Multi-object splicing forgery detection using noise level difference

Splicing forgery is a commonly used operation in digital image synthesize. Exposing multi-object splicing forgery by detecting noise discrepancy is discussed in this paper. The image is firstly segmented into small segments and noise level function, which reveals relationship between image noise and pixels' intensity is estimated. Suspicious regions are detected by checking constrains of noise level function. In the experiment, a new dataset is used for evaluating the proposed method. The experimental results show the effectiveness and robustness in dealing with multi-objects splicing forgery. Besides, comparisons prove our method is superior to the existing state-of-art.