Video Sensing: Final Project idea

Scene Detection

For my final project I am going to work with scene detection on pre-recorded video. I intend to explore not just cut or shot detection, but scene detection (or at the very least an approximation of it). More simply, I also intend to explore commercial detection techniques, in order to mark and remove commercials from recordings of television broadcasts.

My goal is to generate a list of timestamps for the in/out markers of scenes (or commercial breaks), which can then be used for a variety of purposes


  • Video Auto Summary

    Generate a timecode list of 5 second snippets of each scene in a video. Playback the snippets in order, creating a short summary of the video (re-encoding or saving the video is unnecessary; just playback each start/end point.

  • Video Remix

    Build upon the Auto Summary technique. Add the ability to generate summary lists for multiple videos. Create a basic UI for re-arranging the snippets and remixing several videos into one playback stream. Perform the same functions automatically.

  • Commercial Removal

    Generate a list of commercial break markers and feed them to an AppleScript to use in conjunction with PVR Software (in this case, El Gato’s EyeTV), to edit the commercials out of the video.


A few links I found that relate to video scene detection techniques

Automatic video boundary detection
Shot-boundary detection: unraveled and resolved? (PDF)
Neighborhood coherence and edge based approaches to film scene extraction (PDF)

One thought on “Video Sensing: Final Project idea

  1. Pingback: Video Sensing

Leave a Reply

Your email address will not be published. Required fields are marked *