Image Denoising Using Di Strategy Pelican Optimization Algorithm Based on Grayscale and Color images

Image denoising is the process of eliminating unwanted noise from an image by preserving significant information and structures.Its primary goal is to recover clean, visually coherent images and preserve the original content.However, conventional models face challenges in preserving image quality during denoising due to difficulties in effectively distinguishing noise from the true image, which reduces the efficiency of denoised images.This research proposes Di Strategy Pelican Optimization Algorithm (DS-POA) for denoising of Grayscale and color images.In traditional POA, Levy flight and Disturbance Inhibition factor are incorporated to avoid local optima issues with enhanced local and global search abilities.The DS-POA effectively balances global and local searches to ensure efficient image denoising convergence.The obtained images are resized to standardize the input dimensions, after which Gaussian noise is added to the resized Red Green Blue (RGB) images.Then, adaptive guided image filtering is applied or smoothing noise in flat regions, while preserving edges.Finally, DS-POA is used to determine the optimal degree of smoothing and the kernel size.DS-POA achieves commendable Peak-to-Signal-Noise-Ratios (PSNR) of 33.65 (dB), 37.65 (dB), 37.05 (dB), and 34.29 (dB), respectively for Berkeley Segmentation Dataset 68 (BSD68), Kodak24 Dataset, Color BSD68 Dataset (CBSD68), and Set12 Dataset, for a Gaussian noise of 25.This proves the proposed DS-POA model's superiority over the existing methods namely, Multi-scale Dilated Residual convolution Network (MDRN) and Neural Ordinary Differential Equations (NODEs)-ImgNet.