Abstract

The proposed system develops a novel image super pixel segmentation approach using energy optimization algorithm. This algorithm with self-loops has the merit of segmenting weak boundaries by the new global probability maps and the commute time strategy. But it is difficult to find and track exact contours of an object in case of complex images using energy optimization algorithm. In order to mitigate this problem and to improve segmentation quality, this paper presents a seeded region growing algorithm (SRG) along with energy optimization algorithm. In this paper, we discuss about popular seeded region growing methodology used for segmenting weak boundaries. This method use centroid calculation of the different regions appeared in an image and can withstand for almost each and every shape appear in the mage. This work is divided in to two stages, in first stage calculate the region of interest and then place the seed at the centroid of that region. In second stage region starts to grow from the initial seed until the homogeneity criteria satisfied. The experimental results have demonstrated that this method achieves better performance than previous approaches.