Image Splicing Detection Using Camera Characteristic Inconsistency

We propose a passive image splicing detection method based on consistency check of camera characteristics among different areas in an image. For a given image, we compute color sharpness from pixels and inter-channel singular value difference. Each color channel is independently down-sampled with a factor of two in the x and y directions. The down-sampled color image is then expanded by zero-padding, and interpolated to produce an estimate of the CFA-filtered version of the image. Difference of singular values between the interpolated image and the given image is computed. Color sharpness and singular value difference are fed into a statistical classifier to accomplish the image authentication. Experiments on a set of 363 natural and spliced images show satisfactory accuracy of the method.