Video mining using unsupervised clustering of video content
First Claim
Patent Images
1. A method for mining content of a sequence of frames of a video, comprising:
- selecting a low-level feature of the content of the sequence of frames of the video;
generating time-series data from the content of the sequence of frames of the video, the time series data comprising the selected low-level feature;
implicitly constructing windows for the time-series data, in which the windows are multi-resolution, a distance D between two windows of size w at points xi and xi in the time series data is measured as and k is an index for the points xi and xj in the windows;
self-correlating the time-series data, using the multi-resolution windows, to determine similar segments of the sequence of frames of the video according to the low-level feature, and wherein all similar segments are found at a plurality of temporal resolutions; and
clustering the similar segments to discover high-level patterns in the content of the sequence of frames of the video.
1 Assignment
0 Petitions
Accused Products
Abstract
A method mines unknown content of a video by first selecting one or more low-level features of the video. For each selected feature, or combination of features, time series data is generated. The time series data is then self-correlated to identify similar segments of the video according to the low-level features. The similar segments are grouped into clusters to discover high-level patterns in the unknown content of video.
55 Citations
26 Claims
-
1. A method for mining content of a sequence of frames of a video, comprising:
-
selecting a low-level feature of the content of the sequence of frames of the video; generating time-series data from the content of the sequence of frames of the video, the time series data comprising the selected low-level feature; implicitly constructing windows for the time-series data, in which the windows are multi-resolution, a distance D between two windows of size w at points xi and xi in the time series data is measured as and k is an index for the points xi and xj in the windows; self-correlating the time-series data, using the multi-resolution windows, to determine similar segments of the sequence of frames of the video according to the low-level feature, and wherein all similar segments are found at a plurality of temporal resolutions; and clustering the similar segments to discover high-level patterns in the content of the sequence of frames of the video. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26)
-
Specification