A Hybrid Denoising–Edge Detection Framework for Enhanced Medical Image Analysis
Tanusree Saha1, Kumar Vishal2
1Tanusree Saha, Research Scholar, Department of Information Technology, Magadha University, Bodh-Gaya (Bihar), India.
2Dr. Kumar Vishal, Department of Information Technology, Magadha University, Bodh-Gaya (Bihar), India.
Manuscript received on 20 August 2025 | First Revised Manuscript received on 03 September 2025 | Second Revised Manuscript received on 06 September 2025 | Manuscript Accepted on 15 September 2025 | Manuscript published on 30 September 2025 | PP: 1-6 | Volume-13 Issue-10, September 2025 | Retrieval Number: 100.1/ijese.C829614030925 | DOI: 10.35940/ijese.C8296.13100925
Open Access | Editorial and Publishing Policies | Cite | Zenodo | OJS | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: High-quality medical imaging is indispensable for precise diagnosis and treatment planning, but it is confounded by noise and ill-defined edges that impede diagnostic consistency. Noise originating during acquisition, transmission, or reconstruction tends to obliterate delicate structures but makes existing edge detection schemes incapable of producing edge-line sharp and smooth where noisy features abound. It is imperative to resolve such complications to enhance clinical image interpretability and facilitate the development of sophisticated computer-aided diagnosis systems. This research presents a Hybrid Denoising–Edge Detection Framework that merges stateof-the-art filtering and localization of edges to enable better analysis of medical images. An early stage adopts a hybrid filtering methodology that combines Adaptive Median Filtering and Block-Matching and 3D (BM3D) filtering. Adaptive Median Filtering efficiently eliminates speckle and impulse noise while preserving edge detail effectively, and BM3D reduces additional Gaussian noise using collaborative patch-based denoising. Both utilise superior noise resilience without loss of structural fidelity. A hybrid edge detection methodology, combining the Sobel operator and the Canny detector, is incorporated into the final stage. Sobel operates effectively in localising strong gradient variations, while Canny applies non-maximum suppression and hysteresis thresholding to achieve optimal edge continuation. An edge fusion process gains both strategies’ complementary strengths to produce sharper and more consistent edge maps. Experiments were carried out on MRI, chest X-ray, and mammography images corrupted by simulated Gaussian, saltand-pepper, and speckle noise. Quality was compared quantitatively using Peak Signal-to-Noise Ratio (PSNR) against noise suppression, Structural Similarity Index (SSIM) against structural preservation, and Edge Connectivity Ratio (ECR) against continuity of edges supported by quality visual assessment.
Keywords: Medical Image, Canny, Sobel, BM3D (Block Matching and 3D Filtering), Otsu, Adaptive Median.
Scope of the Article: Information Technology