Cellular Neural Networks Templates Learning Approach Based on Mutual Information and Firefly Algorithm for X-Ray Images de-noising

Sharaf, Ahmed I.; E. Abu El-Soud, Mohamed; M. El-Henawy, Ibrahim

doi:10.21608/mjcis.2019.321068

Cellular Neural Networks Templates Learning Approach Based on Mutual Information and Firefly Algorithm for X-Ray Images de-noising

Document Type : Original Research Articles.

Authors

¹ Faculty of Computers & Information Systems, Dept. of Computer Science, Mansoura University, Egypt

² Faculty of Computers & Information Systems, Dept. of Computer Science, Zagazig University, Egypt

10.21608/mjcis.2019.321068

Abstract

The Cellular Neural Network is a 2D array of analog processors which forms a parallel computing framework. The
main key factors in this model are the values of the neighborhood of each cell, which are called templates. These templates are usually set by a domain expert in this framework to determine the optimal values of the templates. In this paper, a novel approach was proposed to discover the templates of the cellular neural networks based on mutual information and firefly optimization. The mutual information discovers the hidden pattern in the templates by measuring the similarities among cells. The firefly algorithm navigates the search space to find the optimal values of the templates. The benchmarking and validation have been performed on the ChestX-ray8, which is a real-world X-ray images dataset. The proposed method achieved significant results when compared to other meta-heuristics algorithms such as Genetic Algorithm and Particle Swarm Optimization.

Keywords