An Evolutionary Approach to Image Denoising Using A Regularized L1 TV Model

Main Article Content

Annie Lyn T. Oliveros
Marrick C. Neri


Total variation models are effective and popular in image reconstruction. In many papers a variation model with L2 fidelity term was
introduced and shown to be capable of removing Gaussian noise. For images corrupted with impulse noise or outliers, the total variation model
with L1 fidelity term exhibit good properties in restoring noise free pixels and in preserving contrast. However, this model is nonstrictly convex
and nondifferentiable. Another research work proposed a regularized version of the L1 model and an efficient semismooth algorithm which
involves second order information was presented to solve the discretization of this model. This paper deals with denoising images corrupted with
impulse noise using an evolutionary approach. Specifically, the Genetic Algorithm (GA) is employed to optimize the regularized L1 model.
Numerical results show the capability of GA in reconstructing n x n noisy images, with n = 256.

Keywords: Image Processing, Genetic Algorithms, Image Denoising, Impulse Noise Removal, Evolutionary Algorithm


Download data is not yet available.

Article Details