ADAPTIVE PERONA–MALIK FILTERING FOR ULTRASOUND SPECKLE REDUCTION USING LOCAL ENTROPY
Ключевые слова:
ultrasound imaging, speckle noise, anisotropic diffusion, Perona–Malik model, entropy, adaptive filtering, medical image processingАннотация
This paper addresses speckle noise reduction in ultrasound (EchoKG) imaging, which strongly impacts diagnostic accuracy. An entropy-driven adaptive extension of the Perona–Malik anisotropic diffusion model is proposed, where the diffusion coefficient is controlled locally using entropy and variance. This enables stronger smoothing in homogeneous regions and reduced diffusion near edges and tissue boundaries, thus preserving anatomical structures. Experiments on echocardiographic B-scan images show that the method outperforms the classical Perona–Malik model by 3–4 dB in PSNR and 0.06 in SSIM, achieving a better balance between noise suppression and edge retention. The results indicate that the proposed entropy-based adaptive diffusion improves visual clarity, preserves myocardial and valve contours, and enhances diagnostic reliability. Overall, the method provides a robust and computationally efficient filtering framework for real-time medical imaging and intelligent diagnostic systems
