Title :
Graph Cut Based Unsupervised Color Image Segmentation
Author :
Liang Bin-mei ; Zhang Jian-zhou
Author_Institution :
Coll. of Math. & Inf. Sci., Guangxi Univ., Nanning, China
Abstract :
This paper presents an unsupervised segmentation algorithm for color images. The algorithm consists of two stages. In the first stage, the optimal number of segments is automatically determined by means of a compactness measure that is formulated to find a clustering with "maximum inter-cluster distance and minimum intra-cluster variance". In the second stage, a multiple terminal vertices weighted graph is constructed based on an energy function and the image is segmented. A large number of performance evaluations have been carried out and the experimental results indicate that the proposed approach is effective, and it obtains satisfied results in comparing with other algorithms.
Keywords :
graph theory; image colour analysis; image segmentation; pattern clustering; clustering; compactness measure; energy function; graph cut based unsupervised color image segmentation algorithm; maximum intercluster distance; minimum intracluster variance; terminal vertices weighted graph; Clustering algorithms; Color; Humans; Image segmentation; Indexes; Minimization; Proposals; clustering; graph cut; k-means; unsupervised color image segmentation;
Conference_Titel :
Multimedia (ISM), 2012 IEEE International Symposium on
Conference_Location :
Irvine, CA
Print_ISBN :
978-1-4673-4370-1
DOI :
10.1109/ISM.2012.100