Anisotropic Diffusion Filtering Based Gray-value Morphological Operator Processes
Wenyuan Wang · Fudan xuebao. Ziran Kexue ban · 2004
Current regularization formulations of anisotropic diffusion denoising processes are unable to prevent degradation of edges and corners. It introduces a novel algorithm that applies gray-value erosion and gradual regularization scale in anisotropic diffusion filtering. It is capable of denoising efficiently while maintaining edge and corners locality. Another idea is to replace central difference gradient by morphological gradient in diffusion coefficient discretization, the latter applying in anisotropic diffusion can enhance one-pixel line feature in noisy image while the former can not process efficiently. Results are presented that demonstrate their efficacy in heavy noisy image denoising and single pixel line enhancing tasks respectively.