An Approach for Learning “Tuned” Mask based on Cuckoo Search

• 1、School of Remote Sensing and Information Engineering, Wuhan University, Wuhan, 430079

Abstract：Texture analysis is important in many applications of computer image process and analysis and is a hot research in these fields．Many texture extraction methods haven been presented, among them the texture energy analysis based on "Tuned" mask is an efficient and sample method; however, the original method of "Tuned" mask suffers from a lot of calculation and falling into local optimal. In order to overcome defects of traditional method for producing texture "Tuned" mask the paper introduces the principle of producing texture "Tuned" masks with Cuckoo Search algorithm; illustrate how to train "Tuned" masks with the proposed method in details. In final, the experiments show that the texture classification using the new approach has satisfying results. In addition, the paper also employs particle swarm optimization to learn "Tuned" mask, by comparing experimental results form the proposed method and PSO, it demonstrates that the proposed method outperforms PSO with respect to learning "Tuned" mask and is efficient and robust in the practical application.

Keywords： Cuckoo Search “Tuned” Mask Texture Image Feature Extraction