An Adaptive Detail Preserving Modified Progressive Pca Thresholding Based Image Denoising

Abstract

: Radiologists use medical imaging tools to assess the anatomy and function of interior organs. However, medical images are frequently distorted by various types of noise generated during the capture, transmission, and reconstruction operations, which can dramatically reduce image quality and hide critical diagnostic features. As a result, developing effective denoising algorithms that suppress noise while maintaining structural information remains a significant challenge in medical image processing. To address this issue, this work presents an adaptive cluster-wise progressive denoising algorithm for maintaining detail in medical images. The suggested method groups similar picture patches using adaptive clustering and then uses progressive Principal Component Analysis (PCA) thresholding guided by the Marchenko-Pastur (MP) law to eliminate noise while retaining significant image information. The denoising technique is further improved by employing a low-rank matrix representation to maintain the image's structural properties. The proposed approach is effective, as demonstrated by experimental results acquired using MATLAB on medical images damaged with Gaussian noise at various noise levels. The method's performance is measured using statistical metrics such as Peak Signal-To-Noise Ratio (PSNR), Mean Squared Error (MSE), and Contrast to Noise Ratio (CNR). The findings show that the suggested method outperforms existing denoising methods in terms of noise suppression while conserving critical anatomical characteristics.