FFmpeg支持多线程编码并保存mp4文件示例

之前介绍的示例：

(1).https://blog.csdn.net/fengbingchun/article/details/132129988 中对编码后数据保存成mp4

(2).https://blog.csdn.net/fengbingchun/article/details/132128885 中通过AVIOContext实现从内存读取数据

(3).https://blog.csdn.net/fengbingchun/article/details/132389734 中将图像加载到视频中

这里将三部分整合到类中，便于后面增加测试代码，下面的示例是两个线程：从内存中读取数据，并将指定的图像加载到视频，将结果保存成mp4。

示例代码如下：

1. 类PacketScaleQueue：用于持续的从指定内存中读取原始数据，上面的示例中已包含此代码

2.类CodecQueue：用于将解码数据存入队列中，并通过单独的线程进行编码

class AVFrameQueue {
public:AVFrameQueue() = default;~AVFrameQueue() {}void push(AVFrame** frame) {std::unique_lock<std::mutex> lck(mtx);queue.push(*frame);cv.notify_all();}void pop(AVFrame** frame) {std::unique_lock<std::mutex> lck(mtx);while (queue.empty()) {//cv.wait(lck);if (cv.wait_for(lck, std::chrono::milliseconds(150)) == std::cv_status::timeout) {fprintf(stderr, "#### Warning: wait timeout\n");*frame = nullptr;return;}}*frame = queue.front();queue.pop();}size_t size() const {return queue.size();}private:std::queue<AVFrame*> queue;std::mutex mtx;std::condition_variable cv;
};class CodecQueue {
public:CodecQueue() = default;void init(unsigned int frame_num) {for (auto i = 0; i < frame_num; ++i) {AVFrame* frame = nullptr;pushDecode(&frame);}}~CodecQueue() { release(); }void release() {AVFrame* frame = nullptr;while (getDecodeSize() > 0) {popDecode(&frame);av_frame_free(&frame);}while (getEncodeSize() > 0) {popEncode(&frame);av_frame_free(&frame);}}void pushDecode(AVFrame** frame) { decode_queue.push(frame); }void popDecode(AVFrame** frame) { decode_queue.pop(frame); }size_t getDecodeSize() const { return decode_queue.size(); }void pushEncode(AVFrame** frame) { encode_queue.push(frame); }void popEncode(AVFrame** frame) { encode_queue.pop(frame); }size_t getEncodeSize() const { return encode_queue.size(); }private:AVFrameQueue decode_queue, encode_queue;
};

3.类VideoCodec：供外面的接口调用，封装了视频的解码和编码过程

声明如下：

typedef struct CodecCtx {char outfile_name[VIDEO_CODEC_MAX_STRING_SIZE];char video_size[VIDEO_CODEC_MAX_STRING_SIZE];char bitrate_str[VIDEO_CODEC_MAX_STRING_SIZE];char pixel_format[VIDEO_CODEC_MAX_STRING_SIZE];char filter_descr[VIDEO_CODEC_MAX_STRING_SIZE];AVFormatContext* ifmt_ctx;AVFormatContext* ofmt_ctx;AVCodecContext* dec_ctx;AVCodecContext* enc_ctx;AVFrame* dec_frame;AVFilterContext* buffersink_ctx;AVFilterContext* buffersrc_ctx;AVFilterGraph* filter_graph;AVPacket* enc_pkt;AVRational frame_rate;int term_status;int stream_index;int frame_count;bool encode_thread_end;
} CodecCtx;class VideoCodec {
public:VideoCodec() = default;~VideoCodec() {  }void setOutfileName(const std::string& name) { outfile_name_ = name; }void setVideoSize(const std::string& size) { video_size_ = size; }void setPixelFormat(const std::string& format) { pixel_format_ = format; }void setFilterDescr(const std::string& filter_descr) { filter_descr_ = filter_descr; }void stopEncode() {while (raw_packet_queue_.getScaleSize() != 0);codec_ctx_->term_status = 1;Buffer buffer;raw_packet_queue_.popPacket(buffer);memset(buffer.data, 0, block_size_);raw_packet_queue_.pushScale(buffer); // for av_read_frame to exit normally}PacketScaleQueue& get_raw_packet_queue(unsigned int buffer_num, size_t buffer_size) {raw_packet_queue_.init(buffer_num, buffer_size);block_size_ = buffer_size;return raw_packet_queue_;}int openEncode();int processEncode();int closeEncode();private:std::string outfile_name_ = "";std::string video_size_ = "";std::string pixel_format_ = "";std::string filter_descr_ = "";PacketScaleQueue raw_packet_queue_;int block_size_ = 0;CodecCtx* codec_ctx_ = nullptr;AVIOContext* avio_ctx_ = nullptr;CodecQueue codec_queue_;std::thread encode_thread_;int get_decode_context();int get_encode_context();int init_filters();int filter_encode_write_frame(AVFrame* frame);int get_output_format_context();int flush_encode_write_frame();int flush_decoder();int flush_encoder();void flush_codec();
};

