Multilevel Thresholding for Coastal Video Image Segmentation Based on Cuckoo Search Algorithm

In the coastal video image segmentation, images are partitioned into land and sea classes, and each of these classes could have different segmentation qualities. In order to cope with variations in image quality and opaque areas, this paper has proposed a multilevel threshold technique based on the Cuckoo Search (CS) algorithm as an optimization algorithm for selecting optimum threshold values. The optimum threshold values are determined by maximizing Otsu's or Kapur's objective function using CS algorithm. The CS algorithm uses McCulloch's method for Levy flight generation and combined with Otsu's and Kapur's objective functions to analyze CS algorithm performance. Based on the evaluations of PSNR, MSE, FSIM and CPU time parameters, the McCulloch's method based on CS algorithm with Otsu's objective function is the most promising and computationally efficient for segmenting coastal video images.