Reachability Distance-Based Vector Median Filter: A Novel Method for Image Denoising
Bogdan Smołka, Milena Smolka, Damian Kusnik · 2025
This paper proposes a generalization of the Vector Median Filter (VMF) for suppressing impulsive noise in color images. The new filtering approach is based on the reachability distance, which enables the incorporation of pixels similarities within their local neighborhoods. The structure of the proposed technique is inspired by the Sharpening VMF (SVMF), which effectively suppresses impulsive noise while enhancing image edges. The proposed method further improves robustness against noise corruption while maintaining edge-sharpening capabilities. Experimental results demonstrate that the new filter significantly outperforms both the VMF and SVMF in terms of objective quality metrics. Visual assessments confirm the efficient removal of noisy pixel clusters without introducing color artifacts. The design is computationally efficient, making it suitable for preprocessing heavily corrupted color images. Additionally, it provides a flexible framework for developing filter families utilizing reduced ordering concepts.