Color classification to improve block-based motion estimation in RGB-image sequences

Author

Hofmeister, Henning ; Brückner, Bernd

Author_Institution

Leibniz Inst. for Neurobiol., Magdeburg, Germany

Volume

3

fYear

1999

fDate

1999

Firstpage

1224

Abstract

The paper describes a neurobiologically motivated method for analysing monocular colored image sequences. A combination of color classification, using fuzzy-ART networks and motion estimation by blockmatching realises a figure-ground separation within an image block. So one requirement of conventional blockmatching, that all pixels of an image block have to move uniformly, isn´t fulfilled. After color classification processing and blockmatching, moving objects can be separated and an object model can be extracted, which is based on motion and color and overcomes the block structure

Keywords

ART neural nets; fuzzy logic; fuzzy neural nets; image classification; image colour analysis; image matching; image sequences; motion estimation; RGB image sequences; block based motion estimation; block structure; blockmatching; color classification; figure-ground separation; fuzzy-ART networks; image block; monocular colored image sequences; moving objects; neurobiologically motivated method; object model extraction; Data mining; Humans; Image analysis; Image color analysis; Image motion analysis; Image sequence analysis; Image sequences; Machine vision; Motion analysis; Motion estimation;

fLanguage

English

Publisher

ieee

Conference_Titel

Neural Information Processing, 1999. Proceedings. ICONIP '99. 6th International Conference on

Conference_Location

Perth, WA

Print_ISBN

0-7803-5871-6

Type

conf

DOI

10.1109/ICONIP.1999.844717

Filename

844717