Distinguishing Computer Graphics from Photographic Images Using Local Binary Patterns

摘要

With the ongoing development of rendering technology, computer graphics (CG) are sometimes so photorealistic that to distinguish them from photographic images (PG) by human eyes has become difficult. To this end, many methods have been developed for automatic CG and PG classification. In this paper, we explore the statistical difference of uniform gray-scale invariant local binary patterns (LBP) to distinguish CG from PG with the help of support vector machines (SVM). We select YCbCr as the color model. The original JPEG coefficients of Y and Cr components, and their prediction errors are used for LBP calculation. From each 2-D array, we obtain 59 LBP features. In total, four groups of 59 features are obtained from each image. The proposed features have been tested with thousands of CG and PG. Classification accuracy reaches 98.3% with SVM and outperforms the state-of-the-art works.