A novel hybrid approach of color image segmentation

Yangxing Liu*, Takeshi Ikenaga, Satoshi Goto

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contribution

1 Citation (Scopus)


Color image segmentation is probably the most important task in image analysis and understanding. In this paper, we present a novel approach to segment color images by integrating region clustering result with edge detection result. In contrast to existing region-based clustering method, we do not cluster all pixels in an image at one time. We divide clustering process into two steps. First we only cluster those reliable pixels, whose colors are not affected by shadow or highlight, to get more reasonable initial clustering results. Then we cluster left unreliable pixels into classes obtained in previous step or new classes. To avoid over-segmenting an image, edge detection result and spatial information are utilized to merge some neighboring regions, a significant part of whose common boundary consists of weak edges, together as a whole. Experimental results demonstrate the efficacy of our algorithm to segment color images without any prior knowledge.

Original languageEnglish
Title of host publicationAPCCAS 2006 - 2006 IEEE Asia Pacific Conference on Circuits and Systems
Number of pages4
Publication statusPublished - 2006 Dec 1
EventAPCCAS 2006 - 2006 IEEE Asia Pacific Conference on Circuits and Systems - , Singapore
Duration: 2006 Dec 42006 Dec 6

Publication series

NameIEEE Asia-Pacific Conference on Circuits and Systems, Proceedings, APCCAS


ConferenceAPCCAS 2006 - 2006 IEEE Asia Pacific Conference on Circuits and Systems

ASJC Scopus subject areas

  • Electrical and Electronic Engineering


Dive into the research topics of 'A novel hybrid approach of color image segmentation'. Together they form a unique fingerprint.

Cite this