Adaptive Switching Anisotropic Diffusion model for universal noise removal

In this paper, a novel method is presented for universal noise removal from corrupted digital images based on Adaptive Switching Anisotropic Diffusion (ASAD) model. The originality of ASAD is utilizing Local Difference Factor (LDF) to identify impulse noise or Gaussian noise. Initially, LDF is computed from intensity values of pixels in a neighborhood using weighted statistics. Subsequently, directional weighted median (DWM) and anisotropic diffusion (AD) are adopted to filter noise respectively. In addition, we use LDF to control the diffusion process adaptively incorporating with local gradient. As LDF indicates the local statistical property of image pixels, image edges and details can be finely preserved while filtering out noise. Simulation results show that the restored images by our method have high peak signal-to-noise ratio and great image quality by efficiently removing salt-and-pepper noise, uniform impulse noise, Gaussian noise and mixed noise.