Segmentation of Noise Stained Gray Scale Images with Otsu and Firefly Algorithm

Background/Objectives: The major aim of thework is to propose an efficient multi-level thresholding for gray scale image using Firefly Algorithm (FA). Methods/Statistical Analysis: The multi-level image thresholding is attempted using Otsu's function and Firefly Algorithm (FA) using standard 512 x 512 sized gray scale image dataset. The robustness of the attempted segmentation process is tested by staining the test images with universal noises. The superiority of the FA based segmentation is validated with the heuristic algorithms, such as Bat Algorithm, Bacterial Foraging Optimization and Particle Swarm Optimization existing in the literature. Findings: The simulation result in this work conforms that, FA assisted segmentation offers better result compared to the alternatives. The robustness of the FA and Otsu based segmentation is also superior and offered improvedcost function, SSIM, PSNR value and reduced CPU time compared with the alternatives. Application/Improvements: In future, the proposed technique can be experienced using standard RGB images availablein the literature.