Abstract: For the structure characteristics usually become instable in traditional graph based image representation methods, a novel image representation and recognition method based on complex network is proposed in this paper. Key points are extracted for an image and an initial complex network is constructed in which nodes correspond to the key points. A novel dynamic evolution process is devised for the initial complex network using the minimum spanning tree decomposition. The features of the networks in different evolution stages are extracted to finally achieve image structural information extraction. This method can simply describe an image by using geometrical feature of the image key points. Experimental results on both classification and clustering demonstrate that the proposed method outperforms the traditional edge weight threshold evolution method and it can describe the structure of images more effectively.
minimum spanning tree,