A cross entropy and whale optimization algorithm based image segmentation for aerial images

There are several image segmentation techniques for image processing such as, thresholding, region based, edge based, watershed etc. However, the multilevel thresholding based image segmentation is the most widely used approach. This approach mostly suffers high computational complexity with the increasing number of threshold levels. The researchers have been using optimization techniques to resolve this issue. This paper presented a multilevel thresholding technique for segmenting aerial images using Whale Optimization Algorithm (WOA). It uses Minimum Cross Entropy (MCE) as an objective function for selecting a set of optimum threshold values. The presented method is evaluated by using a set of standard aerial images. The evaluated performance of the proposed technique is compared with Bacterial Foraging Optimization (BFO), Moth-flame Flame optimization (MFO) and Particle Swarm Optimization (PSO). The overall performance and the quality of segmented images are evaluated based on the following parameters, Average Mean Structural Similarity Index (MSSIM), Average Feature Similarity Index (FSIM), Average Peak Signal to Noise Ratio (PSNR), average mean objective functions values and average CPU rum time values. In terms of quality of the segmented images based on MSSIM, FSIM and PSNR values the presented method performs better than other methods and we found the average CPU run time also to a considerable range.