|
What is VideoQuery ?
VideoQuery is an application for automatic scene detection and querying of video files.
It uses wavelet analysis of frames of video on the detection of transitions between shots in a video stream, thereby dividing the stream into segments (scenes).
Then this scenes are inserted into a video database using an indexing scheme that involves a wavelet-based "signature".
Therefore, one can automatically search for a scene in the database by supplying an image, which is transformed into it's wavelet signature, and then quickly searched for the best match among all scenes in all videos in the database.
Currently, only mpeg2 videos are supported (therefore libmpeg2 is a requirement), but more input plug-ins are planned.
I'm using C++ for the UI and database, and C for the video input plug-ins. The GUI toolkit I'm using is QT 3.x, on Linux systems. Later I may port it to Win32.
|
|
Future plans
I'm thinking of extending VideoQuery to handle image databases as well, using a similar
wavelet transform and query metric.
I'll start a prototype in Python, so I can get a lot of functionality done quickly. The drawback
is that adding images to the database will be slow, so I could perhaps write a quick C program to
add them to a database, and then use python for querying (which is a fast operation), for displaying
the results and even for a web front-end, so you could let people browse your collection of wallpapers
by just clicking an image to get others similar to this one.
If you have any ideas on how all these querying systems should be integrated, and which tools and programming
languages I should use to achieve so, feel free to email me at nieder at mail dot ru.
|
|
|
|
