On Discrimination between Photorealistic and Photographic Images

Abstract

This paper presents a classifier built for differentiating digital photorealistic images from digital photographs. Results show that our 144-dimensional (144-D) feature vector extracted from characteristic functions of wavelet histograms is more efficient than Lyu and Farid's 216-D feature vector (S. Lyu and H. Farid), and Ng et al.'s 192-D feature vector (2005). Our classifier outperforms Lyu and Farid's state-of-art method while only requiring half of their feature extraction and testing time