views:

39

answers:

2

Hello everyone,

For a video file (e.g. wmv file or other format), I need to generate abstract for the video. The abstract is expected to be in text format. Any mature tools or algorithms? Better (not a must) if open source tools/SDKs so that I can modify for my special needs.

thanks in advance, George

+1  A: 

If I understand 'abstract' correctly, you should look into Hachoir-Metadata, see their example page:

http://bitbucket.org/haypo/hachoir/wiki/hachoir-metadata/examples

JesperSM
Or Kaa Metadata: http://doc.freevo.org/api/kaa/metadata/index.html
JesperSM
I think this tool works in this way -- it reads the meta-data stored inside the binary video file, correct? If yes, it is not what I need. I want to generate text abstract for the video, not based on meta-data of video, but the video's binary content (e.g. frames). Any comments or tools?
George2
+1  A: 

If you're looking for a textual description of a video's content given only the video's frames, then I'm afraid you're out of luck -- the current state of computer vision research is well short of that goal. If you can settle for a non-textual description, video keyframe or shot/scene detection algorithms can give you an image-based summary.

jeff7
Thanks jeff7! If I want to generate non-textual information, any tools and algorithms to use?
George2
A simple approach is can be found in this project: http://moviethumbnail.sourceforge.net/and for quick review of academic techniques google "Video Shot Boundary Detection and Condensed Representation" by C Cotsaces.
jeff7
Cool, thanks! :-)
George2