1. 注册所有容器格式和CODEC:av_register_all()
2. 打开文件:av_open_input_file()
3. 从文件中提取流信息:av_find_stream_info()
4. 穷举所有的流,查找其中种类为CODEC_TYPE_VIDEO
5. 查找对应的解码器:avcodec_find_decoder()
6. 打开编解码器:avcodec_open()
7. 为解码帧分配内存:avcodec_alloc_frame()
8. 不停地从码流中提取出帧数据:av_read_frame()
9. 判断帧的类型,对于视频帧调用:avcodec_decode_video()
10. 解码完后,释放解码器:avcodec_close()
11. 关闭输入文件:av_close_input_file()
FfmpegEncoder.h
/* * FfmpegEncoder.h * * Current, Can Support YUV422sp encoder and decoder * * Created on: Dec 5, 2010 * Author: Henry.Wen */ #ifndef _H264ENCODER_H #define _H264ENCODER_H void save_image(const char* filePath, const void* bufferBase, int width, int height); int encoder_init(const char* filePath, int width, int height); int encoder_frame(const void* frame); int encoder_frame_yuv422(const void* frame); void encoder_close(); #endif
FfmpegEncoder.cpp
#include <math.h> #include <stdio.h> #include <stdlib.h> #include <string.h> extern "C" { #include <libavutil/mathematics.h> #include <libavformat/avformat.h> //#include <libswscale/swscale.h> } #include <skia/core/SkBitmap.h> #include <skia/images/SkImageEncoder.h> #include <android_runtime/AndroidRuntime.h> #include "FfmpegEncoder.h" AVOutputFormat *g_fmt = 0; AVFormatContext *g_oc = 0; AVCodec *g_video_codec = 0; AVStream *g_video_st = 0; AVFrame *g_frame = 0; AVPicture g_picture; int g_frame_count = 0; double g_video_pts = 0; int g_flagInit = 0; int g_width = 0; int g_height = 0; using namespace android; static Mutex sg_mutexLock; #ifndef LOGI #define LOGI(...) ((void)__android_log_print(ANDROID_LOG_INFO, "H264ENCODE", __VA_ARGS__)) #define LOGE(...) ((void)__android_log_print(ANDROID_LOG_ERROR, "H264ENCODE", __VA_ARGS__)) #endif void save_image(const char* filePath, const void* bufferBase, int width, int height) { Mutex::Autolock lock(sg_mutexLock); SkBitmap b; #if (ANDROID_r4_4_0) b.setConfig(SkBitmap::kARGB_8888_Config, width, height,(size_t)0); #else b.setConfig(SkBitmap::kARGB_8888_Config, width, height); #endif b.setPixels((void*)bufferBase); SkImageEncoder::EncodeFile(filePath, b, SkImageEncoder::kJPEG_Type, SkImageEncoder::kDefaultQuality); LOGI("save_image image ok===================="); } AVStream *add_stream(AVFormatContext *oc, AVCodec **codec, enum AVCodecID codec_id, int width, int height) { AVCodecContext *c; AVStream *st; /* find the encoder */ *codec = avcodec_find_encoder(codec_id); LOGI("encoder_init add_stream find encoder='%s'", avcodec_get_name(codec_id)); if (!(*codec)) { LOGE("encoder_init add_stream could not find encoder for '%s' ", avcodec_get_name(codec_id)); return 0; } st = avformat_new_stream(oc, *codec); if (!st) { LOGE("encoder_init add_stream could not allocate stream"); return 0; } st->id = oc->nb_streams-1; c = st->codec; avcodec_get_context_defaults3(c, *codec); c->codec_id = codec_id;//AV_CODEC_ID_MPEG4; c->bit_rate = 40000; /* Resolution must be a multiple of two. */ c->width = width; c->height = height; c->time_base.den = 25; c->time_base.num = 1; c->gop_size = 12; /* emit one intra frame every twelve frames at most */ c->pix_fmt = AV_PIX_FMT_YUV420P; if (c->codec_id == AV_CODEC_ID_MPEG2VIDEO) { c->max_b_frames = 2; } if (c->codec_id == AV_CODEC_ID_MPEG1VIDEO) { c->mb_decision = 2; } /* Some formats want stream headers to be separate. */ if (oc->oformat->flags & AVFMT_GLOBALHEADER) c->flags |= CODEC_FLAG_GLOBAL_HEADER; if(!strcmp(oc->oformat->name, "mp4") || !strcmp(oc->oformat->name, "mov") || !strcmp(oc->oformat->name, "3gp")) c->flags |= CODEC_FLAG_GLOBAL_HEADER; return st; } AVFrame* alloc_picture(AVPixelFormat pix_fmt, int width, int height) { AVFrame *picture; uint8_t *picture_buf; int size; picture = avcodec_alloc_frame(); if (!picture) return NULL; size = avpicture_get_size(pix_fmt, width, height); picture_buf = (uint8_t *)av_malloc(size); if (!picture_buf) { av_free(picture); return NULL; } avpicture_fill((AVPicture *)picture, picture_buf, pix_fmt, width, height); return picture; } int open_video(AVFormatContext *oc, AVCodec *codec, AVStream *st) { int ret = 0; AVCodecContext *c = st->codec; /* open the codec */ ret = avcodec_open2(c, codec, NULL); if (ret < 0) { LOGE("encoder_init open_video could not open video codec: %s", av_err2str(ret)); return -1; } g_frame = avcodec_alloc_frame(); if (!g_frame) { LOGE("encoder_init open_video could not allocate video frame"); return -1; } ret = avpicture_alloc(&g_picture, c->pix_fmt, c->width, c->height); if (ret < 0) { LOGE("encoder_init open_video could not allocate picture: %s", av_err2str(ret)); free(g_frame); g_frame = 0; return -1; } *((AVPicture *)g_frame) = g_picture; return 0; } void close_video(AVStream *st) { if(st->codec) { avcodec_close(st->codec); st->codec = 0; } if(g_frame) { av_free(g_picture.data[0]); av_free(g_frame); g_frame = 0; } } int encoder_init(const char* filePath, int width, int height) { if(g_flagInit) return 0; Mutex::Autolock lock(sg_mutexLock); int ret = 0; LOGI("encoder_init ============begin"); if(!filePath || width <= 0 || height <= 0) { LOGE("encoder_init input parameters error ret = %d", (ret = -1)); return -1; } av_register_all(); /* allocate the output media context */ avformat_alloc_output_context2(&g_oc, NULL, NULL, filePath); if (!g_oc) { LOGI("Could not deduce output format from file extension: using MPEG."); return -1; } g_fmt = g_oc->oformat; g_video_st = add_stream(g_oc, &g_video_codec, g_fmt->video_codec, width, height); av_dump_format(g_oc, 0, filePath, 1); if (g_video_st) { if(open_video(g_oc, g_video_codec, g_video_st) < 0) { LOGE("encoder_init open_video fail!"); close_video(g_video_st); av_free(g_oc); g_oc = 0; return -1; } } else { LOGE("encoder_init g_video_st is null, not enough memory!"); av_free(g_oc); g_oc = 0; return -1; } /* open the output file, if needed */ if (!(g_fmt->flags & AVFMT_NOFILE)) { LOGI("encoder_init avio_open ============begin"); ret = avio_open(&g_oc->pb, filePath, AVIO_FLAG_WRITE); LOGI("encoder_init avio_open ret:%d============end", ret); if (ret < 0) { LOGE("encoder_init could not open '%s': %s", filePath, av_err2str(ret)); close_video(g_video_st); av_free(g_oc); g_oc = 0; return -1; } } LOGI("encoder_init avformat_write_header video file"); ret = avformat_write_header(g_oc, NULL); if (ret < 0) { LOGE("encoder_init error occurred when opening output file: %s ", av_err2str(ret)); close_video(g_video_st); av_free(g_oc); g_oc = 0; return -1; } if (g_frame) g_frame->pts = 0; g_flagInit = 1; g_width = width; g_height = height; LOGI("encoder_init ============end"); return 0; } static struct SwsContext *swsContext; void fill_yuv_image(AVFrame *pict, const void* frame) { int x, y, tmpIndex = 0, tmpWdith = g_width >> 1, tmpHeight = g_height >> 1; unsigned char* tmpBuffer_yuv = (unsigned char*)frame; /* Y */ for (y = 0; y < g_height; ++y) { for (x = 0; x < g_width; ++x) { pict->data[0][y * pict->linesize[0] + x] = *(tmpBuffer_yuv + tmpIndex); ++tmpIndex; } } tmpIndex = 0; int tmpLength = g_width * g_height; unsigned char* tmpBuffer_uv = tmpBuffer_yuv + tmpLength; /* Cb and Cr */ for (y = 0; y < tmpHeight; ++y) { for (x = 0; x < tmpWdith; ++x) { pict->data[1][y * pict->linesize[1] + x] = *(tmpBuffer_uv + tmpIndex + 1); pict->data[2][y * pict->linesize[2] + x] = *(tmpBuffer_uv + tmpIndex); tmpIndex+= 2; } } } //fill AVFrame with YUV422p buffer void fill_yuv422p_image(AVFrame *pict, const void* frameYUV422p) { int width = g_width, height = g_height; unsigned char * pyuv422 = (unsigned char *)frameYUV422p; unsigned char * pyuv420y = &pict->data[0][0]; unsigned char * pyuv420u = &pict->data[1][0]; unsigned char * pyuv420v = &pict->data[2][0]; int uv_count = 0; int i, j; for (i = 0; i < height; i += 2) for (j = 0; j < width; j += 2) { memcpy(pyuv420y + i * width + j, pyuv422 + i * width * 2 + j * 2, 1); memcpy(pyuv420y + (i + 1) * width + j, pyuv422 + (i + 1) * width * 2 + j * 2, 1); memcpy(pyuv420y + i * width + (j + 1), pyuv422 + i * width * 2 + (j + 1) * 2, 1); memcpy(pyuv420y + (i + 1) * width + (j + 1), pyuv422 + (i + 1) * width * 2 + (j + 1) * 2, 1); //±£ÁôU ·ÖÁ¿ memcpy(pyuv420u + uv_count, pyuv422 + i * width * 2 + j * 2 + 1, 1); //±£ÁôV·ÖÁ¿; memcpy(pyuv420v + uv_count, pyuv422 + (i + 1) * width * 2 + (j + 1) * 2 + 1, 1); uv_count++; } /*int x, y, tmpIndex = 0, tmpWdith = g_width >> 1, tmpHeight = g_height >> 1; unsigned char* tmpBuffer_yuv = (unsigned char*)frameYUV422p; /* Y */ /*for (y = 0; y < g_height; ++y) { for (x = 0; x < g_width; ++x) { pict->data[0][y * pict->linesize[0] + x] = *(tmpBuffer_yuv + tmpIndex); ++tmpIndex; } } tmpIndex = 0; int tmpLength = g_width * g_height; unsigned char* tmpBuffer_uv = tmpBuffer_yuv + tmpLength; // Cb and Cr for (y = 0; y < tmpHeight; ++y) { for (x = 0; x < tmpWdith; ++x) { pict->data[1][y * pict->linesize[1] + x] = *(tmpBuffer_uv + tmpIndex); pict->data[2][y * pict->linesize[2] + x] = *(tmpBuffer_uv + tmpIndex +1); tmpIndex += 2; } tmpIndex += g_width; }*/ } void write_video_frame(AVFormatContext *oc, AVStream *st) { int ret; static struct SwsContext *sws_ctx; AVCodecContext *c = st->codec; /* encode the image */ AVPacket pkt; int got_output; av_init_packet(&pkt); pkt.data = NULL; // packet data will be allocated by the encoder pkt.size = 0; ret = avcodec_encode_video2(c, &pkt, g_frame, &got_output); if (ret < 0) { LOGE("encoder_init error encoding video frame: %s ", av_err2str(ret)); return; } //If size is zero, it means the image was buffered. if (got_output) { if (c->coded_frame->key_frame) pkt.flags |= AV_PKT_FLAG_KEY; pkt.stream_index = st->index; ret = av_interleaved_write_frame(oc, &pkt); } else { ret = 0; } av_free_packet(&pkt); if (ret != 0) { LOGE("encoder_init error while writing video frame: %s ", av_err2str(ret)); return; } ++g_frame_count = 0; } int encoder_frame(const void* frame) { if(!g_flagInit) return 0; Mutex::Autolock lock(sg_mutexLock); fill_yuv_image(g_frame, frame); if (g_video_st) g_video_pts = (double)g_video_st->pts.val * g_video_st->time_base.num / g_video_st->time_base.den; else g_video_pts = 0.0; write_video_frame(g_oc, g_video_st); g_frame->pts += av_rescale_q(1, g_video_st->codec->time_base, g_video_st->time_base); return 0; } int