SUPERPIXEL BASED K-MEANS CLUSTERING ON CHEMICAL BONDING STRUCTURE ANALYSIS

Identification and segmentation of chemical bond images is important in computer aided diagnosis. In this paper, two chemical bonds were secured for superpixel based linear clustering, is proposed. In the clustering, histograms from contrast enhanced image channels and centre surround statistics from centre surround difference bonds are proposed as features to determine each superpixel as information or bonded structures. In the initial step, to handle the unbalanced cluster issue due to the presence of unbalanced images, bootstrapping is done. The proposed method has been tested on two bonds namely propanol and methoxyethane images with its boundaries marked by image processing techniques. The results also show a decrease in overlapping error as the reliability score is reduced. The method can be used in computer aided diagnosis systems to enhance the bonding structure for clinical deployment of the automatic segmentation.