REMOVAL OF PERIODIC NOISE FROM DIGITAL IMAGES USING PARTICLE SWARM OPTIMIZATION (PSO) AND GREY WOLF OPTIMIZATION (GWO) ALGORITHMS COMBINED WITH ADAPTIVE FILTERS

Abstract

Here, a hybrid approach is introduced to remove periodic noise in digital images based on the combination of Particle Swarm Optimization (PSO) and Grey Wolf Optimization (GWO) algorithms and adaptive filtering techniques. The proposed new method, called PSO-GWO Adaptive Filter (PGA-F), attempts to adaptively tune the filter parameters to enhance the image quality with a preservation of fine details and edges.

At the experimental phase, a library of 200 grayscale and color images (256×256 and 512×512 resolutions) with different periodic noise patterns corrupted between 10 Hz and 60 Hz and amplitude ranges of 5–25 dB was employed. The performance of the proposed algorithm was compared against conventional filters such as Wiener, Median, and Gaussian, as well as optimization-based techniques such as PSO-only and GWO-only methods.

Quantitative performance indexes like Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSIM), and Mean Square Error (MSE) were employed. The PGA-F scheme provided a mean PSNR of 38.72 dB, a 9.2% and 7.0% improvement over PSO-only (35.45 dB) and GWO-only (36.18 dB) filters, respectively. The mean SSIM was boosted from 0.914 (PSO) and 0.921 (GWO) to 0.948 using the combined scheme. Moreover, MSE values were reduced by 28.5% compared with conventional methods.

Visual results confirm that the hybrid algorithm effectively removes periodical artifacts without sacrificing image acuteness and structural integrity. The per-image processing time of 512×512 images was approximately 1.48 seconds on a typical Intel Core i7 processor with 16 GB RAM, which is acceptable for real-time image improvement applications.

Overall, the proposed PSO-GWO Adaptive Filter is better in convergence speed, accuracy, and robustness for periodic noise removal and therefore is a worthy competitor for real image processing applications in remote sensing, medical imaging, and industrial vision systems.