0
$\begingroup$

How would you reliably detect scene transitions in a video? The simplest case wouldn't be too hard, but sometimes it might be a fast moving scene, or there might be an effect like lightning that would throw it off.

$\endgroup$
2
$\begingroup$

Scenes generally exhibit a fade-to-black transition. You could capture those frames with simple image processing tools, which are found in many libraries, such as OpenCV.

If you want to rely on the change of the mis-an-scene to be robust against sudden changes, then of cours algorithms for detecting temporal differences is more appropriate. For that, you might want to consider:

http://vis.uky.edu/~cheung/courses/ee639_fall04/readings/spie99.pdf http://www.mitsubishi-electric-itce.fr/uk-rce/pubdocs/VIL07-D071.pdf

If you want a simple algorithm do as described here: https://stackoverflow.com/questions/4801053/video-scene-detection-implementation

For harder cases, you might even try GMM-like background models to test if the transition was a temporarily quick shot or a complete change.

$\endgroup$
1
$\begingroup$

What about computing the sum of squared differences or the correlation between the current frame and the previous frame? For fast moving scenes, you could compute phase correlation via Discrete Fourier Transform, but I think that computation would be too expensive for video processing, especially if there's a real-time requirement.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.