Mixed Poisson-Gaussian noise reduction using a time-space fractional differential equations

Abstract

Images are frequently corrupted by various sorts of mixed or unrecognized noise, including mixed Poisson-Gaussian noise, rather than just a single kind of noise. In this work, we propose a time-space fractional differential equation to remove mixed Poisson-Gaussian noise. Combining fixed- and variable-order fractional derivatives allows us to maintain an image's high- and low-frequency components while eliminating noise. The current model, although primarily intended for mixed noise reduction, can indeed be utilized with great efficacy on images that have been solely degraded by Gaussian noise. In addition to this, a stable discretization strategy is presented. The illustrative results demonstrate that our scheme performs better than earlier models, reduces the staircase effect, and is applicable to electron microscopy and CT images.