One Class Support Vector Machines for Distinguishing Photographs and Graphics

In recent research, distinguishing photographs and graphics was regarded as a typical two-class pattern classification task. In practice, photographs pattern model can be established easily, while it is hard to gain a perfect graphics pattern model. In this paper, we develop a novel approach based on one-class support vector machine (OCSVM) to distinguish photographs and graphics. In this approach, the image detection can be simplified as a one- class pattern problem. Firstly, the use of Gabor wavelet features with different orientation at different scale for texture analysis is proposed to get a feature vector and then support vectors are used to model photographs pattern, and graphics which are rejected based on this model. Finally the results of some automatic classifications are analytically compared. Extensive experiments show that in the only photographs model, C-SVC, nu-SVC, BP network become useless, while the proposed method has an encouraging performance.