Screen content image quality assessment using curvelet transform

Screen content images (SCIs) are gaining widespread popularity due to the increase in computer processing power. Dissimilar to the natural images (NIs), SCIs are a mixture of texts, computer-generated graphics and natural images. Due to this reason, SCI and NI have different characteristics. Therefore, the quality assessment methods proposed for NIs are not suitable for assessing the quality of SCIs. In this paper, curvelet-based method (CurM-SCI) is proposed. Curvelet transform is used to extract edge features in CurM-SCI due to its superior directionality. CurM-SCI considers edge features from all orientations. However, most of the existing methods only deal with edge features from horizontal and vertical directions only. A new similarity equation which can handle negative values is also proposed to compare the coefficients of curvelet transform. Compared with the existing methods, CurM-SCI showed better performance.