<![CDATA[Salt-and-pepper noise severely degrades the visual quality of digital images, par ticularly at high noise densities where conventional denoising techniques often fail. Median- and mean-based filters tend to oversmooth images and blur fine structures when the majority of pixels within a local window are corrupted. This paper proposes a robust dual-layer denoising framework for grayscale images that integrates rank-based prescreening, interquartile range (IQR)-based statis tical outlier detection using Tukey fences, and a lightweight post-processing sharpening stage. In the first layer, a rank-4 trimmed estimator suppresses ex treme impulse values and stabilizes local statistics. In the second layer, adap tive IQR thresholds are employed to detect and replace residual outliers, even in heavily corrupted neighborhoods. A final step involving selective sharpen ing combined with mild smoothing enhances edge details without amplifying residual noise. Extensive experiments on standard grayscale images (Lenna, Barbara, lake, boat, and living room) across salt-and-pepper noise levels from 10% to 90% demonstrate that the proposed approach consistently outperforms conventional methods, including mean, median, Gaussian, modified decision based unsymmetrical trimmed median filter (MDBUTMF), and pixel density based filter (BPDF). Quantitative evaluation indicates peak signal-to-noise ratio (PSNR) values reaching 38.23dB, structural similarity index (SSIM) values up to 0.99, and significant reductions in mean squared error (MSE), particularly at higher noise densities. These results confirm that the proposed framework ef fectively suppresses noise while preserving edges and textures, making it well suited for practical applications such as medical imaging, remote sensing, and surveillance.]]>
Copyrights © 2026
