Robust Mammogram Denoising Under Extreme Impulse Noise via FLBMF ‐Selective Median‐ TV Hybrid Scheme
Benard Nyangena Kiage, Wilson Kipchirchir Cheruiyot, Michael Waema Kimwele · Engineering Reports · 2025
ABSTRACT Critical for the early detection and precise diagnosis of breast cancer, mammogram imaging risks encountering extreme impulse noise which, in some cases, occurs at levels between 60% and 90%. When present, this guarantees that the resulting image is quite unlikely to be of diagnostically reliable quality. This then compromises its overall efficacy. Traditional denoising methods certainly do not completely suppress this noise nor perform the preservation of vital mammogram image details. In this paper, we propose a robust hybrid denoising framework that integrates a Fuzzy Logic‐Based Mean Filter (FLBMF) with Selective Median filtering and total variation (TV) smoothing. The approach is implemented in a streaming, row‐buffered architecture using integer‐only operations, enabling deployment in resource‐constrained environments such as smartphone imaging pipelines and embedded medical devices. Evaluating our method on mammogram image datasets corrupted with extreme impulse noise demonstrates that the proposed method outperforms seven classical and state‐of‐the‐art denoising techniques in PSNR, MSE, SSIM, VIF, and Pratt's FOM, with an average PSNR gain exceeding 1 dB over the next best method. Runtime noise density analysis shows competitive processing speeds with minimal quality tradeoff, and ablation studies confirm the benefit of selective switching. The method's efficiency and robustness make it suitable for both clinical diagnostic systems and real‐time consumer imaging applications.