Sasanian coins classification using discrete cosine transform

Cultural heritage is a real-world application which has attracted attentions from pattern recognition community in recent years. Ancient coins classification is a branch of the application in which high similarity between coins classes as well as high variability within each class make it a challenging pattern recognition problem. In this paper, a method is proposed for addressing this problem. For this purpose, after preprocessing coins images, discrete cosine transform is applied in order to extract most discriminative features so as to be used as inputs of a support vector machine classifier. A database containing 570 images related to both sides of Sasanian coins is used in the conducted experiments. The effect of different number of DCT coefficients in classification rate of the proposed method is studied. Moreover, the performance of the proposed method is compared with Fourier-based and wavelet-based classification. The obtained results suggest that the proposed method outperformed the two other methods with a classification rate of 86.2%.