Multilevel Image Segmentation Using BDSONN Architecture Assisted by Quantum Inspired ACO

Image segmentation has long been an important focus of researchers, yielding many automated huge time consumed procedures to perform segmentation. The bidirectional self-organizing neural network (BDSONN) architecture, assisted by multilevel sigmoidal (MUSIG) activation function used for efficiently segmenting gray scale images into multilevel segmented images. To remove the bottleneck of heuristic class responses, generated by MUSIG activation function, an optimized version of the same, the OptiMUSIG activation function has been proposed. An attempt has been made in this article, to reduce the time complexity of the generation of the optimized class responses of the OptiMUSIG activation function using a quantum inspired ant colony optimization technique (QIACO). Experimental results of the proposed approach are presented on two real life gray scale images and one pixel intensity based brain MRI image with eight classes. Comparative study with the classical ACO reveals that the QIACO based multilevel segmented images show significantly better performance over its conventional counterpart.