Intrascale windowing-based cuckoo search–optimized sub-band thresholding for satellite image denoising

Abstract

Traditional satellite image denoising techniques can minimize Gaussian noise efficiently, but they fail to preserve image feature and may cause blurring effect after denoising. To overcome such limitations, this paper proposes a new intrascale windowing/dependency-based sub-band adaptive thresholding approach for feature-preserved satellite image denoising using a cuckoo search algorithm. In this framework, basically, a 4 × 4 window has been created for each coefficient to find the local parameters in each sub-band instead of the global parameter. Therefore, the most widely used nature-inspired swarm and evolutionary algorithms are exploited via the intrascale dependency method for learning the parameters of adaptive thresholding function. It has been examined that the proposed windowing-based adaptive thresholding with cuckoo search (CS) algorithm yields better edge information and reveals superior outcomes in terms of peak signal-to-noise ratio (PSNR), mean square error (MSE), structure similarity index (SSIM), and feature similarity index (FSIM) fidelity parameters as compared with exiting classical techniques which are corrupted by Gaussian noise. In this paper, several well-recognized traditional (Weiner, median, Bayes, and soft and hard thresholds) denoising approaches are compared with the proposed method.