Quantum Modeled Clustering Algorithms for Image Segmentation

The ability to cluster data accurately is essential to applications such as image segmentation. Therefore, techniques that enhance accuracy are of keen interest. One such technique involves applying a quantum mechanical system model, such as that of the quantum bit, to generate probabilistic numerical output to be used as variable input for clustering algorithms. This work demonstrates that applying a quantum bit model to data clustering algorithms can increase clustering accuracy, as a result of simulating superposition as well as possessing both guaranteed and controllable convergence properties. For accuracy assessment purposes, four quantum-modeled clustering algorithms for multi-band image segmentation are explored and evaluated. The clustering algorithms of choice consist of quantum variants of K-Means, Fuzzy C-Means, New Weighted Fuzzy CMeans, and the Artificial Bee Colony. Data sets of interest include multi-band imagery, which subsequent to classification are analyzed and assessed for accuracy. Results demonstrate that these algorithms exhibit improved accuracy, when compared to classical counterparts. Moreover, solutions are enhanced via introduction of the quantum state machine, which provides random initial centroid and variable input values to the various clustering algorithms, and quantum operators, which bring about convergence and maximize local search space exploration. Typically, the algorithms have shown to produce better solutions.