Opened 4 years ago

Last modified 3 years ago

#3389 new enhancement

Feature Request - Isolate and normalize video content

Reported by: Rafajafar Owned by:
Priority: wish Component: undetermined
Version: unspecified Keywords:
Cc: Blocked By:
Blocking: Reproduced by developer: no
Analyzed by developer: no

Description

Hey,

I have an interesting problem and I'm solving it with OpenCV. It's be great if the solution could come straight from ffmpeg, though!

I need to:

1) cropdetect
2) crop surrounding borders from a video
3) resize video to normalize its dimensions

In other words, all I want is one command that will detect the frame-by-frame borders and in the same step, crop it, and also resize it. And I want this done for the entire video, not a single frame. It's gotta be possible, right?

The applications are very useful in video analysis. I've spoken to colleagues of mine and they are doing this manually themselves. It would be great to be able to work this feature into the automation surrounding ffmpeg in our tools. My personal application, for instance, is in detecting videos that are similar using content perceptual hashing. Another colleague of mine works for the Center for Missing and Exploited Children. He does similar work, but with pHashing. I can see applications for ContentID type systems across the board.

I'm sure this is a low on the totem pole request, but if you have any spare cycles, this would really rock in the world of computer vision and video analysis!

Thanks for your consideration,

--Collin

Change History (3)

comment:1 in reply to: ↑ description Changed 4 years ago by cehoyos

Replying to Rafajafar:

1) cropdetect
2) crop surrounding borders from a video
3) resize video to normalize its dimensions

In other words, all I want is one command that will detect the frame-by-frame borders and in the same step, crop it, and also resize it. And I want this done for the entire video, not a single frame. It's gotta be possible, right?

How?

Are you only interested in the cropdetect output for the first frame and want to use that for the whole file?

comment:2 Changed 4 years ago by Rafajafar

That's not what I specifically would need, but I could still use it.

What would be most useful is to crop detect each frame, removing borders from that frame, resize the frame to the normalized dimension, and then stick that into an output movie.

Example movie which would fail detection due to borders: http://www.youtube.com/watch?v=ZaLdCbTINUM

Check out the borders from the beginning of the video versus the borders from 1:55 of the video.

For content detection, the borders are more damaging than the distorted resolution.

comment:3 Changed 3 years ago by cehoyos

  • Component changed from ffmpeg to undetermined
  • Keywords feature request normalize video removed
Note: See TracTickets for help on using tickets.