Texture Classification Using Optimized Neural Networks Based on Firefly Algorithm

Texture classification is a process to category a texture image into its related class. Texture features can be extracted by different methods, using structural, statistical, model-based and transform information. In this work geometric invariant moments (GM) feature is utilized as a rotation, scale and translation invariant classifier. For omitting the image intensity of the features, thresholding technique is used before feature extraction. After that a new optimized neural classifier employed to classify the input images into their category. The classifier consists of optimizing the weights of neural network by a new algorithm, Firefly algorithm. Brodatz database are used to perform the experiment and final results show 90.18% classification rate as the system efficiency.

Texture Detection, Rotation, scale, translation and intensity invariant, Kapur's Thresholding, Artificial Neural Network, Optimization, Firefly Algorithm.