0]=%d, linesize[1]=%d\n", m_pFrameYUV->width, m_pFrameYUV->height, m_pFrameYUV->linesize[0], m_pFrameYUV->linesize[1]);/*if (0 != iReturn){fwrite(m_pFrameYUV->data[0], 1, m_nFrameWidth * m_nFrameHeight, m_pfOutYUV);fwrite(m_pFrameYUV->data[1], 1, m_nFrameWidth * m_nFrameHeight /4, m_pfOutYUV);fwrite(m_pFrameYUV->data[2], 1, m_nFrameWidth * m_nFrameHeight /4, m_pfOutYUV);}*///用linesize更能兼容特殊的宽if (0 != iReturn){					for (int i = 0; i < m_nFrameHeight; ++i) {fwrite(m_pFrameYUV->data[0] + i * m_pFrameYUV->linesize[0], 1, m_nFrameWidth, m_pfOutYUV2);}for (int i = 0; i < m_nFrameHeight / 2; ++i) {fwrite(m_pFrameYUV->data[1] + i * m_pFrameYUV->linesize[1], 1, m_nFrameWidth / 2, m_pfOutYUV2);}for (int i = 0; i < m_nFrameHeight / 2; ++i) {fwrite(m_pFrameYUV->data[2] + i * m_pFrameYUV->linesize[2], 1, m_nFrameWidth / 2, m_pfOutYUV2);}}				}}av_packet_unref(m_pAVPacket);}
}int CDecodeH264::DecodeH264()
{if (DecodeH264File_Init() != 0){return   -1;}auto video_func = std::bind(&CDecodeH264::H264Decode_Thread_Fun, this);std::thread  video_thread(video_func);video_thread.join(); return 0;
}int CDecodeH264::Start()
{DecodeH264();return  1;
}int CDecodeH264::Close()
{return   0;
}int  CDecodeH264::ReleaseDecode()
{if (m_pSwsContext){sws_freeContext(m_pSwsContext);m_pSwsContext = nullptr;}if (m_pFrameScale){av_frame_free(&m_pFrameScale);//av_frame_alloc()对应}if (m_pFrameYUV){av_frame_free(&m_pFrameYUV);}avcodec_close(m_pVideoDecodeCodecCtx);avformat_close_input(&m_pInputFormatCtx);return 0;}

#include <iostream>
#include <Windows.h>#include "1__DecodeH264/DecodeH264.h"int main()
{CDecodeH264* m_pDecodeVideo = new CDecodeH264();m_pDecodeVideo->Start();return 0;
}

图是雷神博客的：注册函数废弃了，解码函数变了。

二  相关的结构体，方便记忆

1 AVFrame是未压缩的，解码后的数据。AVPacket是压缩的，解码前的数据。知道了这个，编码的send_frame、receive_packet，解码的send_packet、receive_frame，容易记住了。2 2个Context(上下文)：Format(混合文件、流地址)、Codec(单个编码格式，比如H264、AAC，编解码实现)AVFormatContext*  m_pInputFormatCtx;AVCodecContext*   m_pVideoDecodeCodecCtx;m_pInputFormatCtx会用到的函数：avformat_open_input、avformat_find_stream_info、 av_read_frame、avformat_close_input。m_pOutputFormatCtx会用到的函数：avcodec_find_decoder、avcodec_open2、avcodec_send_packet、 avcodec_receive_frame。3 AVCodec结构体
const  AVCodec  ff_h264_decoder = {.name                  = "h264",.long_name             = NULL_IF_CONFIG_SMALL("H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10"),.type                  = AVMEDIA_TYPE_VIDEO,.id                    = AV_CODEC_ID_H264,.priv_data_size        = sizeof(H264Context),.init                  = h264_decode_init,.close                 = h264_decode_end,.decode                = h264_decode_frame,……
}static const AVCodec * const codec_list[] = {
...&ff_h264_decoder,
...
};

三 兼容性问题

1 文件名带中文，需要转换。

2 播放竖屏视频（手机录的那种），获取旋转角度。

截图是ffmpeg做法：获取角度后，使用filter调整。

3 宽比较特殊，不是16，32的整数。(比如544x960，544是32的倍数)。用linesize[i]代替宽。

linesize跟cpu有关，是cpu 16、32的倍数。

其它，待更新。

四  为什么需要sws_scale？转换到统一格式I420

sws_scale作用：1 分辨率缩放、 2 不同YUV、RGB格式转换。

H264有记录编码前的YUV采样格式，chroma_format_idc，在sps里。如果没有这个字段，说明该字段用的默认值1，即yuv 4:2:0。

如果YUV的采样格式是yuv 4:2:0，也不要求缩放，不需要sws_scale。

avcodec_receive_frame(AVCodecContext *avctx, AVFrame *frame);

frame->format记录了yuv的类型。

ffmpeg默认解码成：编码前的yuv格式。即m_pVideoDecodeCodecCtx->pix_fmt。

int ff_decode_frame_props(AVCodecContext *avctx, AVFrame *frame)
{...frame->format              = avctx->pix_fmt;...
}

五  不同格式的time_base

H264的time_base：1/1200000。

flv：音视频都是1/1000。

mp4：视频1/12800(帧率25，怎么算出来的？)，音频：1/48000(1/采样频率)。