A hybrid approach for texture classification based on complex network

This paper proposes a method for image texture classification based on a complex small world network model. Finding important and valuable information in the context of an image is a big problem for image classification. In current analysis methods, image texture features such as spatial information are left out and color histogram is mostly used. In this article, the multi-radial distance analysis method is used to select the nodes, and then based on the identified nodes and calculating the shortest distance between adjacent nodes, a complex network is created to record the texture pattern. In the next step, the topological characteristics of the network such as the number of nodes, the number of triple triangles, the length of the edges and the average diameter of the network are identified and used for evaluation. Brodatz, UIUC and Outex databases have been used to test the system. The results of the proposed method show that this method is effective for texture classification and improves the classification rate compared to the usual traditional methods.

Keywords:

Texture analysis , Texture classification , Dynamic complex network , Network similarity