Image Representation and Recognition Based on Directed Complex Network Model
Image structure representation is a vital technique in the image recognition. A novel image representation and recognition method based on directed complex network is proposed in this paper. Firstly, the key points are extracted from an image as the nodes
- PDF / 2,231,062 Bytes
- 9 Pages / 439.37 x 666.142 pts Page_size
- 27 Downloads / 227 Views
Abstract Image structure representation is a vital technique in the image recognition. A novel image representation and recognition method based on directed complex network is proposed in this paper. Firstly, the key points are extracted from an image as the nodes to construct an initial complete undirected complex network. Then, the k-nearest neighbor evolution method is designed to form a series of directed networks. At last, the feature descriptor of the image is constructed by concatenating the structure features of each directed network to finally achieve image recognition. Experimental results demonstrate that the proposed method outperforms the traditional methods in image recognition and can describe the structure of images more effectively.
Keywords Image recognition Directed complex network Dynamic evolution k-nearest neighbor Feature extraction
1 Introduction Image representation plays a vital role in computer vision and image processing. There are various categories of image representation. Texture features-based methods [1, 2], edge and contour-based methods [3, 4], and the key points-based Y. Chen J. Tang (&) B. Luo School of Computer Science and Technology, Anhui University, Hefei 230601, China e-mail: [email protected] Y. Chen e-mail: [email protected] B. Luo e-mail: [email protected] J. Tang B. Luo Key Lab of Industrial Image Processing and Analysis of Anhui Province, Hefei 230039, China
Z. Yin et al. (eds.), Proceedings of The Eighth International Conference on Bio-Inspired Computing: Theories and Applications (BIC-TA), 2013, Advances in Intelligent Systems and Computing 212, DOI: 10.1007/978-3-642-37502-6_115, Springer-Verlag Berlin Heidelberg 2013
985
986
Y. Chen et al.
methods [5–7] are some of them. The key points-based methods are one of the most popular categories owing to its lower computational complexity. However, it is still a challenge to the traditional structural description methods due to the ubiquitous image noise which will result in the extraction error of key points [5, 7]. Recently, complex network theory is becoming more and more attractive [8, 9] and has been widely applied in pattern recognition. Backes et al. [1] analyzed the image texture features and established a complex network model to achieve the identification of texture image; Backes et al. [3, 4] established a complex network model on the shape contour edge points to achieve the shape feature description. In addition, Tang et al. [10] achieved the feature representation of structure graphs using complex network model. However, the above models were all established based on the undirected graph. In fact, the directed graph often contains more abundant structural information. Therefore this paper gives a novel directed complex network representation model. Furthermore, the image feature extraction will be conducted through the complex network features under the k-nearest neighbor evolution. Benefited from the statistical features, this image feature description has the advantages of fav
Data Loading...