encoder_frame_yuv422(const void* frame) { if(!g_flagInit) return 0; Mutex::Autolock lock(sg_mutexLock); fill_yuv422p_image(g_frame, frame); if (g_video_st) g_video_pts = (double)g_video_st->pts.val * g_video_st->time_base.num / g_video_st->time_base.den; else g_video_pts = 0.0; write_video_frame(g_oc, g_video_st); g_frame->pts += av_rescale_q(1, g_video_st->codec->time_base, g_video_st->time_base); return 0; } void encoder_close() { LOGI("encoder_close ============begin"); Mutex::Autolock lock(sg_mutexLock); if(g_oc) { av_write_trailer(g_oc); if (g_video_st) close_video(g_video_st); for(int i = 0; i < (int)g_oc->nb_streams; ++i) { av_freep(&g_oc->streams[i]->codec); av_freep(&g_oc->streams[i]); } if (!(g_fmt->flags & AVFMT_NOFILE)) avio_close(g_oc->pb); av_free(g_oc); } g_oc = 0; g_video_st = 0; g_flagInit = 0; g_frame_count = 0; g_width = 0; g_height = 0; LOGI("encoder_close ============end"); }
首先第一件事情就是开一个视频文件并从中得到流。我们要做的第一件事情就是使用av_register_all();来初始化libavformat/libavcodec:
这一步注册库中含有的所有可用的文件格式和编码器,这样当打开一个文件时,它们才能够自动选择相应的文件格式和编码器。av_register_all()只需
调用一次,所以,要放在初始化代码中。也可以仅仅注册个人的文件格式和编码。
下一步,打开文件:
AVFormatContext *pFormatCtx;
const char *filename="myvideo.mpg";
av_open_input_file(&pFormatCtx, filename, NULL, 0, NULL); // 打开视频文件
最后三个参数描述了文件格式,缓冲区大小(size)和格式参数;我们通过简单地指明NULL或0告诉 libavformat 去自动探测文件格式并且使用默认的缓
冲区大小。这里的格式参数指的是视频输出参数,比如宽高的坐标。
下一步,我们需要取出包含在文件中的流信息:
av_find_stream_info(pFormatCtx); // 取出流信息
AVFormatContext 结构体
dump_format(pFormatCtx, 0, filename, false);//我们可以使用这个函数把获取到得参数全部输出。
for(i=0; i<pFormatCtx->nb_streams; i++) //区分视频流和音频流
if(pFormatCtx->streams->codec.codec_type==CODEC_TYPE_VIDEO) //找到视频流,这里也可以换成音频
{
videoStream=i;
break;
}
接下来就需要寻找解码器
AVCodec *pCodec;
pCodec=avcodec_find_decoder(pCodecCtx->codec_id);
avcodec_open(pCodecCtx, pCodec); // 打开解码器
给视频帧分配空间以便存储解码后的图片:
AVFrame *pFrame;
pFrame=avcodec_alloc_frame();
/////////////////////////////////////////开始解码///////////////////////////////////////////
第一步当然是读数据:
我们将要做的是通过读取包来读取整个视频流,然后把它解码成帧,最后转换格式并且保存。
while(av_read_frame(pFormatCtx, &packet)>=0) { //读数据
if(packet.stream_index==videoStream){ //判断是否视频流
avcodec_decode_video(pCodecCtx,pFrame, &frameFinished,
packet.data, packet.size); //解码
if(frameFinished) {
img_convert((AVPicture *)pFrameRGB, PIX_FMT_RGB24,(AVPicture*)pFrame, pCodecCtx->pix_fmt, pCodecCtx->width,pCodecCtx->height);//转换
SaveFrame(pFrameRGB, pCodecCtx->width,pCodecCtx->height, i); //保存数据
av_free_packet(&packet); //释放
av_read_frame()读取一个包并且把它保存到AVPacket结构体中。这些数据可以在后面通过av_free_packet()来释 放。函数avcodec_decode_video()把包
转换为帧。然而当解码一个包的时候,我们可能没有得到我们需要的关于帧的信息。因此,当我们得 到下一帧的时候,avcodec_decode_video()为我们设
置了帧结束标志frameFinished。最后,我们使用 img_convert()函数来把帧从原始格式(pCodecCtx->pix_fmt)转换成为RGB格式。要记住,你可以把一个
AVFrame结构体的指针转换为AVPicture结构体的指针。最后,我们把帧和高度宽度信息传递给我们的SaveFrame函数。
到此解码完毕,显示过程使用SDL完成考虑到我们以后会使用firmware进行显示操作,SDL忽略不讲。
音视频同步
DTS(解码时间戳)和PTS(显示时间戳)
当我们调用av_read_frame()得到一个包的时候,PTS和DTS的信息也会保存在包中。但是我们真正想要的PTS是我们刚刚解码出来的 原始帧 的PTS,这样我
们才能知道什么时候来显示它。然而,我们从avcodec_decode_video()函数中得到的帧只是一个AVFrame,其中并 没有包含有用的PTS值(注意:AVFrame并
没有包含时间戳信息,但当我们等到帧的时候并不是我们想要的样子)。。我们保存一帧的第一个包的PTS: 这将作为整个这一帧的PTS。我们 可以通过函
数avcodec_decode_video()来计算出哪个包是一帧的第一个包。怎样实现呢?任何时候当一个包开始一帧的时 候,avcodec_decode_video()将调用一个函数
来为一帧申请一个缓冲。当然,ffmpeg允许我们重新定义那个分配内存的函数。计算前 一帧和现在这一帧的时间戳来预测出下一个时间戳的时间。同时,我
们需要同步视频到音频。我们将设置一个音频时间audioclock;一个内部值记录了我 们正在播放的音频的位置。就像从任意的mp3播放器中读出来的数字一
样。既然我们把视频同步到音频,视频线程使用这个值来算出是否太快还是太慢。