This technology is a software application for algorithmic selection of key frames from unedited or semi-edited videos that can be used to summarize a video for browsing or streaming.
Current methods to summarize a video without detracting from the video’s overall information content require video pre-processing or computationally expensive methods. Key frame selection for summarization may be based on video segmentation, but requires edited or semi-edited videos, while clustering approaches require determination of a number of threshold parameters.
This technology is an algorithm that selects key video frames from a larger video file to optimize the video’s information content for summarization or streaming. This is accomplished by breaking the larger video file into subsequences that may be compared by an algorithm to optimize the distance between frames. After several iterations, the frames with the optimal distance values are retrieved and assembled into the final set of key frames to produce the video summary. Compared to existing methods, this technology achieves video summarization with low computational expense and is applicable to both unedited and semi-edited videos.
This technology has been demonstrated to perform better in key frame selection and in temporal video sampling than existing methods.
Patent Issued (US 7,643,597)
IR M02-053
Licensing Contact: Richard Nguyen