VideoInfo.h

Go to the documentation of this file.

/*
 * Author: Phil Schatzmann
 * Video Information Structure for RTSP Video Streaming
 */
 
#pragma once
#include "AudioTools/CoreAudio/AudioTypes.h"
 
namespace audio_tools {
 
enum class RTSPVideoFormat {
  RGB24,     
  RGB565,    
  YUV420,    
  JPEG,      
  MJPEG,     
  GRAYSCALE, 
  H264       
};
 
struct RTSPVideoInfo {
  uint16_t width = 640;           
  uint16_t height = 480;          
  float framerate = 30.0f;        
  RTSPVideoFormat format = RTSPVideoFormat::MJPEG; 
  uint8_t bits_per_pixel = 24;    
  uint32_t rtp_clock_rate = 90000; 
 
  RTSPVideoInfo() = default;
 
  RTSPVideoInfo(uint16_t w, uint16_t h, float fps = 30.0f,
            RTSPVideoFormat fmt = RTSPVideoFormat::MJPEG, uint8_t bpp = 24)
    : width(w), height(h), framerate(fps), format(fmt), bits_per_pixel(bpp) {}
 
  uint32_t frameSize() const {
    switch (format) {
      case RTSPVideoFormat::RGB24:
        return width * height * 3;
      case RTSPVideoFormat::RGB565:
        return width * height * 2;
      case RTSPVideoFormat::YUV420:
        return (width * height * 3) / 2; // Y + U/2 + V/2
      case RTSPVideoFormat::GRAYSCALE:
        return width * height;
      case RTSPVideoFormat::JPEG:
      case RTSPVideoFormat::MJPEG:
        // Variable size - return max estimate (1/4 compression ratio)
        return (width * height * bits_per_pixel) / (8 * 4);
      case RTSPVideoFormat::H264:
        // Variable size - return a rough estimate (much higher compression
        // ratio than JPEG thanks to inter-frame prediction)
        return (width * height * bits_per_pixel) / (8 * 20);
      default:
        return width * height * (bits_per_pixel / 8);
    }
  }
 
  uint32_t timestampIncrement() const {
    return (uint32_t)(rtp_clock_rate / framerate);
  }
 
  uint32_t framePeriodUs() const {
    return (uint32_t)(1000000.0f / framerate);
  }
 
  float aspectRatio() const {
    return (float)width / (float)height;
  }
 
  bool isCompressed() const {
    return format == RTSPVideoFormat::JPEG || format == RTSPVideoFormat::MJPEG ||
           format == RTSPVideoFormat::H264;
  }
 
  String toString() const {
    String result = "VideoInfo: ";
    result += width;
    result += "x";
    result += height;
    result += " @ ";
    result += framerate;
    result += "fps, ";
 
    switch (format) {
      case RTSPVideoFormat::RGB24: result += "RGB24"; break;
      case RTSPVideoFormat::RGB565: result += "RGB565"; break;
      case RTSPVideoFormat::YUV420: result += "YUV420"; break;
      case RTSPVideoFormat::JPEG: result += "JPEG"; break;
      case RTSPVideoFormat::MJPEG: result += "MJPEG"; break;
      case RTSPVideoFormat::GRAYSCALE: result += "GRAY"; break;
      case RTSPVideoFormat::H264: result += "H264"; break;
      default: result += "UNKNOWN"; break;
    }
 
    result += " (";
    result += bits_per_pixel;
    result += "bpp)";
    return result;
  }
 
  void logInfo(const char* source = "VideoInfo") const {
    LOGI("%s: %s", source, toString().c_str());
  }
};
 
class RTSPVideoInfoSupport {
 public:
  virtual ~RTSPVideoInfoSupport() = default;
  virtual void setVideoInfo(RTSPVideoInfo info) = 0;
  virtual RTSPVideoInfo videoInfo() = 0;
};
 
} // namespace audio_tools