0

我正在尝试使用MediaMetadataRetrieverandgetFrameAt()方法从视频中获取所有帧。使用FFMPEG，我得到了以下信息：

  1. 视频帧数：234 帧

  2. 以毫秒为单位的视频持续时间：90000

    int counter = 0;
long mVideoDuration = 9000;
for (long i = 0; i < mVideoDuration * 1000; i += 1000) {
    Bitmap thumbnail = mMediaMetadataRetriever.getFrameAtTime(i);
}

上面的代码不起作用，它循环了几乎8944次，这太多了，我不想得到所有这些帧，我只想得到234帧。

getFrameAtIndex() 工作得很好，但由于 API 级别的兼容性，我想让它与getFrameAt()方法一起工作

4

1 回答 1

1

To get the frames with getFrameeAtTime() youd have to know the time of every frame in the video. You could roughly calculate that if u already know the frame count by getting the frametime (duration/frames), taking the math.ceil() method on that and taking that as your incrementor. however, youd have to check if the last frame gets read since u can't always get the exact frame time, bc they tend to have alot of decimal places wich can lead to wrong representation as float. A 60fps video for example would have a frame time of 16+1/3 ms. Wich has infinite deecimal places. In addition this method wouldn't even work at all if you have a video with too high fps since in that case the round up done by math.ceil() can lead to a big enough margin to skip frames. To be precise this method only works if:

(math.ceil(frameTime)-frameTime)⋅frameCount<frameTime

I hope, by now u got that it's a pretty bad idea and u should probably use another way to get the individual frames. But still, heres an implementation:

int lastTime;
long mVideoDuration = 9000;
mVideoDuration*=1000;
int frameTime=math.ceil(mVideoDuration/frameCount)
for(int i =0, i<=mVideoDuration, i+=frameTime){
    Bitmap thumbnail = mMediaMetadataRetriever.getFrameAtTime(i);
    lastTime=i;
}
if (lastTime<mVideoDuration){
   mMediaMetadataRetriever.getFrameAtTime(mVideoDuration);
}

Of couse this is already assuming that,

(math.ceil(frameTime)-frameTime)⋅frameCount<frameTime

is given. And i can't guarantee that the declaframeTime will result the correct value since i dont know exactly how decimal dividing is handleed internally and its entirely possible that is causes problems due to floating point overflow

于 2022-02-25T19:39:18.347 回答