A comparison of four PDE-spatial denoising systems for molecular images

Abstract

Partial differential equation (PDE)-based denoising methods are suitable to preserve image edges and describe intrinsic geometry. Besides, spatial filter based image denoising methods are suitable to describe texture while removing noise. In this work, we compare four hybrid filters that combine two spatial based denoising methods, namely the Wiener filter or the first order local statistics (FOLS), and the fourth-order PDE. The noisy image is processed by the Wiener or FOLS filter in the first step, and the result is further processed by PDE. Alternatively, the order of operations is reversed. Each hybrid denoising system was tested on three images corrupted with four types of noise at different levels: Gaussian, Poisson, Salt and Pepper, and speckle. Based on the peak-signal-to-noise ratio (PSNR) metric, the obtained experimental results show that Wiener-PDE denoising system offers the best performance.