Apparatus and method for automatic video summarization using fuzzy one-class support vector machines
Disclosed is an automatic video summarization device and method using a fuzzy OC-SVM (one-class support vector machine) algorithm. A user's subjective decision is reflected in order to generate an effective video summary, and a method for generating flexible video summary information which satisfies the user's environment or requirements is provided. Important video segments are extracted from a given video, and a sequence of key frames is extracted from the video segments, and hence, the user can catch the contents of the video quickly and access desired video scenes.
1. An automatic video summarizer comprising:
an input unit for receiving a video source to be summarized and a desired summarization time from a user;
an importance measurement module for generating importance degrees according to category characteristics of the video and a purpose of desired summary; and
a video summarization generation module for receiving shot information of the video source and a visual characteristic vector, and for applying a combination of the importance degrees, the shot information, and the visual characteristic vector to a fuzzy OC-SVM (one-class support vector machine) algorithm to generate a video summary,
wherein the video summarization generation module comprises a scalability processing module for receiving the summarization time information from the user, repeatedly performing a scalability process, and generating a video summary having a time range desired by the user.
2. The automatic video summarizer of claim 1 , further comprising a shot detection module for extracting the shot information from the video source.
3. The automatic video summarizer of claim 1 , comprising:
an output unit for outputting the generated video summary to a screen; and a storage unit
for storing the generated video summary.
4. The automatic video summarizer of claim 2 , wherein the shot detection module detects a shot from the video source to be summarized, configures a shot list, and transmits the shot list to the video summarization generation module.
5. An automatic video summarization method comprising:
(a) receiving, by an automatic video summarizer, a video source to be summarized and a desired summarization time from a user;
(b) extracting, by the automatic video summarizer, shot information from the video source;
(c) generating, by the automatic video summarizer, importance degrees according to the video's category characteristic and a purpose of desired summary; and
(d) applying, by the automatic video summarizer, a combination of the shot information, the importance degrees, and a visual characteristic vector to a fuzzy OC-SVM (one-class support vector machine) algorithm to generate a video summary,
wherein the automatic video summarizer, in response to the summarization time information from the user, repeatedly performs a scalability process, and generates a video summary having a time range desired by the user.
6. The automatic video summarization method of claim 5 , further comprising:
outputting the generated video summary to the screen; and
storing the generated video summary.
7. An automatic video summarization method comprising:
(a) receiving, by an automatic video summarizer, a video source to be summarized and a desired summarization time from a user;
(b) generating, by the automatic video summarizer, importance degrees according to the video s category characteristic and a purpose of desired summary;
(c) applying, by the automatic video summarizer, a combination of shot information of the video source, the importance decrees, and a visual characteristic vector to a fuzzy OC-SVM (one-class support vector machine) algorithm to generate a video summary;
(d) outputting, by the automatic video summarizer, the generated video summary to a screen; and
(e) storing, by the automatic video summarizer, the generated video summary,
wherein the automatic video summarizer, in response to the summarization time information from the user, repeatedly performs a scalability process, and generates a video summary having a time range desired by the user.
8. A recording medium storing a program for an automatic video summarization method, comprising:
receiving, by an automatic video summarizer, a video source to be summarized and a desired summarization time from a user;
extracting, by the automatic video summarizer, shot information from the video source;
generating, by the automatic video summarizer, importance degrees according to the video's category characteristic and a purpose of desired summary; and
applying, by the automatic video summarizer, a combination of the shot information, the importance degrees, and a visual characteristic vector to a fuzzy OC-SVM (one-class support vector machine) algorithm to generate a video summary,
wherein the automatic video summarizer, in response to the summarization time information from the user, repeatedly performs a scalability process, and generates a video summary having a time range desired by the user.