Recherche avancée

Sur d’autres sites (4283)

• Extract RTP-packages from video

  5 août 2019, par Gregor A. Lamche

  I have a proprietary REST server that acts as a sort of RTSP server and establishes a UDP stream to another program for video streaming (no audio). It works fine and I can already stream videos. The only issue is that it can only stream one (proprietary) file (container) format at the moment as I’m limited by my media source code.

  I need a code/library which I can give any video file of a common video format/encoding (MPEG-4/MJPEG/H264) and it returns me the RTP-packages that I can then stream (or how to extract the content myself). Alternatively I already got a video player which gives me YUV frames of any video I give it, which in turn I could convert into RTP-packages provided I know how.

  Rendering is done on client side, and something I don’t care about. Though, I will have to set the encoding flags correctly.

  I know roughly about the RTP-header, but have no idea what bytes to write into its content. I heard of FFmpeg and Live555, but for neither could I find example code to extract those packages from a file (let alone a c# wrapper).

• Ffmpeg - convert Aces exr sequence to h264

  25 mars 2023, par Michael B

  I would like to convert an exr sequence to an h264 video. This is fairly easy but the exr sequence is in [AcesCG][1] colourspace which is a colourspace created by [OCIO][2] (an open source colour management solution that has become the standard for VFX).

  


  One solution would be to convert the exrs to a more common colourspace (e.g. srgb or rec709) using [oiio][3] and then use ffmpeg to create an h264 from those images, but that adds an extra step and will be slow. I would like to find a solution where I can directly convert the exrs into an h264 using open source software (e.g. ffmpeg)

  


  [1] : https://en.m.wikipedia.org/wiki/Academy_Color_Encoding_System#:~:text=ACEScg%20(ACES%20computer%20graphics%20space,graphics%20rendering%20and%20compositing%20tools.  ;[2] : https://opencolorio.org/
[3] : https://www.google.com/url?sa=t&source=web&rct=j&url=https://github.com/OpenImageIO/oiio&ved=2ahUKEwjZ9_Gc59D8AhWKdcAKHcG_CUEQFnoECDIQAQ&usg=AOvVaw27FWYrbTnUMqgfYMPRKia_

  


• How can I programmatically write and read random video watermarks ?

  13 novembre 2017, par GreenTriangle

  I spent a few minutes trying to think of a clearer way to word my title, but I couldn’t manage it, sorry.

  I want to essentially canary trap video files : I am (hypothetically, this is not real but a personal exercise) offering them up to 5,000 different people, and if one gets leaked, I want to know who leaked it. Metadata is too easily emoved, so what I’d like to do is add a random and subtle watermark to each file, and store information about that in a database.

  For example : on Joe Smith’s copy, a 10x10 pixel 80% transparent red square in the upper left corner for 5 frames. On Diane Brown’s copy, a full-width 5-pixel 90% transparent black bar on the bottom edge for 15 frames. Then, if I find a leaked copy, I could check it against the database.

  I know this still isn’t foolproof : cropping would break co-ordinates, hue/brightness transforms would break colour reading, cutting time would break timestamps. But if I did want to do this anyway, what would be a good strategy for it ?

  My idea was to generate PNG overlays randomly, split the video into parts with mkvtoolnix/ffmpeg, re-encode the middle part with ffmpeg + overlay filter, and then rejoin them. But is this silly when there’s a "proper" way to do it ? And what would I be doing to read the watermarks, which I can’t even really conceive of ?