Shot Transition Detection Based on Locally Linear Embedding

Author

Fu, Maosheng ; Zhang, Daming ; Kong, Min ; Luo, Bin

Author_Institution

Key Lab. of Intell. Comput. & Signal Process., Anhui Univ., Hefei, China

Volume

3

fYear

2010

fDate

13-14 March 2010

Firstpage

338

Lastpage

341

Abstract

Detection of shot transitions servers as the preliminary step to video indexing and retrieval. Locally linear embedding (LLE) algorithm fails when it is applied to video with multi-shot. In this paper, we present a novel framework of shot transitions detection. The method involves two processes: First we extract the manifold feature of shot transition using LLE through addition of virtual frames on an enriched set, and then they are classified by KNN. Experiments show that the recognition rate of shot transition is reached over 90%.

Keywords

feature extraction; image classification; image motion analysis; video retrieval; KNN; feature extraction; k-nearest neighbor; locally linear embedding; shot transition detection; video indexing; video retrieval; virtual frames; Automation; Computer science education; Educational technology; Embedded computing; Feature extraction; Gunshot detection systems; Mechatronics; Signal processing algorithms; Video compression; Video signal processing; feature extraction; locally linear embedding; shot transition detection;

fLanguage

English

Publisher

ieee

Conference_Titel

Measuring Technology and Mechatronics Automation (ICMTMA), 2010 International Conference on

Conference_Location

Changsha City

Print_ISBN

978-1-4244-5001-5

Electronic_ISBN

978-1-4244-5739-7

Type

conf

DOI

10.1109/ICMTMA.2010.382

Filename

5459525