类VideoCodec实现部分：是之前示例的整理，参考之前示例

4.测试代码，即调用VideoCodec接口，以下是同时两个线程进行编码写

namespace {const int total_push_count = 121;
bool flag1 = true;
const size_t block_size_1 = 640 * 480 * 3;
size_t total_push_count_1 = 0;void fill_raw_data_1(PacketScaleQueue& raw_packet)
{unsigned char value = 0;while (total_push_count_1 < total_push_count) {value += 10;if (value >= 255) value = 0;Buffer buffer;raw_packet.popPacket(buffer);memset(buffer.data, value, block_size_1);raw_packet.pushScale(buffer);std::this_thread::sleep_for(std::chrono::milliseconds(33));++total_push_count_1;}flag1 = false;
}void sleep_seconds_1(VideoCodec& video_codec)
{while (flag1) {std::this_thread::sleep_for(std::chrono::milliseconds(33));}video_codec.stopEncode();
}void encode_1()
{VideoCodec video_codec;video_codec.setOutfileName("out1.mp4");video_codec.setVideoSize("640x480");video_codec.setPixelFormat("bgr24");video_codec.setFilterDescr("movie=1.jpg[logo];[in][logo]overlay=10:20[out]");auto& raw_queue = video_codec.get_raw_packet_queue(16, block_size_1);std::thread thread_fill(fill_raw_data_1, std::ref(raw_queue));auto ret = video_codec.openEncode();if (ret != 0) {std::cout << "fail to openEncode: " << ret << std::endl;//return -1;}std::thread thread_sleep(sleep_seconds_1, std::ref(video_codec));ret = video_codec.processEncode();if (ret != 0) {std::cout << "fail to processEncode: " << ret << std::endl;//return -1;}thread_fill.join();thread_sleep.join();video_codec.closeEncode();std::cout << "1 total push count: " << total_push_count_1 << std::endl;
}bool flag2 = true;
const size_t block_size_2 = 640 * 480 * 3;
size_t total_push_count_2 = 0;void fill_raw_data_2(PacketScaleQueue& raw_packet)
{unsigned char value = 0;while (total_push_count_2 < total_push_count) {value += 10;if (value >= 255) value = 0;Buffer buffer;raw_packet.popPacket(buffer);memset(buffer.data, value, block_size_2);raw_packet.pushScale(buffer);std::this_thread::sleep_for(std::chrono::milliseconds(33));++total_push_count_2;}flag2 = false;
}void sleep_seconds_2(VideoCodec& video_codec)
{while (flag2) {std::this_thread::sleep_for(std::chrono::milliseconds(33));}video_codec.stopEncode();
}void encode_2()
{VideoCodec video_codec;video_codec.setOutfileName("out2.mp4");video_codec.setVideoSize("640x480");video_codec.setPixelFormat("bgr24");video_codec.setFilterDescr("movie=1.jpg[logo];[in][logo]overlay=10:20[out]");auto& raw_queue = video_codec.get_raw_packet_queue(16, block_size_2);std::thread thread_fill(fill_raw_data_2, std::ref(raw_queue));auto ret = video_codec.openEncode();if (ret != 0) {std::cout << "fail to openEncode: " << ret << std::endl;//return -1;}std::thread thread_sleep(sleep_seconds_2, std::ref(video_codec));ret = video_codec.processEncode();if (ret != 0) {std::cout << "fail to processEncode: " << ret << std::endl;//return -1;}thread_fill.join();thread_sleep.join();std::cout << "2 total push count: " << total_push_count_2 << std::endl;
}} // namespceint test_ffmpeg_libavfilter_movie_multi_thread()
{std::thread thread_1(encode_1);std::thread thread_2(encode_2);thread_1.join();thread_2.join();std::cout << "test finish" << std::endl;return 0;
}

生成的mp4文件结果如下：在release下生成的两个视频文件完全一致；在debug下编码过程中有时会timeout