Coke microscopic image segmentation based on iterative grid clustering

Separting the coke microstructures from coke microscopic image is a crucial task for automatic recognition by using digital image analysis technology. This paper aims at improving the segmentation accuracy of coke microscopic image by integrating iterative grid clustering procedure into image segmentation algorithm. The proposed algorithm mainly consists of three stages: feature extraction, grid clustring and iterative optimization. At the first step, color features reflecting the difference of coke microstructures are extracted for coke microscopic image segmentation. At the second step, I₁I₂I₃ color space is divided into grid cells through grid division and coke microscopic image is segmented initially by mean shift vectors. Finally, iterative optimization algorithm is adopted for further image segmentation until coke microscopic image is segmented into several parts. Experimental results show that the proposed algorithm is effective for separating the coke microstructures, and offers a reliable foundation for automatic identification.