A new video coding scheme based upon motion pattern recognition

Author

Hongsong, Li ; Baohua, Xu

Author_Institution

Coll. of Inf. Sci. & Technol., Beijing Normal Univ., Beijing

fYear

2007

fDate

12-15 Feb. 2007

Firstpage

1

Lastpage

4

Abstract

To improve the performances of block-based motion estimation and motion complement (ME+MC), a new object-based video coding scheme using motion pattern recognition (MPR) is proposed and tested in conference video coding. The object pattern codebook for motion pattern recognition is trained by frequency sensitive self-organizing feature maps (FSSOM) algorithm. Experimental results for standard testing sequence dasiaWalter Cronkite framepsila show that FSSOM-MPR algorithm has better prediction coding performances than ME+MC algorithm.

Keywords

motion estimation; prediction theory; self-organising feature maps; video coding; Walter Cronkite frame; block-based motion estimation; conference video coding; frequency sensitive self-organizing feature maps algorithm; motion complement; motion pattern recognition training; object pattern codebook; object-based video coding scheme; prediction coding; standard testing sequence; Clustering algorithms; Educational institutions; Image coding; Information science; Motion estimation; Pattern recognition; Performance evaluation; Testing; Video coding; Videoconference;

fLanguage

English

Publisher

ieee

Conference_Titel

Signal Processing and Its Applications, 2007. ISSPA 2007. 9th International Symposium on

Conference_Location

Sharjah

Print_ISBN

978-1-4244-0778-1

Electronic_ISBN

978-1-4244-1779-8

Type

conf

DOI

10.1109/ISSPA.2007.4555519

Filename

4555519