Authors : Saleh Mesbah, Adel A. El-Zoghabi and Mohamed H. Obaid
Abstract: The denoising methods are an important subsystem of any signal processing system used for image enhancement because these methods remove undesirable signal components from the signal of interest. Noise suppression can introduce artefacts or cause image blurring which makes image denoising a complex task. Several image denoising approaches have been investigated in the literature; however, removing noise from images is still a challenging problem as it designed only for a convinced kind of noise and image or requires statistical properties of the corrupting noise. The aim of this research is to propose an optimal image denoising model using niching genetic algorithm by taking into consideration the most discriminative descriptors (features for uncorrupted and corrupted pixels) which can greatly improve the model performance. The suggested model demonstrates that suitable use of Rank-Ordered Logarithmic Difference (ROLD) can play a key role in determining more noisy pixels with less false hits. ROLD serves as an important definition of pixels and distinguishes between noise pixels and normal pixels (noise and normal pixels signature). Based on some statistical measurements of uncorrupted group of pixels, niching GA is utilized to smooth the corrupted pixels depending on PSNR fitness function between the associated group of pixels. In the suggested system, clearing based niching procedure is adapted to force the GA to maintain a heterogeneous population throughout the evolutionary process, thus, avoiding the convergence to a single optimum. Niching method has been employed to minimize the effect of genetic drift resulting from the selection operator in the traditional GA in order to allow the parallel investigation of many solutions in the population. Experimental results indicate that proposed model has a better performance on PSNR and a stronger capacity of preserving the details than previous denoising techniques.
Saleh Mesbah, Adel A. El-Zoghabi and Mohamed H. Obaid, 2020. An Adaptive Image Denoising Model Based on Rank-Ordered Logarithmic Difference and Niching Genetic Algorithm. Journal of Engineering and Applied Sciences, 15: 1743-1758.