USM Sharpening Detection Based on Sparse Coding
Yuzhou Gu, Shilin Wang, Xiang Lin, Tangfeng Sun · 2016
USM sharpening is one of the most widely used sharpening methods. The detection of USM sharpening has attracted much concern in image forensics. Previous research has demonstrated that USM sharpening will cause overshoot artifacts along image edges. In order to detect such artifacts accurately and comprehensively, a sparse-coding based local feature representation is proposed in this paper. The K-SVD algorithm is adopted to build an over-complete dictionary characterizing the local textures along image edges. Then the sparse code for each local patch is calculated by the Orthogonal Matching Pursuit (OMP) algorithm and the overall feature for the image is constructed by max-pooling over local features. Experimental results have demonstrated the superior performance of the proposed approach compared with some state-of-the-art methods investigated.