opencv with ffmpeg 代码解析(二)

桃扇骨 2021-05-16 20:36 574阅读 0赞

1、代码入口

看opencv的代码，最好的切入点就是对外的头文件 opencv.hpp

![20200606172421742.png][]

include 目录下仅有opencv.hpp文件

![2020060617323515.png][]

所有的video read和write相关代码都在modules/videoio这个模块里

2、IVideoCapture  IVideoWriter 实现

在videoio.hpp里声明了VideoCapture VideoWriter IVideoCapture IVideoWriter 四种类

其中  IVideoCapture IVideoWriter 为接口类，定义位于 cap\_interface.hpp

当需要使用第三方库如ffmpeg实现video读写时，可以继承接口类

这两个类定义了很多虚函数，方便调用者调用，而继承者只要继承并实现这些函数，就可以实现video read和write功能

class IVideoCapture
    {
    public:
        virtual ~IVideoCapture() {}
        virtual double getProperty(int) const { return 0; }
        virtual bool setProperty(int, double) { return false; }
        virtual bool grabFrame() = 0;
        virtual bool retrieveFrame(int, OutputArray) = 0;
        virtual bool isOpened() const = 0;
        virtual int getCaptureDomain() { return CAP_ANY; } // Return the type of the capture object: CAP_DSHOW, etc...
    };
    
    class IVideoWriter
    {
    public:
        virtual ~IVideoWriter() {}
        virtual double getProperty(int) const { return 0; }
        virtual bool setProperty(int, double) { return false; }
        virtual bool isOpened() const = 0;
        virtual void write(InputArray) = 0;
        virtual int getCaptureDomain() const { return cv::CAP_ANY; } // Return the type of the capture object: CAP_FFMPEG, etc...
    };

3、FFmpeg 读功能实现

下面我们以ffmpeg为例简述这种功能实现方式，功能实现位于 cap\_ffmpeg.cpp、cap\_ffmpeg\_impl.hpp、cap\_ffmpeg\_legacy\_api.hpp 文件中

CvCapture\_FFMPEG\_proxy继承了IVideoCapture，CvCapture\_FFMPEG\_proxy通过 CvCapture\_FFMPEG类实现了video的读取

CvCapture\_FFMPEG定义如下：

struct CvCapture_FFMPEG
    {
        bool open( const char* filename );
        void close();
    
        double getProperty(int) const;
        bool setProperty(int, double);
        bool grabFrame();
        bool retrieveFrame(int, unsigned char** data, int* step, int* width, int* height, int* cn);
    
        void init();
    
        void    seek(int64_t frame_number);
        void    seek(double sec);
        bool    slowSeek( int framenumber );
    
        int64_t get_total_frames() const;
        double  get_duration_sec() const;
        double  get_fps() const;
        int64_t get_bitrate() const;
    
        double  r2d(AVRational r) const;
        int64_t dts_to_frame_number(int64_t dts);
        double  dts_to_sec(int64_t dts) const;
    
        AVFormatContext * ic;
        AVCodec         * avcodec;
        int               video_stream;
        AVStream        * video_st;
        AVFrame         * picture;
        AVFrame           rgb_picture;
        int64_t           picture_pts;
    
        AVPacket          packet;
        Image_FFMPEG      frame;
        struct SwsContext *img_convert_ctx;
    
        int64_t frame_number, first_frame_number;
    
        double eps_zero;
    /*
       'filename' contains the filename of the videosource,
       'filename==NULL' indicates that ffmpeg's seek support works
       for the particular file.
       'filename!=NULL' indicates that the slow fallback function is used for seeking,
       and so the filename is needed to reopen the file on backward seeking.
    */
        char              * filename;
    
    #if LIBAVFORMAT_BUILD >= CALC_FFMPEG_VERSION(52, 111, 0)
        AVDictionary *dict;
    #endif
    #if USE_AV_INTERRUPT_CALLBACK
        AVInterruptCallbackMetadata interrupt_metadata;
    #endif
    
        bool setRaw();
        bool processRawPacket();
        bool rawMode;
        bool rawModeInitialized;
        AVPacket packet_filtered;
    #if LIBAVFORMAT_BUILD >= CALC_FFMPEG_VERSION(58, 20, 100)
        AVBSFContext* bsfc;
     #else
        AVBitStreamFilterContext* bsfc;
    #endif
    };

[20200606172421742.png]: /images/20210516/d7ec9f57bbba40e78adc23ddcba6cdc4.png
[2020060617323515.png]: /images/20210516/3dfdb35491174c4eaa81f3f9931d80ce.png