FFMPEG，使用“图像选择区域”插件模糊视频区域

Question

I am building an online video editor. I need to allow the users to blur an area of movies. This must be done graphically, I mean user should be able to select an area of a video, using an screenshot, then the selected area must be blurred. Some thing like this

Is there anyway to map this selected area dimension and its distance from borders to the real values that must be applied to the video?

I mean four numbers, width, length, top, left will be provided using this plug in and I need to use these numbers to blur an area of videos.

I take an screenshot of video. In order to keep the aspect ratio, I will fix the width to 800px and let the height to be scaled up/down. It is clear that the width, length, top, left of the selected area of the screenshot is a factor of the width, length, top, left of the area that must be blurred in video. but I don't know how to get this factor. Besides that I don't know how to get the video resolution.

Thanks in advance.

---

Update

This is my PHP code that gets the selected area dimensions and offsets

$iLeft = $_POST['left'];
$iTop = $_POST['top'];
$iWidth = $_POST['width'];
$iHeight = $_POST['height'];
exec('ffmpeg -i '.$url.' -filter_complex "[0:v]scale=iw*sar:ih,setsar=1,split[bg][bb];[bb]crop='.$iWidth.'*iw/800:iw*'.$iHeight.'/800:'.$iWidth.'*iw/800:'.$iHeight.'*iw/800,boxblur=10[b0];[bg][b0]overlay='.$iLeft.'*W/800:'.$iTop.'*W/800"" '.$name.' > block.txt 2>&1');

$iLeft, $iTop, $iWidth, $iHeight are the left, top, width and height of the selected area via the image plugin.

It blurs many selected areas very well, but areas like this

The Image Left, Top, Width, Height is 257, 39.26666259765625, 10, 391

don't get blurred. Also a video with Dimension 207x207 didn't get blurred as well.

Answer 1

Running

ffprobe in.mp4 -loglevel quiet -select_streams v -show_entries stream=width,height,sample_aspect_ratio -of compact=p=0:nk=1

will produce an output that looks like this:

1280|720|1:1

The first entry is the video width, then the height, and finally the pixel or sample aspect ratio, shown as num:den so it's num/den. That's your video resolution.

With that info, the factor is video-width*sar/800 . This assumes that you'll be scaling all anamorphic videos to square pixels using the scale and setsar filters, and that the 800px screenshot is undistorted as well. If you're using FFmpeg to scale screenshot, it's scale=800:ih*800/iw/sar

So FFmpeg value for area width is screenshot area width * video-width*sar/800 .

and for FFmpeg Y co-ordinate is screenshot top * video-width*sar/800 , and so on.

---

Assuming the code in the comment is for the screenshot, for the full-sized movie, it would be

"[0:v]scale=iw*sar:ih,setsar=1,split[bg][bb];[bb]crop=50*iw/800:50*iw/800:20*iw/800:10*iw/800,boxblur='min(min(cw/2\,ch/2)\,10)'[b0];
 [bg][b0]overlay=20*W/800:10*W/800"

I've scaled the movie to make sure we're always dealing with a square-pixel video.