Article Preview
TopIntroduction
One of computer vision major goals is to identify the objects in digital images (Haralick 1992). The complexity in digital objects identification follows the complexity of the image varying from homogeneous regions to a set of region in a specific structure. Thus, the first step to identify objects is to partition images into consistent segments (Comaniciu 2002; Lezoray 2002). Different algorithms have been developed to segment images. Many approaches require identifying the number of segments in advance such as region growing which requires identifying number of region (Al-Mamun 2007; Arbelaez 2011; Comaniciu 2002; Hedjam 2009; Herman 2001; Lucchese 1999; Nock 2005).
Other segmentation approaches rely on stochastic model-based for gray level images (Lucchese 1999; Shah-Hosseini 2002; Tu 2002). Dana et.al, (2011) reviewed image segmentation algorithm based on color and texture. Recently, many researchers adopted statistical segmentation algorithms for color images combined with spatial information to segment color images (Comaniciu 2002; Lee 2002; Horiuchi 2006; Kang 2008; Jain 2010; Panda 2007; Sengur 2011; Tu2002; Yang 2012). Deng and Manjunath (2001) approach relied on quantizing color space followed by spatial segmentation. Furthermore, Ozden, and Polat (2007) employed color, space and texture features to segment color images. Chaabane, Sayadi, and Fnaiech, (2010) used fuzzy homogeneity histogram thresholding and data fusion to segment colored images. Also, Subakan, and Vemuri (2011) proposed a quaternionic Gabor filter that aims to combine color channel with spatial information. Then, they utilized this filter to segment images.